Description
Social Business Intelligence How and where firms can use social media data for performance measurement, an exploratory study
Faculty of Technology, Policy and Management
Social Business Intelligence
How and where ?rms can use social media data for
performance measurement, an exploratory study
Final
Joeri Heijnen
M
a
s
t
e
r
T
h
e
s
i
s
Social Business Intelligence
How and where ?rms can use social media data for performance
measurement, an exploratory study
Master Thesis
Empty
Joeri Heijnen
[email protected]
1320319
December, 2012
Faculty of Technology, Policy and Management · Delft University of Technology
The work in this thesis was supported by KPMG Advisory N.V. Their cooperation is hereby gratefully
acknowledged.
Copyright c Faculty of Technology, Policy and Management
All rights reserved.
Graduation Committee
Formal Chair First Supervisor
prof. dr. Y.H. (Yao-Hua) Tan dr. ir. G.A. (Mark) de Reuver
Full professor Assistant professor
Delft University of Technology Delft University of Technology
Faculty of Technology, Policy and Management Faculty of Technology, Policy and Management
Section Information- and Communication Technology Section Information- and Communication Technology
Second Supervisor External Supervisor
dr. M.E. (Martijn) Warnier ir. M.H. (Han) Horlings AITAP
Assistant professor Manager Business Intelligence
Delft University of Technology KPMG Advisory N.V.
Faculty of Technology, Policy and Management IT Advisory
Section Systems Engineering Business Intelligence
Abstract
Introduction
Both for individuals and for organisations the ?rst decade of the 21
st
century is characterised by the social
media trend. Social media platforms are increasingly popular, and are amongst others used by individuals
to express their opinions. Also ?rms acknowledge the opportunities o?ered by social media and are therefore
increasingly pursuing to realise their goals through means of social media (Murdough, 2009). The value of the
data produced on these platforms lies in the fact that consumers – i.e. (potential) clients – produce these data.
In addition, the information is created instantly, real-time and by many people. It is therefore not surprisingly
that Dey and Haque (2008) state that data generated from online communication acts as “potential gold mines”
for discovering knowledge.
Next, ?rms are increasingly hungry for information that reveals underlying trends and dependencies a?ecting
the ?rm’s performance. Business intelligence systems are used to obtain such insights (Lonnqvist & Pirttimaki,
2006). The demand for (real-time) business intelligence systems and the popularity of social media o?er room for
synthesis. Systems that are purposed to derive actionable information from social media to support managerial
decision-making are referred to as social business intelligence systems. Thus far, business intelligence systems
particularly derive management information from internal data. With the rise of a new data source – social
media platforms – the question rises how a ?rm should process these external data, what kind of managerial
information could be derived from the new data sources, and whether or not each ?rm is able to apply social
business intelligence. In business intelligence, indicators representing the strategy of a ?rm are established.
These indicators are termed ‘key-performance indicators’. Consequently, data re?ecting the performance of
di?erent processes are linked to these key-performance indicators.
Whereas links between social media data and key-performance indicators may leverage the opportunities of social
media for ?rms, a fundamental prerequisite allowing social business intelligence is the existence of user-generated
social media content. After all, user-generated content that does not exist can not be analysed. Thus, an
organisation is dependent for the generation of content on social media users and needs to determine whether
social media data exists before considering to invest in social business intelligence systems. So far, it is not clear
which organisational characteristics a?ect the existence of social media content. In this research, two general
characteristics describing a ?rm are used to investigate the existence of social media data; (i) industry type and
(ii) customer relation type.
Research Objective
On the one hand social media is a new phenomenon and acknowledged as a source of data of which valuable
information can be derived. On the other hand, it is unclear which ?rms are able to collect social media data
that is related to their ?rm and how ?rms should process these new data in accordance with existing business
intelligence processes. Therefore, the objective of this research has been formulated as:
The objective of this research is to develop a procedure to utilise social media data for business
intelligence, for which the applicability is investigated for ?rms in di?erent industries and for
di?erent customer relations.
Method
Our sample consists of social media messages related to eighteen di?erent ?rms, in seven di?erent industries
performing di?erent customer relations. Because the sample ?rms operate in di?erent industries and execute
iii
di?erent customer relations, it is possible to gain insight in potential di?erences between the social media
messages related to these ?rms. During a period of two weeks, social media messages from various platforms
have been crawled into a local database to allow further analyses. The content in the dataset is sourced from
Twitter, Facebook public pages, Flickr, Newssites, Google+ public pages, (Wordpress) Blogs, Picasa, YouTube
and Friendfeed. These platforms are popular in Western Europe.
To gain insight in the amount of ?rm-related social media messages, the average daily mentions of ?rms served
as a proxy to compare the volume of messages related to di?erent ?rms. Next, using a content analysis, a
portion of the collected messages have manually been classi?ed into di?erent categories based on the messages’
subjects. These categories correspond with generally applied categories of key-performance indicators. As such,
the results of the content analysis are directly linked to ?rms’ key-performance indicators, allowing to draw
conclusions on the relatedness of social media messages to di?erent key-performance indicators.
Incorporating the new external data source requires traditional business intelligence systems to be adjusted. A
social business intelligence procedure should be consistent with these traditional systems, and should additionally
consider the challenges involved when processing social media data. As such, the requirements for a social
business intelligence procedure have been established based on generally applied business intelligence concepts.
Furthermore, the challenges involved in the processing of social media data are discovered by the collection
of social media messages for the content analysis. Based on the traditional BI concepts and the challenges
discovered in the content analysis, a business intelligence procedure is developed. The procedure is veri?ed by
analysing its consistency with existing BI systems and its ability to solve the issues emerging when processing
social media data.
Results
The results of this research are twofold. Firstly, we gained insight in the applicability of social business
intelligence by investigating the existence and content of ?rm-related social media messages. Secondly,
a procedure to collect, process and analyse social media data for business intelligence purposes has been
established.
(ii) Applicability of social business intelligence
The applicability of social business is investigated on two facets. Firstly, the volume of ?rm-related social media
messages is investigated to obtain insight in the amount of data that is available for ?rms. The volume of
?rm-related social media content is however not su?cient to draw conclusions on the applicability of social
business intelligence. Therefore, the second facet on which the sample data is analysed relates to the content
of the social media messages. Especially, the subjects of the messages were analysed.
Volume
The average daily mentions di?ers from ?rm to ?rm. This implies that the applicability of social business
intelligence will not be possible for all ?rms, since not for each ?rm data is generated. Figure 1 illustrates the
average daily mentions of di?erent ?rms in our sample.
Heineken
Coca-Cola
Philips
3.000
3.500
TomTom
KLM
2.500
x
1
/d
a
y
]
1.500
2.000
A
v
e
r
a
g
e
D
a
ily
M
e
n
t
io
n
s
[
x
1
/d
a
y
Bol.com NS
Albert Heijn
C-1000
ABN AMRO Arcelor
Mittal
500
1.000
A
v
e
r
Bol.com NS
PostNL
Blokker
Aegon Unibail
Rodamco
AkzoNobel
Arcelor
Mittal
Arcadis Fugro
0
500
B2B B2C
Figure 1: Average Daily Mentions of Firms, Clustered per Customer Relation Type
Figure 1 shows the daily volume of ?rm-related social media content, in which the ?rms are clustered on their
customer relation type and consequently ordered descending. This ?gure suggests that B2C ?rms – coloured
iv Abstract
in red – are more likely to ?nd social media content that is related to their ?rm than ?rms performing B2B
relations (coloured in blue).
The second dimension on which the volume of ?rm-related social media content is investigated relates to
industries. Our sample consists of eighteen di?erent ?rms active in seven di?erent industries. As a ?rst step to
identify possible di?erences in the volume of daily messages between industries, the ?rms have been clustered
on industry type in ?gure 2, and have consequently been sorted in descending order.
Heineken
Coca-Cola
Philips
3.000
3.500
TomTom
KLM
2.500
x
1
/d
a
y
]
1.500
2.000
A
v
e
r
a
g
e
D
a
ily
M
e
n
t
io
n
s
[
x
1
/d
a
y
Albert Heijn
C-1000
ABN AMRO
NS Bol.com
Arcelor
Mittal
500
1.000
A
v
e
r
NS Bol.com
Arcelor
Mittal
Blokker
Aegon PostNL
AkzoNobel
Unibail
Rodamco Arcadis Fugro
0
500
Information&
Communication
Industry Transport &Storage Wholesale &Retail Financial Institutions Mining&Quarrying
Consultancy, Research
&Other Specialised
Business Services Communication Business Services
Figure 2: Average Daily Mentions of Firms, Clustered per Industry Type
Figure 2 strongly suggests that there exists a di?erence in the amount of user-generated content between di?erent
industries, with industrial ?rms being highly mentioned on social media, while consulting ?rms are the least
mentioned.
Subjects
Next to an assessment of the amount of social media posts that are created on the web, this thesis examined
the subjects of the social media posts in order to link the messages to ?rms’ key-performance indicators. The
social media messages of the ?rms have been classi?ed into categories based on their subject. These categories
are based on ten categories of commonly applied key-performance indicators. Consequently, the collected social
media posts of the ?rms in the sample have manually been classi?ed into one of these categories.
Our analysis shows that the subjects of social media messages di?er from ?rm to ?rm. The majority of social
media messages related to ?rms (41%) express how the external stakeholders of a ?rm perceive the company. In
this thesis, such posts have been classi?ed as community posts. 18% of the social media messages in our dataset
contained the name of a ?rm, but did not contain any valuable information for the ?rm and have consequently
been assigned as unde?ned posts. About 11% of the social media messages relate to ?nancial results, which
consist of ?nancial performance discussions (5%) and stock related discussions (6%).
The content analysis of this research suggests that the subjects of social media messages related to B2B
?rms contain a higher percentage of short term ?nancial results, news and professionals related messages than
messages related to B2C ?rms. Unfortunately for B2B ?rms, such type of information is yet available internally.
Acquiring social media data to gain additional management information is therefore of less value for B2B ?rms.
Next, the analysis indicates that the social media messages related to B2C ?rms contain a higher percentage
of posts related to customer relations, product and service quality and product and service innovation than
messages related to B2B ?rms. It are these types of information that deliver additional value to the ?rm, since
this information is not available at ?rms internally.
In addition, the content analysis of this research suggests that the subjects of social media posts di?er between
industries, but that the majority of the subjects in each industry relates to community, i.e. social media posts
revealing how the community perceives the company. The results indicate that ?rms active in the information
& communication, ?nancial institutions and transport & storage industries are more subjected to social media
messages related to customer relations, while ?rms active in the mining and quarrying and consulting industries
will ?nd messages related to ?nancial performance.
v
(ii) Procedure for social business intelligence
Based on (i) traditional business intelligence frameworks and (ii) the experience we gained in collecting,
processing and analysing social media data in the content analysis, a social business intelligence (“SBI”)
procedure has been developed. Figure 3 schematically shows the social business intelligence procedure.
Reaction based on
social intelligence
Strategic
mapping of
KPIs
Reacting
Search terms
Action plan(s) to
respond to gained
intelligence
g
Collecting
Mapping
insights to
business
units
Data pre
Unstructured
data
Information for
business units
Data pre-
processing
Categorising
Analysing
Structured, combined
(and anonymised) data
Categorised data
Figure 3: Blueprint: Social Business Intelligence Procedure
Our SBI procedure consists of seven main components, being (i) strategic mapping of KPIs, (ii) collecting,
(iii) data pre-processing, (iv) categorising, (v) analysing, (vi) mapping insights to the business units, and (vii)
reacting. The seven steps can be interpreted as a cycle, i.e. the output of the last step in?uences the ?rst step.
The very ?rst step of social business intelligence sets the scene for the objects that are to be collected and
analysed. Namely, in the ?rst step the key-performance indicators that are to be measured by social media data
are selected. Not each type of KPI is to be measured by social media data since there does simply not exist
any related social media data to these types of KPIs. Firms should mainly focus on KPIs related to customer
relations, public image and – to a less extent – on product and service innovation when selecting KPIs that are
to be measured using social media data.
The second step of the SBI procedure relates to data collection. In contradiction to regular BI systems, the
data is to be sourced from external parties in social business intelligence. People create ?rm-related messages
on di?erent platforms, of which the vast majority of publicly accessible messages are created on Twitter. The
search terms that are used to ?lter out the content at which the ?rm is interested should be based on the social
KPIs selected in the previous step.
The social media data has been collected from multiple platforms which adhere to their own data format. The
di?erent format are to be combined into one uniform database, so that – in a later step – data analysis can be
applied on the complete dataset. Furthermore, the ?rm should select those attributes that are necessary for the
analysis, not each platform o?ers the same richness of attributes to a social media post. In addition, the data
should be anonymised to be in compliance with new Regulations regarding data privacy. Finally, spam – i.e.
social media posts that do not relate to the ?rm – should be removed from the collected data.
The data pre-processing step resulted in a structured database in which the social media messages from multiple
platforms are combined. In the categorising step, the messages are clustered on di?erent issues of interest,
depending on the ?rm’s subject of interest. E.g., messages related to certain products can be categorised, or
one can cluster the messages that are created by people with many followers, etc. Again, the criteria at which
the messages are categorised are determined by the selection of the social KPIs in the ?rst step.
So far, the collected data has not provided any insights. It is in this analysis step of the procedure where data
is transformed into information. The categories that were established in the previous step are analysed in this
step. For instance, sentiment analysis can be applied on the categories related to the ?rm’s products in order
to acquire intelligence related to customer experiences of the products. However, the most valuable intelligence
vi Abstract
is gained when social media data is related to internal data. For instance, the volume of social media messages
related to a certain product may be correlated with the sales volume of that product. It is in this phase of the
SBI procedure where such relations are explored.
In the ?rst step of the procedure, KPIs have been selected. These KPIs typically relate to a certain function of
the ?rm, and hence have an “owner”. The intelligence gained in the previous step relates to KPIs, and should
feed back to the owner of the KPI. Generally, it are the people in the ?rm that are responsible for the KPI who
are the ones that can reason how the KPI is in?uenced. Therefore, these people are the ones that can draft an
action plan in case the KPI needs improvement.
The ?nal step of the social intelligence procedures consists of the execution of the action plans that are developed
in collaboration with people from the business lines that are responsible for the respective KPIs. Actions on
the gained intelligence may involve revisions of internal processes or strategies, or external interventions such
as social media engagement.
Contents
Graduation Committee i
Abstract ii
Preface xi
1 Research Problem 1
1-1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1-2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1-3 Research Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1-4 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1-5 Coherence of Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1-6 Research Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1-6-1 Exploratory Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1-6-2 Description of Research Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1-6-3 Data Collection and Research Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1-7 Project Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1-8 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1-9 Scienti?c Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1-10 Societal Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1-11 Project Deliverable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2 Conceptual Frame of Research 15
2-1 Business Intelligence Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2-2 Registering the Right Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2-2-1 Strategy and Business Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2-2-2 Frameworks Supporting the Formulation of Performance Indicators . . . . . . . . . . . . . . . 17
2-2-3 Performance Measurement System Design Process . . . . . . . . . . . . . . . . . . . . . . . 20
2-2-4 Typology of Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2-2-5 KPI Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2-3 Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2-4 Sub Conclusion: How Business Intelligence is Applied . . . . . . . . . . . . . . . . . . . . . . . . . . 25
viii Contents
3 Research Domain 27
3-1 Web 2.0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3-2 Social Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3-2-1 Social Media Platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3-2-2 User-Generated Content on Social Media Platforms . . . . . . . . . . . . . . . . . . . . . . . 31
3-2-3 Current Applications of Social Media in Firms . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3-3 Social Business Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3-3-1 Current State of Social Business Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3-4 EU Legislation on Social Media Data Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3-4-1 What Firms are allowed to do with Public Data . . . . . . . . . . . . . . . . . . . . . . . . . 40
3-5 Sub Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4 Content Analysis 42
4-1 Theoretical Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4-1-1 Hypotheses Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4-1-2 Material to Investigate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4-2 Establishment of Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4-2-1 Operationalising the Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4-2-2 Determining the Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4-2-3 Description of the Measuring Period . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4-3 Pretest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4-3-1 Categories of Social Media Posts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4-3-2 Revised Taxonomy of Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4-4 Data Collection and Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4-4-1 Search Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4-4-2 Scraping Social Media Content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4-4-3 Data Cleaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4-5 Descriptive Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4-5-1 Channel Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4-5-2 Volume of Firm-Related Social Media Messages . . . . . . . . . . . . . . . . . . . . . . . . . 62
4-5-3 Subjects of Social Media Posts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4-6 Interpretation of the Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4-6-1 Volume of Social Media Posts related to Firms . . . . . . . . . . . . . . . . . . . . . . . . . 68
4-6-2 Subjects of Social Media Posts related to Firms . . . . . . . . . . . . . . . . . . . . . . . . . 69
4-7 Sub Conclusion: Social Media Posts that relate to KPI Categories and the Performance Prism
Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5 Blueprint of a Social Business Intelligence Procedure 72
5-1 Requirements Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5-1-1 Description of Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5-1-2 Requirements Check on Business Intelligence Concepts . . . . . . . . . . . . . . . . . . . . . 76
5-2 Social Business Intelligence Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5-2-1 Strategic mapping of KPIs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5-2-2 Collecting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5-2-3 Data Pre-Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5-2-4 Categorising . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5-2-5 Analysing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5-2-6 Mapping insights to Business Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5-2-7 Reacting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5-3 Veri?cation of Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5-4 Real-Time Social Business Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5-5 Social Business Intelligence versus Business Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . 85
5-6 Sub Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Contents ix
6 Conclusions & Discussion 90
6-1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6-2 Contributions to Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6-2-1 Methodological Innovation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
6-3 Implications for Practice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6-4 Re?ection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6-4-1 Twitter Scraper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6-4-2 If I had More Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6-4-3 Stepwise Description of Data Collection Process . . . . . . . . . . . . . . . . . . . . . . . . . 96
6-5 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6-6 Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6-6-1 Classi?er . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6-6-2 Social Media Posts Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6-6-3 The Real Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6-6-4 Case Study: Relations of Social Media Metrics and Key-Performance Indicators . . . . . . . . 98
A Performance Prism Perspectives and Key-Performance Indicators Categories 99
B Classi?cation of Social Media Posts 101
C Social Media Platform Distribution 108
D Descriptive Statistics of Social Media Post Categories 112
E Corporate Engagement 116
Bibliography 118
x Contents
Preface
Social media is a trend in the ?rst decade of this century, and the concept is increasingly incorporated in the
daily lives of people. Scepticism towards the new technology is losing support, and companies are aware that
the new trend cannot be denied. Though the new phenomenon is gaining attention in the scienti?c world, social
media was not yet part of the curriculum at my faculty. I am grateful that I was o?ered the opportunity and the
con?dence to dive in the rather unexplored world of research into social media, and explore the opportunities
for companies o?ered by social business intelligence.
First of all I would like to thank my graduation committee. My ?rst supervisor of Delft University of Technology,
Mark de Reuver, critically reviewed my work on a regular basis. I hereby thank Mark for his constructive
comments and suggestions for improvements, I experienced our meetings as pleasant and useful. Mark’s
experience in scienti?c research and knowledge of ICT and business models contributed to the quality of my
thesis. Martijn Warnier supervised my work as second supervisor from the Systems Engineering section. Martijn
indicated issues concerned with social media (data) that I did not think of in the ?rst place, for which I am
grateful. Harry Bouwman chaired my committee as professor from the ICT section. Harry contributed to this
thesis by critically reviewing my work and o?ering suggestions for improvements, which were mainly related to
scienti?c concepts. Thank you for these comments, the critical notions improved the level of this work.
I would like to express my gratitude to my supervisor at KPMG, Han Horlings. Thanks to weekly meetings with
Han I was driven to progress my thesis. I found a sparring partner to discuss especially business intelligence
related aspects of my work. Han, thanks for your time, contributions and coaching! In addition I would like to
thank all employees and co-interns of KPMG’s Business Intelligence department for their interest in my thesis,
their contributing opinions and ideas on the subject, for being challenging competitors during the karting event
and for the fun at the Amsterdam Parade last summer.
Amstelveen, December 2012
Joeri Heijnen
xii Preface
Chapter 1
Research Problem
1-1 Introduction
More and more, customers are using content sharing sites to express their opinions about almost anything, from
soccer matches to ?nancial statements of large corporations. Examples of platforms where these expressions
are shared to the world are blogs and forums, social network sites and wikis. In 2008, “75% of internet surfers
used social media” (A. M. Kaplan & Haenlein, 2010), and the usage of social media is not limited to teenagers.
Members of generation X, now 35–44 years old, are increasingly active on social media sites (A. M. Kaplan &
Haenlein, 2010). Anno 2012, people express how they feel, what they do, what they think of, and what they
intend to do in over 340 million daily Twitter posts (Twitter, 2012). The value of the information produced
on these platforms lies in the fact that consumers produce these data. In addition, the information is created
instantly, real-time and by many people. Since social media posts are often non-anonymous and directly linked
to a person, ?rm or brand, the content produced on social media platforms can be interpreted as an indicator
of people’s attitude towards a ?rm, product or service. The user-generated content is considered as a driver for
future sales by Dhar and Chang (2009), hence containing economic value for ?rms (Ghose & Panagiotos, 2010).
In the ?rst decade of the 21
st
century, business intelligence (“BI”) has evolved to one of the critical processes
for organisations to provide useful insight, to support decision-making, and to drive organisational performance
(Ramakrishnan, Jones, & Sidorova, 2012). According to Watson and Wixom (2007) BI has become a “strategic
initiative and is regarded as an instrument in driving business e?ectiveness and innovation”. For organisations,
it is increasingly important to quickly respond to changes in the environment (Gessner & Volonio, 2005).
Therefore, BI systems are required to contain a component that allows monitoring the real-time environment.
We de?ne such systems as ‘real-time BI’ systems.
From the above, we can derive two trends in the current business landscape:
(i) an increase in the usage of social media, and,
(ii) an increase in the usage of business intelligence systems.
Trend (i): Social Media in Organisations Organisations are increasingly pursuing to realise their goals
through social media (Murdough, 2009). Social media applications support organisations in creating value
in many of their activities, e.g. in marketing, services, human resource management and customer relationship
management (A. N. Smith, Fischer, & Yongjian, 2012). In addition, ?rms are able to acquire data from social
media at low costs. Dey and Haque (2008) state that data generated from online communication acts as
“potential gold mines” for discovering knowledge. It is therefore that this thesis focuses on the extraction of
information based on the data created by consumers on social media.
The increased application of social media has serious consequences for an organisation’s exposure to the actors in
their environment, which include (potential) customers, suppliers and competitors. It seems that the power has
been taken from the corporate marketing departments by individual consumers that create, share and discuss
online blogs, tweets, Facebook entries, movies, pictures, etc. (Kietzmann, Hermkens, McCarthy, & Silvestre,
2011). With or without permission from the organisation, communication about brands will happen. In an
2 Research Problem
environment where customers gain more and more power, an organisation needs to carefully treat its actions and
control its exposure. Therefore, companies empower employees to talk, listen, and respond to what consumers
post on social media (A. N. Smith et al., 2012).
Though many organisations acknowledge the opportunities in the application of social media, there also exists
a fair degree of uncertainty with respect to allocating marketing e?ort and budget to social media, and “limited
understanding” of the social media platforms (Weinberg & Pehlivan, 2011). Kietzmann et al. (2011) argue that
many executives avoid or ignore social media because they do not understand what it is, how to engage with it
and learn from it. This thesis contributes to a further understanding of social media and discovers opportunities
to leverage the valuable content on these platforms for business purposes.
Trend (ii): Business Intelligence in Organisations Business intelligence systems are applied to obtain a better
understanding of underlying trends and dependencies – often coming from the external context – that a?ect the
business (Lonnqvist & Pirttimaki, 2006). Whereas BI systems were initially perceived as tools that were used
exclusively to support strategic decision-making, organisations have recently commenced to further exploit the
capabilities of BI systems to support wider business activities (Elbashir, Collier, & Davern, 2008).
The scale of recent investments in BI systems re?ects the growing importance and highlights the need for more
attention in research studies. Elbashir et al. (2008) estimated that global spending on BI systems and related
products reached USD 6.1 billion in 2008. A paper by Gartner (2009) predicted that organisations will increase
spending on “packaged analytic applications, including corporate performance management (“CPM”), online
marketing analytics that optimise processes, not just report on them”. Azvine, Cui, and Nauck (2005) predict
that in the future, “business intelligence will be available to everyone in the enterprise, and will be embedded
in many business systems”.
1-2 Problem Statement
The demand for (real-time) business intelligence and the popularity of social media o?er room for synthesis.
The opportunities o?ered by linking both concepts are acknowledged in the literature, e.g by Dey and Haque
(2008) and Lovejoy, Waters, and Saxton (2012). However, search queries
1
related to the subject of this thesis
into the scienti?c databases ScienceDirect and JStore, and the search engine Google Scholar resulted in the
understanding that social media applications for BI purposes are relatively underexposed in the literature.
Generally, research in the area of social media is related to marketing activities, sales, promotions, public
relations and customer relationship management, e.g. by Dong-Hun (2010); Ratner (2003); Klassen (2009);
Kozinets, de Valck, Wojnicki, and Wilner (2010); Kirtis and Karahanb (2011); Hanna, Rohm, and Crittenden
(2011); A. M. Kaplan and Haenlein (2012); You, Xia, Liu, and Liu (2012). The focus of the research conducted
in the literature is mainly focused on the organisation expressing itself to the outside (social media) world,
whereas this thesis focuses on the incoming aspect. A reason for the shallow results discovered in the literature
may be the relatively new character of combining social media and business intelligence.
Zeng, Chen, Lusch, and Li (2010) distinguish social media research between social media analysis and social
media intelligence. Social media analysis is concerned with “developing and evaluating informatic tools and
frameworks to collect, monitor, analyse, summarise, and visualise social media data”. Social media intelligence –
on the other hand – “aims to derive actionable information from social media in context-rich application settings,
develop corresponding decision-making or decision-aiding frameworks, and provide architectural designs and
solution frameworks for existing and new applications that can bene?t from the wisdom of crowds through the
web”.
Many social media monitoring tools, like Socialmention.com, Radian6, RowFeeder, Trackur, uberVU, SAS
Social Media Analytics, Finchline, Sprout Social, etc. mainly reveal the performance of a ?rm on social media
(number of mentions, number of likes, % of positive mentions), and treat the social media component of a ?rm
as a separate business unit executing its own strategy. However, the purpose of business intelligence is to reveal
the underlying parameters that determine the performance of the organisation, that is, not limited to solely
social media performance. In order to understand the in?uence of social media content on a ?rm’s performance,
a link between the company’s key performance indicators (“KPIs”) and social media parameters is required
because KPIs measure the performance of an organisation with respect to its strategy. Some social media
1
(SOCIAL BUSINESS INTELLIGENCE), (BUSINESS INTELLIGENCE 3.0), (SOCIAL MEDIA) AND (ORGANISATION),
(SOCIAL MEDIA) AND (BUSINESS), (SOCIAL MEDIA) AND (BUSINESS INTELLIGENCE), (TWITTER) AND (BUSINESS
INTELLIGENCE), (WEB 2.0) AND (BUSINESS INTELLIGENCE), (SOCIAL MEDIA) AND (STRATEGY)
1-3 Research Objective 3
monitoring tools, like Kapow Software and ListenLogic seem – at a glance – to establish this link. Zeng et al.
(2010) highlight the need for clearly de?ned social media performance measures because much of the research
is conducted in a setting which aims to support decisions in organisations. We argue that the possibilities of
social media for business intelligence purposes reaches further than what is currently o?ered by the social media
analytics tools. This argument is supported by Reinhold and Alt (2011), who state that “existing tools still have
a limited functional scope”. The key bene?ts will be gained whenever the KPIs of an organisation are linked to
the parameters that are measured by social media tools. Only in that case, one can speak about ‘social business
intelligence’. This thesis contributes to a transition from social media ‘monitoring’ towards social ‘business
intelligence’.
Whereas links between organisational performance and social media content can leverage the opportunities
of social media for ?rms, a fundamental prerequisite allowing social business intelligence is the existence of
user-generated social media content. After all, user-generated content that does not exist can not be analysed.
Thus, an organisation is dependent for the generation of content on social media users and needs to determine
whether social media data exists before considering to invest in social business intelligence systems. However,
it is not clear which organisational characteristics a?ect the existence of social media content. The following
section illustrates which factors are to be considered when one tries to categorise the availability of social media
content that is related to ?rms.
Firstly, it is likely that within some industries users express their opinions more often than in other industries.
We expect that one expresses his or her opinion more often about a product that is purchased on a frequent
basis. For example, domestic products are purchased more frequent than a car or a house. Therefore, the
consumer industry is probably discussed more often than the real-estate market. Secondly, the relation with
end-users makes it that people discuss the company on social media, or not. Some ?rms are more visible
for consumers than others. Zhang, Jansen, and Chowdhury (2011) support this factor by concluding that
“business engagement on social media relates directly to consumer’s engagement with online word-of-mouth
communication”. When users experience malfunctions in a mobile network they complain at the ?rm at which
they signed the contract, while the ?rm that delivered the network equipment – which may be responsible for the
errors – remains una?ected. This example illustrates that it is necessary to make a distinction between companies
in the same industry based on their position regarding consumers. Turban, Lee, King, and Chung (1999) classify
e-commerce into either business-to-business (“B2B”), business-to-consumer (“B2C”), consumer-to-consumer
(“C2C”), consumer-to-business (“C2B”), non-business e-commerce, or intra-business e-commerce” (as cited in
Chen, Jeng, Lee, and Chuang (2008)). We will use this classi?cation to assign an organisation’s position
regarding consumers since it clearly illustrates how close an organisation acts to the end consumer. As such the
network service provider can be positioned as a B2C ?rm, while the provider of the equipment performs B2B
relations.
Next, in the case that there exists social media content, an organisation should be able collect and analyse the
data. The unstructured nature of the data, various languages, various data formats, interpretation di?culties,
unveri?ed information and privacy issues are aspects that make the usage of social media data for business
intelligence di?erent from ‘regular’ – i.e. internal management information – BI data.
Knowledge Gap From the previous, we can conclude the following. It is unclear in which industries and for
which type of customer relations ?rms can apply social media data for business intelligence. Secondly, there
is no understanding how organisations should process social media data in relation with business intelligence.
Taking into account the previous, the following knowledge gap is formulated:
It is unclear how ?rms can process social media data for business intelligence, and how the
applicability of social media data for business intelligence varies among di?erent industries and
di?erent customer relation types.
1-3 Research Objective
Social media is a new phenomenon, and increasingly popular for both consumers and organisations. Business
intelligence is applied in organisations to measure organisational performance and to provide managerial
information. The literature agrees that social media posts may contain valuable insights for organisations
that managers can use in their decision-making. Hence, the two concepts o?er room for synthesis. However,
there does not exist a structured procedure that prescribes how organisations should acquire and analyse these
4 Research Problem
social media posts in order to generate managerial information. In addition, it is unknown how (i) di?erent
industries and (ii) di?erent customer relations a?ect the existence of social media data on the web. After
all, if (potential) clients do not generate social media posts related to a ?rm, it will not be possible to derive
information from the posts. Therefore, the objective of this thesis is formulated as:
The objective of this research is to develop a procedure to utilise social media data for business
intelligence, for which the applicability is investigated for ?rms in di?erent industries and for
di?erent relations with end-users.
As such, insight in (i) the suitability of social media for business intelligence for di?erent organisations and (ii)
a procedure prescribing the steps required for social business intelligence is obtained.
Concepts in Research Objective In order to clarify the research objective, the key concepts are listed and
explained below.
• Procedure to utilise social media data for business intelligence
A procedure to utilise social media data for business intelligence prescribes which steps are necessary when
an organisation applies social media data for the measurement of organisational performance. Within business
intelligence procedures, managers endeavour to measure organisational performance based on metrics that re?ect
the performance of organisational activities. Generally, these activities are performed by di?erent departments.
In this thesis we look for performance metrics that are in?uenced by social media data.
• Social media data
Social media data can be quantitative or qualitative in nature. Examples of quantitative social media data are the
number of likes, views or shares of a certain page, the number of followers, friends or retweets through the course
of time. Qualitative social media data contains the text of the posts. In this thesis, we investigate how social
media data can be used for business intelligence.
• Business intel ligence
Business intelligence is a process in which information is derived from data to support decision making. The
acquired information is required to measure organisational performance, at which managers can base their decisions.
Information may for example relate to trends in the level of inventory of a certain product, or the amount of sales
in a certain period.
• Firm contexts
Though there are various ways to de?ne a ?rm’s context, we describe the context of a ?rm based on two dimensions
in this thesis: (i) industry and (ii) relation with end-consumers. We employ this de?nition of context in this thesis
because we are particularly interested in the variations of the applicability of social business intelligence on these
two dimensions. Next, a generic classi?cation of a ?rm’s context on these two dimensions allows the conclusions
of the research to be applicable at a broad range of ?rms.
i. Industry
Organisations can be classi?ed in industries. All organisations in the same industry deliver similar products
/ services. We apply CBS’ (2012) classi?cation to position ?rms in certain industries. Examples of industries
are the telecommunications industry, or the ?nancial industry.
ii. Relation with end-users
Each organisation has di?erent customers. Generally, a distinction between Business-To-Business (“B2B”)
and Business-To-Consumer (“B2C”) is made to described the relation with an organisation’s customer. In
B2C relations, the end-user is part of the relation.
1-4 Research Questions
From the research objective, the following main research question is formulated:
How can ?rms use social media data for business intelligence, taking into account the ?rm’s speci?c
industry and relationship with end-users?
1-5 Coherence of Research Questions 5
In order to describe the domain of this thesis, the ?rst sub question describes the current state of social media,
the role of business intelligence in ?rms and the developments towards social business intelligence. Therefore,
the ?rst sub question is formulated as:
1. What is the current state of social media in relation with business intelligence?
(a) What are social media?
(b) How are social media generally applied within ?rms?
(c) How is business intelligence generally applied within ?rms?
(d) How are key-performance indicators established within ?rms?
(e) How can key-performance indicators be categorised?
(f) What is social business intelligence?
The main research objective contains a component in which we reveal in which contexts – i.e. for which
industries and for which customer relation type – ?rms are able to acquire social media data, and in which not.
This objective follows from the fact that ?rms are dependent on the users of social media whether or not social
media data is available. Therefore, the second sub question investigates for which ?rms social media posts are
available, and to what subjects the posts are related. The subjects of social media posts are consequently used
to assign social media posts to the KPIs of a ?rm. The composition of the second sub question is twofold, sub
questions 2(a) and 2(b) are quantitative in nature and provide insight in the volume of social media posts. On
the other hand, 2(c) and 2(d) are qualitative in nature and provide insight in the content of the social media
posts related to ?rms. The second sub question is formulated as:
2. In which ?rm contexts
2
are ?rms able to acquire social media data for business intelligence?
(a) How does the volume of social media posts related to ?rms vary between di?erent industries?
(b) How does the volume of social media posts related to ?rms vary between di?erent relations with
end-users?
(c) How do subjects of social media posts related to ?rms vary between di?erent industries?
(d) How do subjects of social media posts related to ?rms vary between di?erent relations with
end-users?
Secondly, the research objective contains a component in which we describe how a ?rm can acquire and process
social media data for business intelligence purposes. The third sub question focuses on the development of a
procedure to process social media data so that it can be joined up in business intelligence processes. A key
requirement of this process is that it should ?t within existing business intelligence activities. Therefore, 1(c)
investigates how business intelligence is generally applied in organisations, and will result in requirements for
a procedure in which social media data is applied for business intelligence. As discussed, social media data
di?ers from data that is generally processed in BI systems. Question 3(a) discusses the potential problems and
pitfalls when processing social media data. Consequently, 3(b) provides solutions for these problems. In 3(c),
we determine how social media data can be linked to KPIs. Finally, 3(d) describes how a ?rm can process social
media data while following the generally applied BI approach. The third sub question is de?ned as:
3. Which processes are required to incorporate social media data into general business intelligence
frameworks?
(a) What problems arise when applying social media data for business intelligence?
(b) How can the problems discovered in 3(a) be tackled?
(c) How can social media data be linked to key-performance indicators?
(d) How can social media data be processed in accordance with general business intelligence systems?
1-5 Coherence of Research Questions
Each research question delivers information that is required to answer another question. The coherence of the
research questions is presented in ?gure 1-1. The arrows represent the output of a research question which, in
turn, serve as input to answer an other research question.
2
In this thesis, we de?ne a ?rm context based on the ?rm’s industry and customer relation type.
6 Research Problem
Requirements
for a social
business intelligence
procedure
Main Research Question
How can firms use social media data for business intelligence,
taking into account the firm’s specific industry and relationship with end-users?
Sub Question 2
2. In which contexts are firms able to acquire
social media data for business intelligence?
Sub Question 3
3. Which processes are required to incorporate
social media data into general business
intelligence frameworks?
Procedure prescribing how
to collect, process and analyse
social media data
How does the volume of
social media posts
related to firms vary
between different:
2. (a) industries?
2. (b) relations regarding
end-users?
Quantitative description
of the availability of
social media posts in
different contexts.
Qualitative description
of the subjects of social
media posts in different
contexts.
How do the subjects of
social media posts
related to firms vary
between different:
2. (c) industries?
2. (d) relations regarding
end-users?
1. (c) How is business
intelligence generally
applied within
organisations?
3. (d) How can social
media data be
processed in
accordance with general
business intelligence?
3. (a) What problems
arise when applying
social media data for
business intelligence?
Understanding of how
business intelligence is
applied in firms, and
what the consequences
are when adding
social media data in
this process
3. (b) How can the
problems discovered in
3(a) be tackled?
Understanding of the
pitfalls of social media
data
Solutions for the
pitfalls of processing
and analysing social
media data
Sub Question 1
1. What is the current state of social media in
relation with business intelligence?
Overview of available social media
data in different contexts
Knowledge about current
technologies related to
social business intelligence
3. (c) How can social
media data be linked to
key-performance
indicators?
Understanding of relations
between social media data
and KPIs
1. (a) What are social
media?
1. (b) How are social
media generally applied
within firms?
1. (d) How are key-
performance indicators
established within firms?
Understanding
how business intelligence
is applied in organisations
1. (f) What is social
business intelligence?
Definition of
social business
intelligence
Understanding of
currently exploited
opportunities offered
by social media
Understanding of
generally applied
KPIs in firms
Knowledge about social
media platforms
and the data that is
created on such platforms
1. (e) How can key-
performance indicators
be categorised?
Understanding of the
role of KPIs within
business intelligence
Figure 1-1: Coherence of Research Questions
1-6 Research Method 7
1-6 Research Method
This section describes the type of research (section 1-6-1), the research method (section 1-6-2) and the approach
of the research (section 1-6-3).
1-6-1 Exploratory Research
Exploratory research is conducted for a problem that has not been clearly de?ned. It relies on reviewing
literature and/or data. Often, the results of exploratory research are not usually useful for decision-making by
themselves, but they can provide signi?cant insight into a given situation. The goal is to learn “what is going on
there?”, and to investigate social phenomena without explicit expectations. Mainly, the purposes of exploratory
research are exploratory, descriptive and explanatory in nature. This thesis researches an area that is relatively
unexplored, and of which the functioning is not clearly documented in theories and frameworks. Therefore, this
thesis can be positioned under exploratory research.
1-6-2 Description of Research Methods
The research questions formulated in section 1-4 individually require di?erent research methods in order to be
answered. The research consists of a mix of literature studies, consulting experts, and content analysis on the
acquired data. All methods and the corresponding requirements for data and other resources are discussed in
this section. Table 1-1 schematically lists the corresponding research method for each research question.
Table 1-1: Research Questions versus research Methods
L
i
t
e
r
a
t
u
r
e
r
e
v
i
e
w
C
o
n
t
e
n
t
a
n
a
l
y
s
i
s
C
o
n
s
u
l
t
i
n
g
B
I
e
x
p
e
r
t
s
1. What is the current state of social media in relation with business intelligence?
(a) What are social media?
(b) How are social media generally applied within ?rms?
(c) How is business intelligence generally applied within ?rms?
(d) How are key-performance indicators established within ?rms?
(e) How can key-performance indicators by categorised?
(f) What is social business intelligence?
2. In which contexts are ?rms able to acquire social media data for business intelligence?
(a) How does the volume of social media posts related to ?rms vary between di?erent industries?
(b) How does the volume of social media posts related to ?rms vary between di?erent relations with end-users?
(c) How do subjects of social media posts related to ?rms vary between di?erent industries?
(d) How do subjects of social media posts related to ?rms vary between di?erent relations with end-users?
3. Which processes are required to incorporate social media data into general business
intelligence frameworks?
(a) What problems arise when using social media data for business intelligence?
(b) How can the problems discovered in 3(a) be tackled?
(c) How can social media data be linked to key-performance indicators?
(d) How can social media data be processed in accordance with general business intelligence systems?
Literature Review
Scienti?c articles are studied, mainly in the Journal of Electronic Markets, Journal of Information Systems
Management, Journal of Business Research, Journal of New Media & Society, Business Horizons, Journal of
8 Research Problem
Strategic Information Systems and the Journal of Computer-Mediated Communication. The literature review
was supported by books in the related research context. In addition, reports and white papers by acknowledged
consulting ?rms in the ?eld of information technology have been studied. The novel character of social media
and social business intelligence makes it that especially in these reports social business intelligence is mentioned,
whereas this term is less visible in the scienti?c area. These reports often contain examples from innovations
and practical experiences. A such, a variate overview will be presented about related research and theories to
this thesis.
Consulting Business Intelligence Experts
Firstly, interviewing experts contributes to an understanding of the actual situation of business intelligence in
organisations and the potential role of social media in this ?eld. This allows to scope the research in a topic
that is actual and relevant. Secondly, a part of the research will describe how business intelligence is applied
in organisations. Whereas this is mainly investigated using literature in the ?eld of business intelligence, BI
experts can validate the ?ndings. Thirdly, a procedure prescribing how to execute social business intelligence
will be develop. Such a procedure is required to be applicable in organisations as an integral part of the existing
– regular – BI process.
Content Analysis
Content analysis is appropriate for this research since it o?ers a systematic method to compare content for a
large sample of data. Content analysis is a research technique that can be used to identify what people are
sharing on social media. The research technique is described by Stephens (2012) as an “in-depth look at recorded
information” and as “a means of analysing texts” by Bos and Tarnai (1999). The sources of these texts can be
various, for example newspapers, articles, web sites, or – as in this research – social media posts. Neuendorf
(2002) de?nes content analysis as a “systematic, objective, quantitative analysis of message characteristics”.
As discussed, this thesis purposes to analyse the characteristics of social media posts, and link these posts to
organisational functions. Krippendor? (2004) states that a “content analysis entails a systematic reading of a
body of texts”, and argues that every content analysis requires the following six questions to be considered:
1. Which data are analysed?
2. How are they de?ned?
3. What is the population from which they are drawn?
4. What is the context relative to which the data are analysed?
5. What are the boundaries of the analysis?
6. What is the target of the inferences?
Bos and Tarnai (1999) provide a procedure for analysing content, which is schematically shown in ?gure 1-2. In
the ?rst step, the problem is formulated at the theoretical level, research questions are de?ned and the object of
investigation is determined. Secondly, the unit of analysis is de?ned by establishing categories and determining
the sample. The third step consists of pretesting the reliability of the data, and the validation of the categories
that were established in step 2. Discovered de?ciencies are consequently renovated. In the fourth step of the
content analysis procedure the data is collected and analysed. Finally, the results are interpreted and discussed
on the basis of the problem.
It is the stepwise approach of Bos and Tarnai (1999) that is applied on the content analysis of this thesis. We will
retrieve user-generated content from various social media platforms, store it into a database, and consequently
analyse the collected posts. By analysing the social media content, it is possible to classify the nature of the
content into categories, and ?nd di?erences between posts related to di?erent organisations.
1-6-3 Data Collection and Research Approach
Figure 1-3 illustrates the sequence and the links of the research steps in a schematic manner. A sample consisting
of several organisations across di?erent industries and with di?erent customer relation types will be established.
The selection of the organisations forms the point of departure for the collection of social media data. The
1-7 Project Scope 9
Research outline, research questions,
formulation of hypotheses, material to
investigate
Operationalising the categories,
determining the sample, determining the
unit of analysis
Establishment of categories
Theoretical level
Determining reliability and validating the
categories
Pretest
Appropriate statistical analyses
Data collection and evaluation
Immanent interpretation of the results,
discussion of the results on the basis of
the problem
Interpretation of the results
Figure 1-2: A procedure for analysing content (Bos & Tarnai, 1999).
content analysis requires that the social media data is available in a database. Therefore, social media posts
need to be loaded from the web into our database. This process is called scraping. The selection of the data
will be executed based on keywords corresponding to the selected organisations.
Scraping content from social media platforms results in unstructured data. In addition, the data is expected to
be polluted by e.g. spam or by users who apply nicknames related to the search terms used to scrape the content.
Therefore, the data needs to be cleaned before commencing the analysis. Once the spam and irrelevant posts
are removed from the dataset, the content analysis can start. In this analysis, social media posts are classi?ed
in relation to KPI categories based on the subject of the posts. Once the content analysis has been performed
for the ?rms, it is possible to identify di?erences between the subjects of social media posts across industries
and di?erent positions regarding end-users. Consequently, we can draw conclusions on the applicability of social
media for business intelligence purposes.
The third research question relates to how organisations should execute social business intelligence. For that
reason, a procedure prescribing how to execute social business intelligence will be designed. However, not
before the requirements of a social business intelligence systems are clear, the framework can be designed. The
framework is veri?ed by (i) BI experts and (ii) the ?t in the system that is currently executed in general BI
systems. Finally, conclusions are drawn regarding the applicability of social business intelligence in ?rms.
1-7 Project Scope
Business intelligence and social media are broad concepts. In order to describe the focus of the proposed research,
this section describes the scope of the research. Firstly, the research is scoped by a focus on a particular process
of business intelligence; registering and processing. Next, the research analyses social media activities on a set
of platforms, while others are excluded. Finally, some ?rms are part of our analysis while other are not.
Registering and Processing One possible way to represent BI, is through a cycle. Though many of these
cycles exist in the literature, they do not di?er much from each other (Pirttimäki & Hannula, 2003). Van Beek
(2006) describes BI as a cycle of registering, processing, and reacting on gathered data. Figure 1-4 highlights
the focus of this thesis. The gathering of data, ‘getting the data in’, is the most challenging aspect of BI,
requiring about 80% of the time and e?ort (Watson & Wixom, 2007). The fundamental scope of the proposed
research will be on this part; the gathering and registering of unstructured data generated on social media, and
is highlighted in ?gure 1-4. One of the core activities related to business intelligence, is the formulation of key
performance indicators. Not before these metrics are de?ned, the registering of data can commence. Therefore,
key performance indicators take a central role in this research.
10 Research Problem
Sample selection
Content analysis | Text mining | Data analysis
Literature review
on social media
Literature review
on (real-time)
business
intelligence
Literature review
on big data
Literature study
Analyse related
research
Create framework
to position this
thesis in exististing
theories
Content Analysis
Draw conclusions
on applicability of
social media
content for BI
Literature review
on (e-)business
Determine
knowledge gap
Formulate
research
questions
Relate social
media messages
to KPIs
Analyse
differences in
volume across
industries
Analyse
differences in
content across
industries
Classify
applicability of
social media for BI
Determine
keywords to scrape
Twitter data
Scrape social
media data
Record social
media posts in
database
Determine content
to scrape from
social media data
Scraping
Data cleaning
Analyse
differences in
content across
customer relations
Create sample of
companies to
analyse
Categorise
companies in
customer relation
types
Categorise
companies in
industries
Framework creation
Create social
business
intelligence
framework
Experience
in processing
social media
data
Understanding
‘where’ to apply
social BI
Understanding
‘how’ to apply
social BI
Formulating
requirements for
social BI
framework
Verification of
framework
Analyse
differences in
volume across
customer relations
Figure 1-3: Research Approach
Social Media Platforms Many social media platforms are available, and the range of social media platforms
is vast and growing (A. N. Smith et al., 2012). These platforms di?er in scope, functionality and in culture
(Boyd & Ellison, 2007). “Some sites are for general masses, like Twitter, Hi5 and Facebook. Other sites, like
LinkedIn, are more focused on professional networks. Media sharing sites such as MySpace, YouTube, and Flickr
concentrate on shared videos and photos” (Kietzmann et al., 2011). In addition, there also exist platforms that
are explicitly not purposed to be publicly accessible. An example of such a platform is Yammer, which is used by
1-8 Literature Review 11
G
D
A
Register
React
Process
Combination
Distribution
E
v
a
l
u
a
t
i
o
n
Operational
Tactical
Strategical
C
o
n
t
e
x
t
u
a
l
T
r
a
n
s
a
c
t
i
o
n
a
l
E
n
v
i
r
o
n
m
e
n
t
Figure 1-4: Thesis Scope, visualised in the BI cycle (van Beek, 2006)
organisations for internal communication. The focus of this thesis however is on publicly accessible platforms,
since the key purpose is to investigate what kind of information ?rms can derive from publicly accessible social
media. When investigating the opportunities of social media data for business intelligence purposes it is valuable
to collect data from a great variety of platforms, so that possible di?erences in the nature of the content can be
identi?ed. Our analysis includes 25 platform types that are monitored, among them Facebook’s public pages,
Twitter, Google+’s public pages, Identi.ca, YouTube, Flickr, Vimeo, Picasa, Wordpress based blogs, Blogger,
Typepad, RSS enabled blogs, Yahoo! Answers and Newssites. The di?erent platforms are monitored, implying
that each time a post is generated containing the predetermined keywords (e.g. ‘Albert Heijn’, or ‘Heineken’),
the post is extracted and saved into a database containing all gathered social media posts. Anno 2012, these
platforms are the most popular social media platforms in the Western World. However, there are many other
social network sites in the world. E.g Sina Weibo (the Chinese counterpart of Twitter), Qzone (China), Habbo,
Badoo (Latin America) and many other platforms are not part of our analysis. The selected social media
platforms – as well as the ?rms – are active in Europe.
Selection of Firms 18 ?rms have been selected for the analysis, that are active in di?erent industries. The
starting point of the sample selection has been the list of ?rms that are part of the Amsterdam Exchange Index
(“AEX”). The main reason for this selection criterion is the fact that these organisations are stock listed, and
hence publicise annual reports containing information about strategic initiatives, ?nancial ?gures, etc. In case
the analysis shows inter sector di?erences – e.g. between two comparable ?nancial institutions – the annual
reports may provide company speci?c information (e.g. amount of employees, attitude towards social media,
etc.) clarifying these di?erences. Whenever a sample containing privately owned companies would have been
selected, access to additional information would be limited. In addition, organisations listed in the AEX are
generally well-established, visible to the public and regularly subject to news articles. It is therefore expected
that these ?rms are subject of discussion on social media. The sample is further elaborated in section 4-2-2
(page 45).
1-8 Literature Review
In the following section research that relates to this thesis is presented. The literature related to the topic of
this thesis has been found using search queries (SOCIAL MEDIA), (SOCIAL BUSINESS INTELLIGENCE),
(SOCIAL MEDIA DATA), (SOCIAL MEDIA) AND (DATA EXTRACTION), (SOCIAL MEDIA) AND
(CRAWLING), (WEB 2.0) AND (CRAWLING), (TWITTER) AND (BUSINESS INTELLIGENCE) and (WEB
2.0) AND (BUSINESS INTELLIGENCE) in the scienti?c literature databases ScienceDirect, JStore and the
search engine Google Scholar. Existing research related to this thesis have been found in the scienti?c journals
of Electronic Markets, Computer Science, Journal of the American Society for Information Science, Public
Relations, Journal of Marketing, Public Relations Review, Expert Systems with Applications and the Journal of
Interactive Marketing. Though not all of these studies are explicitly related to business intelligence, the central
theme is the extraction of information from social media sites. We do not limit our review of related work to
12 Research Problem
one social media platform. Instead, the presented research consists of a mix in which Twitter, Facebook, Blogs,
and Questioning and Answering sites served as the data source.
Jansen, Zhang, Sobel, and Chowdury (2009) analysed 150,000 tweets containing branding comments, sentiments,
and opinions. The researchers analysed the content of the tweets, and found that 19% of microblogs’ posts
contain a mention of a brand. Of these branding microblogs, nearly 20% contained some expression of brand
sentiments. Of these, more than 50% were positive and 33% were critical of the company or product. The
research concludes that microblogging is an online tool for customer word of mouth communications, and is
especially suited for brand management activities.
Zhang et al. (2011) – in their quest to uncover the Twitter community dynamics – studied the “in?uences of
business engagement in online word-of-mouth communication” and investigated “the trajectories of a business’
online word-of-mouth message di?usion in the Twitter community”. They studied nine-brands on Twitter, and
concluded that “business engagement on Twitter enhances consumers’ engagement with online word-of-mouth
communication”. Therefore, the authors argue that “businesses must go beyond simply being aware of or taking
into consideration electronic word-of-mouth messages and instead must engage in the communication process
as both initiators and active participants. Next, Zhang et al. (2011) found that “retweeting, as an explicit way
to show consumers’ response to business engagement, only reaches consumers with a second-degree relationship
to the brand” and that the “life cycle of a tweet is generally 1.5 to 4 hours at most”.
McCorkindale (2010) investigated – based on a content analysis – how the Fortune 50 companies used Facebook.
The research studied how many fans an organisation had, what organisational information was included, if they
used photos and videos, if they used discussion boards, whether they generated feedback, etc. She found that
companies are using Facebook extensively, but that most companies were not using the site to “disseminate
news and information about the organisation”. Next, the research indicated that the companies should focus
more on “relationship-building strategies in order to encourage users to revisit the sites”. The content analysis
of McCorkindale (2010) revealed that there are several reasons why people post on Facebook pages. “Some
were current employees who identi?ed where they worked and for how long, while some were former employees
reconnecting with past coworkers. Headhunters posted jobs at competing corporations on the wall, and job
seekers posted they were looking for employment. Customers having product problems, especially in the
technology ?eld, would post their issues on the wall hoping to ?nd solutions. Journalists also posted on pages
requesting interviews”.
Agichtein, Castillo, Donato, Gionis, and Mishne (2008) argue that the “quality of user-generated content varies
drastically from excellent to abuse and spam”, and that the “task of identifying high-quality content in sites
based on user contributions – social media sites – becomes increasingly important”. Therefore, the authors
developed a method to exploit “community feedback to automatically identify high quality content”. Agichtein
et al. (2008) applied their model on a popular questioning and answering site (Yahoo! Answers). The system
of Agichtein et al. (2008) models all user relations, and applies the user ratings on the individual answers. As
such, the system determines high-quality content based on the ratings that users assigned to the content.
Guo, Zhang, Tan, and Guo (2012) developed a system that detects popular topics on Twitter. According to
the authors, the key technology in mining web text includes the modules “text classi?cation, clustering, topic
detection and tracking, opinion tendency identi?cation, and multi-document automatic summarisation”. Guo
et al. (2012) argue that popular topic detection systems should entail these ?ve modules. However, the nature
of tweets – “very short, sparse and spreading rapidly” – is di?erent from regular web text. Therefore, Guo et
al. (2012) propose a more “?exible and practical approach based on frequent pattern mining”.
Kozinets et al. (2010) qualitatively studied 83 blogs in order to understand how marketing departments in?uence
consumer-to-consumer communications. The authors distinguished the strategies of the marketeers into four
categories – evaluation, embracing, endorsement, and explanation.
Araujo and Neijens (2012) researched how top global brands participated in social network sites by investigating
which factors in?uence the presence and level of engagement of these brands on social network sites. The authors
reviewed the corporate websites of 129 brands in di?erent markets, targeting di?erent ages of audience, di?erent
home markets, di?erent web operations and in di?erent countries. Consequently, the authors determined
whether or not the companies refer to their presence at social network sites. The research found that social
network site “presence was signi?cantly higher for information technology and telecommunication brands”
(Araujo & Neijens, 2012), implying that the presence of ?rms on social media di?ers between ?rms in di?erent
industries. Furthermore, Araujo and Neijens (2012) found that “brands targeting younger audiences also engage
at higher levels than brands targeting generic audiences” and that the “country in which the brand operates
plays a signi?cant role in a brand’s likelihood of adopting social network sites”. The ?ndings of Araujo and
1-9 Scienti?c Relevance 13
Neijens (2012) indicate that the applicability of social media for business purposes di?ers between ?rms, which
support the basis of this thesis.
Dey and Haque (2008) acknowledge that “the data generated from online communication acts as potential gold
mines for discovering knowledge”. However, as Dey and Haque (2008) illustrate, “the quality of texts generated
from online sources can be extremely poor and noisy” because the “text data typically comprises spelling errors,
ad-hoc abbreviations and improper casing, incorrect punctuation and malformed sentences”. It is therefore
that text mining techniques based on “pure linguistic strategies fail to extract information from noisy text”.
According to Dey and Haque (2008), “statistical techniques on the other hand which though not as successful
as the linguistic methods, are more suited to extract information from noisy text. However, lack of appropriate
training data often poses as a bottleneck”. Dey and Haque (2008) conclude that – when processing unstructured
social media data – “domain related training sets” are required to the clean the text before the text can be
processed by Natural Language Processing Tools. With such domain related training sets, the word ‘small’ can
be classi?ed as either positive or negative, depending on its context.
Lovejoy et al. (2012) performed a content analysis of the tweets related to 73 non-pro?t organisations to examine
“how these organisations use Twitter to engage stakeholders”. Within that analysis, the researchers looked at
“the organisations’ utilisation of tweet frequency, following behaviour, hyperlinks, hashtags, public messages,
retweets, and multimedia ?les”. Lovejoy et al. (2012) conclude that non-pro?t organisations use social media as
a “one-way communication channel”, and not as a platform for “conversation and community building”.
Lee (2012) acknowledge that the “contents of microblogs preserve valuable information”. In his study, Lee
(2012) focused on real-world o?ine events, and the information that was generated on social media sites related
to those events. With his system, it is possible to detect real-world events through the content on social media
sites. Also Lee (2012) argues that the challenge of automatically classifying social media posts is the informal
structure of the text.
Dhar & Chang’s (2009) research is one of the few that studied the relation between social media activity and
organisational performance. More speci?cally, they employed social media data to predict sales in the music
industry. Using linear and nonlinear regression, Dhar and Chang (2009) found that “(a) the volume of blog
posts about an album is positively correlated with future sales, (b) greater increases in an artist’s Myspace
friends week over week have a weaker correlation to higher future sales, (c) traditional factors are still relevant
– albums released by major labels and albums with a number of reviews from mainstream sources also tended
to have higher future sales”.
Tirunillai and Tellis (2012) studied the relationship between user-generated content and stock market
performance of the ?rm. The authors found that “of all the metrics of UGC, volume of chatter has the
strongest positive e?ect on abnormal returns and trading volume. Whereas negative UGC has a signi?cant
negative e?ect on abnormal returns, positive UGC has no signi?cant e?ects on these metrics. The volume of
chatter and negative chatter have a signi?cant e?ect on trading volume”. In addition, Tirunillai and Tellis
(2012) found that “an increase in o?-line advertising signi?cantly increases the volume of chatter and decreases
negative chatter”.
From the literature review, we can conclude that there is scienti?c attention in the research ?eld of social media
and the relation with organisational performance. However, no research has been found that investigates the
applicability of social media for organisations in (i) di?erent industries and with (ii) di?erent customer relation
types. Next, though some studies individual tackle di?culties that are inherent to the usage of social media data,
no research has been found that integrally describes how social media should be collected and processed within
a ?rm. As illustrated, the opportunities for social media are bene?cial on many aspects. However, managers
are also reluctant to allocate budget to social media activities (Weinberg & Pehlivan, 2011) and incorporate
social media data in the ?rm’s BI process, because they do not fully understand what social media intelligence
may bring to the ?rm. In addition, it is unclear which type of ?rms are subject of discussion on social media
and – if they are – how a ?rm should collect and process these data so that it adds value to the ?rm.
1-9 Scienti?c Relevance
The proposed research touches the world of e-business, which implies “the transformation of key business
processes through the use of internet technologies” (Cha?ey, 2009). The monitoring of opinions, customer
thoughts, etc. by electronic means – for instance by social media sites – can be positioned under the denominator
‘e-business’. Many literature exists in which the world of e-business is described. This research contributes to
existing models and theories by positioning social media content as an external factor in these theories.
14 Research Problem
Next, science is built on data. The more data is available to scientists, the “greater the level of transparency
and reproducibility and hence the more e?cient the scienti?c process becomes” (Molloy, 2011). Historically,
scienti?c data has not been openly available. In recent years, several scientists advocate the application of open
data. The proposed research will be based on publicly accessible data – coming from social media – and will
hence contribute to understanding the opportunities and threats of applying public data for scienti?c purposes.
The literature contains many de?nitions of business intelligence, and provides theories describing how BI
processes internal as well as external data. Data from social media can be positioned under external factors.
This thesis positions explicitly adds social media data into the existing theories of business intelligence.
Next, the research will be executed based on the research methodology of content analysis. Though this method
is yet widely applied in many research areas, the fact that the source of the content in this research is social
media, makes it new. The lessons learned in this research from applying a content analysis on social media data
contribute to the research methodology of content analysis.
Finally, the literature of customer relationship management (“CRM”) describes how organisations interact with
their customers. Recent literature also includes social media solutions into CRM activities. This research reveals
how user-generated content varies between industries, departments and the position regarding consumers. As
such, the conclusions of this thesis contribute to the applicability of CRM through means of social media.
1-10 Societal Relevance
Many executives avoid or ignore social media because they do not understand what it is and how to engage
with it and learn, though they sense that social media is – and will remain – an important “fabric of commerce”
(Kietzmann et al., 2011; Weinberg & Pehlivan, 2011). This thesis contributes to a further understanding of social
media, and to leverage the opportunities of applying the valuable content on these platforms for organisational
e?orts. The social media phenomenon is relatively fresh, Facebook was launched in 2004, Twitter in 2006.
Because of the novelty, the opportunities for organisations’ social BI activities are rather unexplored. This
research also contributes to an understanding for ?rms whether or not social media data can be applied for
business intelligence purposes in which context. In addition, we expect that legacy BI vendors – such as SAP,
Oracle and IBM – are soon asked by their clients to add a social media component to their BI suite. For these
organisations social business intelligence is also a new phenomenon, and social media data can not be directly
applied to their existing systems (Reinhold & Alt, 2011). The conclusions of this thesis support BI vendors in
the development of social media components within their product range.
1-11 Project Deliverable
This research will reveal two central questions that describe (i) where and (ii) how ?rms can derive information
from social media data for their decision-making process. Therefore, the project has two deliverables.
Deliverable 1: Where? This deliverable speci?es in which contexts a ?rm can implement social business
intelligence. For reasons explained in section 1-2, it is expected that a ?rm’s contexts and aspects determine the
applicability of social media data for BI purposes. This deliverable allows organisations to determine whether
or not they are suited for the applicability of social media for business intelligence.
Deliverable 2: How? This deliverable consists of a procedure that prescribes how ?rms can execute a social
business intelligence system. The procedure can be considered as a document prescribing how an organisation
can execute social business intelligence, and which technical and institutional elements are involved in a social
BI system.
Chapter 2
Conceptual Frame of Research
In this chapter, we de?ne and illustrate what business intelligence (“BI”) is, how BI is applied in ?rms, what
the most important elements are and how BI is regarded in this thesis. In a later stadium of this research a
procedure for processing social media data to support business intelligence will be developed. Such a procedure
is required to ?t in the current method that ?rms adhere to in executing BI. Therefore, an understanding of
business intelligence within ?rms is essential. This chapter provides the knowledge of business intelligence that
is required when we develop a procedure to collect and process social media data for BI purposes in a later
stadium.
Section 2-1 starts by a description of the various de?nitions of BI, and consequently formulates the perspective
on BI that is adhered to throughout this research. In section 2-2 the elements relating to the determination
of ‘what to measure?’ are discussed, a key activity in business intelligence. Section 2-2 explains the relation
between a ?rm’s strategy and the ?rm’s performance metrics. As we will see, performance metrics take a central
role in BI. In section 2-3, the processing of data is described. Finally, section 2-4 concludes this chapter by a
description of how business intelligence is applied within ?rms.
2-1 Business Intelligence Perspectives
Business intelligence is a process in which data is translated into information that is required for managerial
decision-making. The literature contains many de?nitions of business intelligence. Elbashir et al. (2008) state
that “business intelligence systems provide the ability to analyse business information in order to support
and improve managerial decision making across a broad range of business activities”. Van Beek (2006)
de?nes business intelligence as “a continuous process that helps organisations gathering and registering data,
analysing it and consequently applying the resulting information and knowledge in decision-making processes
to improve organisational performance”. Rouibah and Ould-ali (2002) describe business intelligence as “a
strategic approach for systematically targeting, tracking, communicating and transforming relevant weak signs
into actionable information on which strategic decision-making is based”. Lonnqvist and Pirttimaki (2006)
de?ne BI as “an organised and systematic process by which organisations acquire, analyse, and disseminate
information from both internal and external information sources signi?cant for their business activities and for
decision making”. Although these de?nitions vary slightly from each other, the common aspect is that business
intelligence is perceived as a process that translates data into interpretable information that supports managerial
decision-making.
Van Beek (2006) visualises business intelligence (“BI”) as a cycle, consisting of three main processes (?gure 2-1).
For the remainder of this thesis, we follow van Beek’s – loosely de?ned – perspective on business intelligence
because it captures the various de?nitions of BI found in the literature. The three main processes – register,
process and react – are discussed in the following paragraphs.
Register The BI cycle starts with carefully listening – registering – to the environment. Within the
environment, a distinction is made between contextual and transactional environments. The contextual
environment consists of aspects that (may) have an e?ect on the organisation. The transactional environment
16 Conceptual Frame of Research
G
D
A
Register
React
Process
Combination
Distribution
E
v
a
l
u
a
t
i
o
n
Operational
Tactical
Strategical
C
o
n
t
e
x
t
u
a
l
T
r
a
n
s
a
c
t
i
o
n
a
l
E
n
v
i
r
o
n
m
e
n
t
Figure 2-1: Business Intelligence Cycle (van Beek, 2006).
consists on the one hand of actors that have a direct relation with the company, like customers, suppliers,
employees and competitors. On the other hand, the transactional environment is made up of institutions
a?ecting the organisation, like new policies or legislation. BI registers signals arising from this environment.
Process Consequently, when data (in whatever format) is registered, it is required to be processed. Processing
the gathered data will reveal trends and provide valuable information. Van Beek (2006) positions a ‘small BI
cycle’ within this process; data is gathered, analysed, and distributed to the right organisational departments.
This part of the BI cycle will be further described in section 2-3.
React Following the results provided by the processing of the data, the company can react. Van Beek
(2006) argues that a company can react on three levels; operational, tactical or strategical. Consequently,
the environment evaluates the companies’ changes in interactions, resulting in new signals for the ?rm’s BI
cycle.
2-2 Registering the Right Indicators
As illustrated in the previous section, the business intelligence process starts with registering. But, what is it
that a ?rm has to register? Not unless a ?rm has clearly set what will be registered, the BI cycle can commence.
The determination of ‘what to measure’ is a process on itself, which is described in this section. In the following
sections, the necessary steps to determine what a ?rm has to register are described. Section 2-2-1 describes that
?rms formulate – or, should formulate – their performance metrics based on their strategy. In section 2-2-2, two
widely accepted frameworks that support ?rms in the formulation of performance metrics are discussed. Section
2-2-3 discusses a framework that illustrates how a ?rm should design and implement a system for measuring
organisational performance. Next, section 2-2-4 describes the various types of performance measures that are
applied by ?rms. Finally, section 2-2-5 describes ten commonly applied categories of key-performance indicators,
which take a central role in the business intelligence process.
2-2-1 Strategy and Business Model
Firms align their business model with their strategy. The determination of ‘what to measure’ initially starts with
the ?rm’s strategy. A strategy consists of a mission, values, vision, goals, objectives and plans. The “mission
and values de?ne why the organisations exists, what it does, and its guiding principles. The vision combines
an overarching purpose with an ideal, future-state competitive positioning. Goals are broad statements that
embody what the organisation would like to achieve in three to ?ve years, while objectives represent short-term
goals of one to three years” (Eckerson, 2009).
Strategies are translated into business models (Bouwman, Faber, Haaker, Kijl, & de Reuver, 2008). There exists
a variety of views on business models in the literature. Osterwalder and Pigneur (2010) state that “a business
model describes the rationale of how an organisation creates, delivers, and captures value”. Bouwman et al.
2-2 Registering the Right Indicators 17
(2008) provide an all-embracing de?nition of a business model for service-oriented organisations by stating that
“a business model is a blueprint for a service to be delivered, describing the service de?nition and the intended
value for the target group, the sources of revenue, and providing an architecture for the service delivery, including
a description of the resources required, and the organisational and ?nancial arrangements between the involved
business actors, including a description of their roles and the division of costs and revenues over the business
actors”. In other words, a business model describes what a ?rm does, why it does that, how it does that, with
whom it does that and for whom it does that.
By aligning the business model with the ?rm’s strategy, managers question themselves “why are we performing
this activity?”, and “what does it contribute to?”. “It is only through consistency of action that strategies
are realised” (Neely, Gregory, & Platts, 1995). Gates (2001) argues that applying value driver maps to
analyse a company’s vision and drivers of performance helps a company aligning its business model with its
strategy. By aligning the business model with the strategy, a ?rm ensures that it performs those activities that
contribute to the intended strategy. It is in this phase of the BI process where managers determine where in
the organisation and by which activities value is added to the main objective of the organisation, and why and
how the performance of these activities are to be measured. There are many frameworks developed to craft
a business model that is based on the ?rm’s strategy, at which – in the end – performance metrics can be
formulated. These frameworks are the topic of the following section.
2-2-2 Frameworks Supporting the Formulation of Performance Indicators
Performance measurement is “the process of quantifying the e?ciency and e?ectiveness of action” (Neely et al.,
1995). To quantify actions – forming the ?rm’s business model – indicators are required that represent these
actions. Such indicators are called performance indicators. Performance indicators should be derived from a
?rm’s strategy (R. S. Kaplan & Norton, 1992, 1993, 1996; Kennerley & Neely, 2003; Tsai & Cheng, 2012;
Fortuin, 1988). The previous section revealed that a business models amongst others speci?es which activities
are required to deliver value for the customer of the ?rm. Hence, the business model prescribes which activities
are to be executed. The activities are consequently required to be measured, so that managers can determine
the organisation’s performance in accordance with its strategy. These activities are measured by performance
indicators, indicating the performance of the individual activities.
But how do we determine the right performance indicators? One of the ?rst widely recognised frameworks
(Neely, Bourne, & Kennerley, 2000) that help managers to decide ‘what to measure’ is the balanced scorecard,
developed by R. S. Kaplan and Norton (1992). R. S. Kaplan and Norton (1992) provide a framework that
supports managers in formulating performance indicators based on four perspectives. Because the balanced
scorecard ensures that managers do not only focus on ?nancial ?gures, it gives managers a “comprehensive view
of the business”, which is required in the competitive environment of the ?rm (R. S. Kaplan & Norton, 1992).
Since the balanced scorecard links a company’s strategy with concrete actions (R. S. Kaplan & Norton, 1996),
it is considered as a tool helping managers to align the company’s business model with its strategy. In response
to the balanced scorecard, Neely, Adams, and Crowe (2001) developed a scorecard that adopts a multi-actor
view in formulating performance metrics, and hence incorporates the perceptions of multiple stakeholders into
the performance metrics formulation process. Both frameworks are discussed in the following sections.
The Balanced Scorecard
Because “you get what you measure” (Kennerley & Neely, 2003), R. S. Kaplan and Norton (1992) advocate
that a ?rm should measure those metrics that contribute to the ?rm’s strategy. In addition, R. S. Kaplan
and Norton (1992) argue that managers should not only focus on ?nancial ?gures, but also on other areas
representing organisational performance. Therefore, R. S. Kaplan and Norton (1992) developed the “balanced
scorecard”. The balanced scorecard is not without reason called a ‘balanced’ scorecard. In essence, it stimulates
managers to not only think in ?nancial ?gures when measuring organisational performance, but also on other
areas. The framework distinguishes organisational activities in the (i) customer-, (ii) internal business-, (iii)
innovation and learning-, and (iv) ?nancial perspective, which are discussed below:
• The customer perspective describes how customers view the ?rm, and ensures that customer’s needs are
ful?lled. R. S. Kaplan and Norton (1992) further categorise the customer’s concerns into time, quality,
performance and service, and cost. Hence, when a manager adopts the customer perspective in formulating
performance metrics, he or she will formulate performance metrics that involve these categories.
18 Conceptual Frame of Research
• The internal business perspective describes where the particular ?rm must excel at and speci?es what a
“company must do internally to meet its customers’ expectations” (R. S. Kaplan & Norton, 1992). It is
in this perspective where managers “attempt to identify and measure their company’s core competencies,
the critical technologies needed to ensure continued market leadership. Companies should decide what
processes and competencies they must excel at and specify measures for each” (R. S. Kaplan & Norton,
1992).
• The third perspective of the balanced scorecard – innovation and learning – ensures that the organisation
continues to improve and create value. Due to competition, “the targets for success keep changing”
(R. S. Kaplan & Norton, 1992). It is therefore that companies must design its organisation in a way
that it can innovate. It is in this perspective where managers consider the development of new products,
entering new markets, etc.
• The ?nancial perspective ensures that the shareholders’ needs are ful?lled. “Financial measures
indicate whether the company’s strategy, implementation, and execution are contributing to bottom-line
improvement” (R. S. Kaplan & Norton, 1992).
The balanced scorecard set the scene for the development of a variety of other performance measurement
frameworks at the beginning of the 1990s.
The Performance Prism
In response to the various types of scorecards that have been developed after Kaplan & Norton’s (1992)
balanced scorecard, Neely et al. (2001) developed a “second generation performance measurement framework”
called the performance prism. According to Neely and Adams (2005), there were three fundamental reasons
why the balanced scorecard was outdated, and why a new framework was required. Firstly, the balanced
scorecard solely focuses on the needs of two groups of (internal) stakeholders; shareholders and customers. In
today’s business environment, ?rms can no longer consider only these two groups of stakeholders. For example,
employees, environmental parties, labour unions, other communities, regulatory bodies, etc. have been fully
denied in the balanced scorecard, while these groups truly in?uence a ?rm in practice. Second, an organisation’s
“strategy, processes, and capabilities have to be aligned and integrated with one another” (Neely & Adams,
2005), e.g. with the processes of the ?rm’s suppliers. Third, ?rms “have to recognise that their relationships are
reciprocal – stakeholders have to contribute to organisations as well as receive something from them” (Neely &
Adams, 2005). Thus, the key innovation that is captured in the performance prism is the fact that it takes a
comprehensive stakeholder orientation, i.e. a multi-actor perspective, whereas former frameworks (such as the
balanced scorecard) adopt a mono-actor perspective. It is necessary to adopt a multi-actor perspective, since
?rms need to have “contributions from their stakeholders – usually capital and credit from investors, loyalty
and pro?t from customers, ideas and skills from employees, materials and services from suppliers, and so on”
(Neely & Adams, 2005).
The performance prism consists of ?ve perspectives, whereas the balanced scorecard comprises of four. The
central element in all these ?ve performance prism perspectives is the stakeholder aspect.
• Stakeholder Satisfaction
The stakeholder satisfaction perspective ensures that managers consider “who the ?rm’s stakeholders are,
what they do, and what they need”. Whereas the balanced scorecard explicitly focuses on two groups
of stakeholders, i.e. on customers through the customer perspective and on shareholders through the
?nancial perspective, the performance prism does not specify stakeholder groups but rather allows for
ambiguity. As such, the ?rm is stimulated to consider all stakeholder groups in its ecosystem, and specify
what these groups want.
• Strategies
The second facet of the performance prism focuses on strategies. With the identi?cation of the needs in the
stakeholder satisfaction perspective, strategies can be developed that ful?l the needs of the stakeholders.
The key question in this facet is: “What strategies should the organisation adopt to ensure that the
wants and the needs of its stakeholders are satis?ed?” (Neely & Adams, 2005). Whereas the balanced
scorecard method of formulating measures starts with the ?rm’s strategy, Neely and Adams (2005) argue
that strategies should be designed in accordance with the needs of the stakeholders.
2-2 Registering the Right Indicators 19
• Processes
In the third facet of the performance prism, the processes are de?ned that are required to ful?l the
strategies de?ned in the second perspective. It is in this perspective where general business processes,
such as product development, demand generation, demand ful?lment, planning, etc. are de?ned. Neely et
al. (2001) stress the importance of measures that re?ect the performance of the processes. Thus, it is in
this perspective where organisations consider which measures are required to determine the organisational
performance.
• Capabilities
Within the fourth aspect of the performance prism, the capabilities required to execute the processes are
identi?ed. Capabilities consist of a combination of “people, practices, technology, infrastructure, etc.”.
• Stakeholder Contribution
Finally, the performance prism “recognises the fact that not only do organisations have to deliver value to
their stakeholders, but also that organisations enter into a relationship with their stakeholders which should
involve the stakeholders contributing to the organisation” (Neely et al., 2001). It is in the stakeholder
contribution perspective where ?rms consider what they want from their stakeholders.
5. Capabilities: What capabilities do we need to put
in place to allow us to operate our processes?
Together, these ?ve perspectives provide a comprehen-
sive and integrated framework for thinking about organi-
zational performance in today’s operating environment
(see Fig. 1).
The performance prismalso seeks to address the short-
comings of the ?rst-generation measurement frameworks
and methodologies, such as the balanced scorecard,
the work on shareholder value, and the various self-
assessment frameworks, such as the Malcolm Baldrige
Award criteria and the business excellence model of
the European Foundation for Quality Management
(EFQM).
The Nature of the Measurement
Problem
Why is this new performance measurement framework
needed? After all, everyone knows that ‘‘you can’t manage
what you don’t measure.’’ And given that people have
been managing organizations for years, then surely by
now they must have perfected their measurement
systems.
Sadly, as in so many walks of life, theory does not re?ect
practice. The number of organizations with weak perfor-
mance measures and measurement systems in place is
immense. Examples abound of organizations that have
introduced performance measures that quite simply
drive entirely the wrong behaviors. There must be
a better way.
There has been a revolution in performance measure-
ment and management during the last few years. Various
frameworks and methodologies—such as the balanced
scorecard, shareholder value added, activity-based cost-
ing, cost of quality, and competitive benchmarking—have
each generated vast interest, activity, and consulting rev-
enues, but not always success. Yet therein lies a paradox.
It might reasonably be asked: how can multiple, and
seemingly inconsistent, business performance frame-
works and measurement methodologies exist? Each
claims to be unique and comprehensive, yet each offers
a different perspective on performance.
Kaplan and Norton’s balanced scorecard, with its four
perspectives, focuses on ?nancials (shareholders), cus-
tomers, internal processes, plus innovation and learning.
In doing so, it downplays the importance of other stake-
holders, such as employees, suppliers, regulators, and
communities. The business excellence model combines
results, whichare readily measurable, withenablers, some
of which are not. Shareholder value frameworks incorpo-
rate the cost of capital into the equation, but ignore ev-
erything (and everyone) else. Both the activity-based
costing and the cost of quality frameworks, on the
other hand, focus on the identi?cation and control of
cost drivers (non-value-adding activities and failures/
non-conformances, respectively), which are themselves
often embedded in the business processes. But this highly
process-focused view ignores any other perspectives on
performance, such as the opinions of investors, custom-
ers, and employees. Conversely, benchmarking tends to
involve taking a largely external perspective, often com-
paring performance with that of competitors or some-
times other ‘‘best practitioners’’ of business processes
or capabilities. However, this kind of activity is frequently
pursued as a one-off exercise toward generating ideas
for—or gaining commitment to—short-term improve-
ment initiatives, rather than the design of a formalized
ongoing performance measurement system.
How can this be? How can multiple, seemingly
con?icting, measurement frameworks and methodologies
exist? The answer is simple: they can exist because they all
add value. They all provide unique perspectives on per-
formance. They all furnishmanagers with a different set of
lenses through which they can assess the performance of
their organizations. The key is to recognize that, despite
the claims of some of the proponents of these various
approaches, there is no one best way to address the mea-
surement and management of business performance. The
reason for this is that business performance is itself
a multifaceted concept, the complexity of which the ex-
isting frameworks only partially address. Essentially, they
provide valuable point solutions.
A Better Solution to the
Measurement Problem
Our solution is the three-dimensional framework that we
call the performance prism. This framework has been
deliberately designed to be highly ?exible so that it can
provide either a broad or a narrow focus. If only a partial
• Stakeholder satisfaction
• Strategies
• Processes
• Capabilities
• Stakeholder contribution
Figure 1
42 Performance Prism
aspect of performance management is required, such as
a single stakeholder focus or a particular business process
agenda, then the performance prism can be applied to
designing a measurement system and appropriate mea-
sures (and their attendant metrics) that address that con-
text. Conversely, if a broad corporate or business unit
performance management improvement initiative is re-
quired, the performance prism is equally capable of sup-
porting that, too. How does it help to achieve these aims?
The performance prism has ?ve perspectives. The top
and bottom perspectives are stakeholder satisfaction and
stakeholder contribution. The three side perspectives are
the organization’s strategies, processes, and capabilities
for addressing those sets of wants and needs. Figure 2
illustrates these ?ve basic perspectives of performance
measurement and management.
Why does the framework look like this and why does it
consist of these constituent components? It is clear that
those organizations aspiring to be successful in the long
termwithin today’s business environment need to have an
exceptionally clear picture of who their key stakeholders
are and what they want or need. But having a clear picture
is not enough. In order to satisfy their own wants and
needs, organizations have to access contributions from
their stakeholders—usually capital and credit from
investors, loyalty and pro?t from customers, ideas and
skills from employees, materials and services from sup-
pliers, and so on. They also need to have de?ned what
strategies they will pursue to ensure that value is delivered
to their stakeholders. In order to implement these strat-
egies, they have to understand what processes the enter-
prise requires and must operate both effectively and
ef?ciently. Processes, in themselves, can only be executed
if the organization has the right capabilities in place—the
right combination of people skill sets, best practices, lead-
ing technologies, and physical infrastructure.
In essence, then, the performance prism provides
a comprehensive yet easily comprehensible framework
that can be used to articulate a given business’s operating
model. Its components are described in the following
sections.
Stakeholder Satisfaction
Where should the measurement design process begin?
One of the great myths (or fallacies) of measurement
design is that performance measures should be derived
from strategy. Listen to any conference speaker on the
subject. Read any management text written about it. Nine
times out of ten the statement ‘‘Derive your measures
from your strategy’’ will be made. This is such a concep-
tually appealing notion that nobody stops to question it.
Yet to derive measures from strategy is to misunderstand
fundamentally the purpose of measurement and the role
of strategy. Performance measures are designed to help
people track whether they are moving in the intended
direction. They help managers establish whether they
are going to reach the destination they set out to reach.
Strategy, however, is not about destination. Instead, it is
about the route that is chosen—how to reach the desired
destination.
At one level, this is a semantic argument. Indeed,
the original work on strategy, carried out in the 1970s
by Andrews, Ansoff, and Mintzberg, asserted that
a strategy shouldexplain both the goals of the organization
and a plan of action to achieve these goals. Today,
Stakeholder satisfaction
Investors
Customers &
Intermediaries
Employees
Regulators &
Communities
Suppliers
What measures? What measures? What measures?
What measures?
Investors
Customers &
Intermediaries
Employees
Regulators &
Communities
Suppliers
Investors
Customers and
intermediaries
Employees
Regulators and
communities
Suppliers
Which
strategies?
What measures? What measures? What measures?
Which
processes?
Which
capabilities?
What measures?
What measures?
Stakeholder satisfaction
Figure 2
Performance Prism 43
Figure 2-2: Five Facets of the Performance Prism. Adopted from Neely & Adams (2005).
Figure 2-2 presents the performance prism. The ?ve areas of the prism represent one of the perspectives.
The triangular – outside facing – surfaces represent the two stakeholder perspectives, which are unique for
the performance prism. The three rectangular – inside facing – surfaces represent the strategy, processes and
capabilities perspectives. The fact that stakeholders are an important element in the performance prism is
also highlighted by the substance of the prism, which mentions groups of stakeholders. As argued by Neely
and Adams (2005), the performance prism “has been deliberately designed to be highly ?exible so that it can
provide either a broad or a narrow focus”. As a result, the perspectives of the performance may seem vague.
The authors explicitly made the perspectives vague so that the framework is broadly applicable.
To conclude, a ?rm should register those metrics that re?ect the performance of the activities that contribute
to the ?rm’s strategy. Since organisational activities are tailored to the ?rm’s strategy, and metrics are derived
from these activities, there exists a link between performance metrics and an organisation’s strategy. Though
the alignment of performance metrics and strategy may sound self-evident, many organisations struggle with
strategic alignment: even at the healthiest companies, about 25% of the employees are unclear about their
company’s direction. KPMG (2009) argues that in many organisations, there is “no explicit linkage between
20 Conceptual Frame of Research
the strategy and the information used to manage the business”, implying that managers are measuring activities
that do not contribute to the ?rm’s strategy. Managing without or the wrong metrics “gives one the feeling of
being lost with no hope”, and leads to a “lack of management control” (R. Smith, 2006). Once the performance
metrics are established and mutual di?erences in importance are assigned, the BI process can commence. The
values of the performance indicators will then reveal the performance of the ?rm. As illustrated, R. S. Kaplan and
Norton (1992) introduced the ?rst framework that considered other metrics than solely ?nancial ?gures. Next,
Neely et al. (2001) developed – in response to the ?aws relating to the mono-actor perspective of the balanced
scorecard – a framework that considers the ?rm’s stakeholders; the performance prism. The performance prism
is deemed as a framework that is – given today’s business landscape – better suited for performance metrics
formulation than the balanced scorecard. Moreover, since the purpose of this thesis is to incorporate social
media data – created by multiple actors – in management information for di?erent ?elds and departments, a
multi-actor perspective is required.
2-2-3 Performance Measurement System Design Process
Wisner and Fawcett (1991) – as quoted in Neely et al. (2000) – developed a “process for performance
measurement system design”. Figure 2-3 shows the nine-step process proposed by Wisner and Fawcett (1991),
which clearly illustrates that performance metrics should be derived from a ?rm’s strategy, and that performance
indicators should be assigned to functional areas – performing individual activities – of the ?rm.
Clearly define the firm’s mission statement.
Identify the firm’s strategic objectives using the mission statement as a guide.
Develop an understanding of each functional area’s role in achieving the various strategic objectives.
For each functional area, develop global performance measures capable of defining the firm’s overall competitive
position to top management.
Communicate strategic objectives and performance goals to lower levels in the organisation. Establish more
specific performance criteria at each level.
Assure consistency with strategic objectives among the performance criteria used at each level.
Assure the compatibility of performance measures used in all functional areas.
Use the performance measurement system to identify competitive position, locate problem areas, assist the firm in
updating strategic objectives and making tactical decisions to achieve these objectives and supply feedback after
the decisions are implemented.
Periodically re-evaluate the appropriateness of the established performance measurement system in view of the
current competitive environment.
Figure 2-3: Performance measurement system design process (Wisner and Fawcett, 1991).
In a joint research, the business intelligence system vendors SAP, IBM, Corda and Pentaho examined how
organisations formulate performance metrics. Also they found that managers determine performance metrics
based on the ?rm’s strategy. Next, they concluded that performance metrics should be “tailored to every
individual and role in the organisation” (Eckerson, 2009). As a result, departments and individuals consequently
understand how their activities contribute to the company’s strategy, which is often stated in generic and vague
terms. Consequently, employees will focus on those activities that are important, because “what’s get measured,
gets done” (Kennerley & Neely, 2003). This typical management quote illustrates the imposing consequences
that descent from the determination of performance metrics, that is, “what’s not get measured, gets not done”.
As illustrated by Eckerson (2009), “if the metrics do not accurately translate the company’s strategy, the
organisation will ?ounder”. It is therefore that determining performance metrics is a critical activity of business
intelligence.
2-2 Registering the Right Indicators 21
2-2-4 Typology of Performance Indicators
Performance indicators are key elements in business intelligence, since they re?ect the performance of the
activities that contribute to the ?rm’s strategy. In this section, we elaborate more about performance indicators
and the types of metrics that exist.
Leading and Lagging Indicators There are two fundamental types of indicators; leading indicators and lagging
indicators. Leading indicators lead to results, and are also referred to as ‘(value) drivers’. Lagging indicators
are the results that measure the output of past activities, and are also known as ‘outcomes’ (R. Smith, 2006).
Leading indicators are used to manage, while lagging indicators measure how well has been managed.
With leading indicators it is possible to respond directly when poor results are found. With lagging
indicators, “we get value from knowing how well we performed but have little opportunity to immediately
a?ect under-performance” (R. Smith, 2006). Hence, leading indicators are more powerful, and can be perceived
as short-term indicators of an organisations’ results. It is therefore that ?rms manage by leading indicators.
Illustratively, table 2-1 lists some examples of leading and lagging indicators. It is noteworthy that among
di?erent departments and individuals in organisations there could exist pluriformity in the perception of the
type of indicators, “one man’s outcome measure can be another man’s value driver” (Eckerson, 2009).
Table 2-1: Examples of leading and lagging indicators
Leading indicators Lagging indicators
New sales today Revenues
Planned rework today Cost
Customer cases currently open Capacity
Contracts in negotiation for Q2 Return on equity
Identi?ed software bugs Customer satisfaction
Employee retention
Margins
Reliability
Failures
Downtime
Quantitative and Qualitative Indicators Another distinction between metrics is the di?erence between
quantitative or qualitative based indicators. Quantitative indicators measure processes by counting, adding,
averaging, etc. numbers. Examples of quantitative measures are inventories, number of orders, number of clients,
delivery time of goods, sales, other ?nancial ?gures, etc. In contrast with qualitative indicators, quantitative
indicators are relatively easy to measure.
However, there are many other criteria to judge performance than solely on (?nancial) quantitative indicators
(Neely et al., 2000; Eccles, 1991). Other metrics are qualitative in nature and require a proxy to be measured.
An example of a qualitative measure is customer satisfaction. The measurement of customer satisfaction
results in quantitative data, but is primarily based on subjective interpretation of customers’ opinions.
Customer satisfaction is therefore traditionally measured by surveys (Peterson & Wilson, 1992). “Traditionally,
performance evaluation has depended to a great extent on ?nancial indicators. However, given the current
environmental uncertainties, ?nancial indicators can no longer give a complete view of business operations”
(Tsai & Cheng, 2012). It is therefore that qualitative measures are as much as important as quantitative
measures. The trick is to identify the links between qualitative measures and ?nancial measures. Firms can for
instance conduct statistical analyses to correlate qualitative indicators with ?nancial performance. Regression
analysis can be applied to identify the key drivers that impact sales, pro?tability, etc. The performance prism
framework allows for the incorporation of qualitative indicators next to quantitative indicators.
Key Performance Indicators To distinguish between performance indicators that are more important than
others, some indicators are termed ‘key-performance indicators’ (“KPIs”). But what is it that makes a
performance indicator ‘key’? PWC (2007) argues that the performance indicators that are key to a ?rm are
those that are used to manage the business. According to Tsai and Cheng (2012), KPIs “are the groundwork of
the performance system which turns the strategic goals of a company into long-term objectives”. The addition of
22 Conceptual Frame of Research
the word ‘key’ to a performance indicator indicates that these metrics are assigned more attention than others.
Thus, it are the KPIs that represent processes that are paramount for the success of a ?rm. Table 2-2 lists the
elements that a key-performance indicator should ful?l, it should be speci?c, measurable, attainable, realistic
and time-sensitive (“SMART”) (Shahin & Ali Mahbod, 2007).
Table 2-2: Requirements of a key-performance indicator (Shahin & Ali Mahbod, 2007).
Requirement Description
Speci?c KPIs should be detailed and as speci?c as possible.
Measurable A KPI should be measurable against a standard of performance and a standard of
expectation.
Attainable The goal of a KPI should not be out of reach. They should be reasonable and attainable.
Realistic A goal should be realistic taking into account the particular working environment.
Time sensitive Goals should have a time frame for completion, to monitor the progress.
2-2-5 KPI Categories
The previous sections illustrated that ?rms manage their business by measuring key-performance indicators, and
that these indicators should represent – whether indirectly – the ?rm’s strategy and stakeholders’ needs. Because
not every ?rm executes the same strategy and not each ?rm has the same stakeholder groups, di?erent ?rms will
apply di?erent KPIs for performance measurement (Shahin & Ali Mahbod, 2007). Generally, managers apply
value driver maps to determine the performance metrics that correspond with the ?rm’s speci?c strategy (Gates,
2001). A value driver map is a break-down of the ?rm’s strategy into activities – drivers – that are required
to achieve the ?rm’s strategy. On the top level of a value driver map, drivers are generic and for many ?rms
identical. Examples of generic performance metrics are net result, operating result, operating expenditures and
operating margin. These high-level, mostly ?nancial metrics, are generally applied within ?rms. As indicated
by R. S. Kaplan and Norton (1993), all ?rms should focus on the four perspectives; ?nancial, customer, internal
business and innovation and learning when de?ning performance metrics. However, the authors also note that
“speci?c measures within these categories should be tailored to the ?rm’s strategy” (Ittner, Larcker, & Randall,
2003). Thus, ?rms with di?erent strategies require di?erent metrics. How can we categorise metrics that are
speci?c for each ?rm?
Table 2-3: Categories of Key-Performance Indicators (Ittner et al., 2003).
KPI Category Example KPI
1 Short-term ?nancial results Annual earnings, return on assets, cost reduction
2 Customer relations Market share, customer satisfaction, customer retention
3 Employee relations Employee satisfaction, turnover, workforce capabilities
4 Operational performance Productivity, safety, cycle time
5 Product and service quality Defect rates, quality awards
6 Alliances Joint marketing or product design, joint ventures
7 Supplier relations On-time delivery, input into product/service design
8 Environmental performance Government citations, environmental compliance or certi?cation
9 Product and service innovation New product or service development success, development cycle time
10 Community Public image, community involvement
In order to categorise the many performance indicators that one can think of, Ittner et al. (2003) reviewed
literature in the ?eld of the balanced scorecard, intangible assets, intellectual capital, and value-based
management to ?nd the most applied categories of KPIs. Based on the models and frameworks that have been
developed in these research areas, Ittner et al. (2003) distinguish ten performance categories, being short-term
?nancial results, customer relations, employee relations, operational performance, product and service quality,
alliances, supplier relations, environmental performance, product and service innovation, and community. These
categories are listed in table 2-3. The ?nal column of the table shows example metrics. The classi?cation
of Ittner et al. (2003) clearly takes a multi-actor perspective into account, and is therefore considered as
an appropriate classi?cation in line with the performance prism. The search terms (PERFORMANCE
INDICATORS CATEGORIES), (KPI CATEGORIES), (KPI CLASSIFICATION), (KEY PERFORMANCE
2-2 Registering the Right Indicators 23
INDICATORS) AND (CATEGORISATION) in the scienti?c databases ScienceDirect and JStore, and the
scienti?c search engine Google Scholar did not result in literature containing other KPI classi?cations than
Ittner et al.’s (2003) categories. Business intelligence professionals from KPMG have acknowledged that the ten
categories are representative for the KPIs that are actually used by ?rms in practice.
Stakeholder
Satisfaction
Strategies
Processes
Capabilities
Stakeholder
Contribution
1. Short-term financial results
2. Customer relations 3. Employee relations
4. Operational performance
5. Product and service quality
6. Alliances
7. Supplier relations
8. Environmental performance
9. Product and service innovation
10. Community
Figure 2-4: The ?ve Performance Prism Perspectives and corresponding Key-Performance Indicator Categories
As described earlier, performance indicators, and especially key-performance indicators should contribute to
a ?rm’s strategy. And, as illustrated by R. S. Kaplan and Norton (1992), performance metrics should be
established from multiple perspectives, that is, not only ?nancial ?gures. Furthermore, Neely and Adams
(2005) argued that ?rms should formulate performance metrics from a multi-actor perspective. Therefore,
the performance metrics categories established by Ittner et al. (2003) must somehow relate to one of the ?ve
performance prism perspectives. Figure 2-4 schematically shows the relations between the ?ve performance
prism perspectives and the key performance categories. For readability issues, the ?ve perspectives have been
visualised in a pie chart, rather than in a prism. For an explanation of the assignment of the KPI categories to
the ?ve performance prism perspective, see appendix A. The performance categories of Ittner et al. (2003) allow
us to systematically assign social media posts to one of the ten categories. As a result, we can draw conclusions
from the applicability of social media data for certain KPI categories. For the remainder of this thesis, we will
apply the categorisation of Ittner et al. (2003) to categorise key-performance indicators.
24 Conceptual Frame of Research
2-3 Processing: From Data to Information
The registering of signals results in raw data which needs to be processed before it represents information.
Figure 2-1 showed the business intelligence cycle. In the second phase of BI, registered signals are processed.
Van Beek (2006) describes this process as a cycle on itself, which is discussed in this section. It is important
to understand the theory underlying the processing of signals when considering to apply social media data for
BI purposes, because an organisation usually applies business intelligence already. A social media component
should hence be consistent with the existing system(s) and process(es). Van Beek (2006) distinguishes 15
activities making up the processing of gathered data and turning it into information. Figure 2-5 shows the
activities in the BI cycle.
Register
React
Process
Combination
Distribution
E
v
a
lu
a
t
io
n
Operational
Tactical
Strategical
C
o
n
t
e
x
t
u
a
l
T
r
a
n
s
a
c
t
io
n
a
l
E
n
v
ir
o
n
m
e
n
t
1. Collecting
2. Filtering
3. Combine
4. Aggregating
5. Visualising
6. Interpret
7. Internalise
8. Revise &
Recalibrate
9. Verify
10. Enrich
11. Share &
Communicate
12.
Remember
13. Decide
14. Distribute
15. Anticipate
on Changes
Figure 2-5: Processing data to information. Based on van Beek (2006).
The ?fteen steps are discussed in the following section.
1. Collecting
Signals, which are stored as data in di?erent systems are collected in a separate system. When considering
social business intelligence, access to di?erent social media platforms – like Twitter and Facebook – is
required in order to collect the data. Each social media platform has its own method of storing data, and
not each platform is as publicly accessible as the other.
2. Filtering
Only signals that contribute to the deduction of information pass the ?ltering process. Data that are
outdated or of poor quality are removed. Especially when applying social media data for BI purposes,
this step requires much e?ort. The data can consist of spam, polluting the data.
3. Combine
The data that is collected and ?ltered on separate systems are combined and integrated into one single
source, so that analyses are based on one ‘version of the truth’.
4. Aggregating
Detailed data are aggregated to a level so that users can quickly understand the data and ?nd information.
5. Visualising
In order to make the data quickly interpretable for the users, the data is visualised.
The ?rst ?ve steps consist of automated activities that convert signals into information. So far, the signals are
translated into information that is now interpretable for the users. The next ten steps of the process consist of
non-automated activities that involve humans to interpret the information, and act on it.
2-4 Sub Conclusion: How Business Intelligence is Applied 25
6. Interpreting
The information generated by the ?rst ?ve steps are interpreted by humans. For example, the automated
process generated a graph showing the amount of sales in a given region over a given time span. The
meaning of the graph is interpreted by the user.
7. Internalise
In this step, the information derived from the interpreting step is combined with other information in the
problem’s context. It is in this step that the real underlying trends and explanatory factors are analysed
so that the information is embedded in one’s cognitive understanding of the system.
8. Revise & Recalibrate
The new information may a?ect existing information. This step ensures that existing information is
revised and adjusted based on the new information.
9. Verify
This step veri?es the new information with other mechanisms. For example, a decrease in market share
is compared with the companies’ turnover development. Whenever turnover increases while market share
decreases, it may indicate an increase of the overall market. If such mechanisms contradict, the process
of turning the data into information has to be checked for errors.
10. Enrich
In this step, the information – graphs, ?gures, numbers, etc. – are enriched by textual explanation of the
information. A decrease in market share, which is visible in e.g. a pie chart, may be enriched by a textual
explanation of two new competitors on the market.
11. Share & Communicate
By sharing and communicating about the information with other members of the organisation, the
information is brought under submission of various perceptions and views.
12. Remember
Some information do not require immediate action. However, the information may be relevant whenever
future information is acquired. It is therefore important that the information is remembered.
13. Decide
This step involves the reaction on the information. Managers decide how they act on the information, for
instance by launching an advertisement campaign, or to sell a part of the organisation.
14. Distribute
The decision in the previous step is generally taken by managers on higher levels of the organisation. The
new information and decisions following from that information are distributed to the right persons in the
organisation in this step.
15. Anticipate on Changes
The new information may be of a negative character, requiring (structural) organisational change. An
organisation should adopt a positive attitude to change according to the new information.
These ?fteen steps describe how a signal is generally translated into information at which managers can act.
When an organisation intends to implement a (sub) system that extracts signals from social media platforms
to derive information, it should be designed according to this method of processing signals.
2-4 Sub Conclusion: How Business Intelligence is Applied
With the rise of a new data source – social media platforms – for ?rms to access customer perceptions, the
question rises how a ?rm should process these data. The process should in any case correspond with existing
business intelligence processes in ?rms. Therefore, it is essential to understand the general business intelligence
process that ?rms adhere to. This chapter reviewed literature in the ?eld of business intelligence, of which the
conclusions are presented in this section.
Though there exist many views on business intelligence, the common aspect is that BI collects and translates
data into information that supports managerial decision-making. BI can be regarded as a process of three
steps; registering data, processing the data into information and reacting on the conclusions derived from
26 Conceptual Frame of Research
G
D
A
Register
React
Process
STRATEGY
BUSINESS MODEL
KEY PERFORMANCE
INDICATORS
Values, mission,
vision, objectives,
goals, plans
Performance
metrics
Revise
strategy
External data (e.g. social media data)
Internal data (e.g. level of inventory)
Stakeholders beliefs,
perceptions, values
External factors
Business Intelligence Process
Strategy Alignment
S
h
o
r
t
-
t
e
r
m
f
i
n
a
n
c
i
a
l
r
e
s
u
l
t
s
C
u
s
t
o
m
e
r
r
e
l
a
t
i
o
n
s
E
m
p
l
o
y
e
e
r
e
l
a
t
i
o
n
s
O
p
e
r
a
t
i
o
n
a
l
p
e
r
f
o
r
m
a
n
c
e
P
r
o
d
u
c
t
a
n
d
s
e
r
v
i
c
e
q
u
a
l
i
t
y
A
l
l
i
a
n
c
e
s
S
u
p
p
l
i
e
r
r
e
l
a
t
i
o
n
s
E
n
v
i
r
o
n
m
e
n
t
a
l
p
e
r
f
o
r
m
a
n
c
e
P
r
o
d
u
c
t
a
n
d
s
e
r
v
i
c
e
i
n
n
o
v
a
t
i
o
n
C
o
m
m
u
n
i
t
y
Figure 2-6: Strategy and Business Intelligence
that information. It is essential that ?rms measure those activities that contribute to its business model and
corresponding strategy, because ‘what’s get measured, get’s done’. The key notion that should be concluded
from this chapter is that an organisation’s strategy should be based on the needs and preferences of the ?rm’s
stakeholders, and that a company’s strategy drives the values, objectives, goals and plans of company, which, in
turn, determine the (key) performance indicators. Therefore, a link between the BI process and an organisation’s
strategy is required. This link is established by de?ning key-performance indicators that are based on the ?rm’s
business model. In turn, the business model should re?ect the organisation’s strategy. This perspective on
business intelligence is schematically presented in ?gure 2-6.
Though each ?rm executes its own speci?c strategy and will consequently measure organisational performance on
speci?c KPIs, we can classify KPIs into ten categories; short-term ?nancial results, customer relations, employee
relations, operational performance, product and service quality, alliances, supplier relations, environmental
performance, product and service innovation and community. Furthermore, ?gure 2-6 illustrates that the BI
cycle can not commence before the ?rm has determined ‘what it should measure’.
Chapter 3
Research Domain
The purpose of this chapter is to describe the research domain. As illustrated in chapter 1, this thesis intends
to develop a procedure that allows ?rms to process social media data for business intelligence purposes. It is
therefore that this chapter explains what social media is, how ?rms currently apply social media, for which
purposes, what social business intelligence is and what developments currently take place in the world of social
business intelligence. Since social media is one of the many applications enabled by Web 2.0, we start with a
description of Web 2.0.
3-1 Web 2.0
Web 2.0 is the generation of web pages that not only provide information, but additionally allow users to interact
with these web pages. In contradiction with the ?rst phase of the web’s evolution, Web 2.0 allows anyone to
create and share content. The content-creating feature makes it that Web 2.0 is also referred to as “the wisdom
web, people-centric web, participative web, and read/write web” (Murugesan, 2007). Web 2.0 allows users to
do more than just retrieve information. Whereas the internet was traditionally applied to read, watch, and buy
products in Web 1.0, it is increasingly utilised to create, modify, share, and discuss content in the Web 2.0 era.
It is the enabling of the creation of user-generated content that distinguishes Web 2.0 from Web 1.0. O’reilly
(2007) – who sees Web 2.0 as “the web as platform” – indicates that the power of Web 2.0 is “collective
intelligence”, and turns the web into a kind of global brain”. The best-known example is probably Wikipedia
(launched in 2001), an online encyclopedia created by the Internet users that contains now 23 million (Wikipedia,
2012) articles. The existence of Wikipedia illustrates that people are willing to share their knowledge with others.
For example, people that searched for ‘Hasbro’s Easy-Bake Oven’ in Web 1.0 would have found a static web
page promoting the product, while in the Web 2.0 era, people also ?nd in the top 5 of search results a warning
that the Easy-Bake Oven may lead to serious burns on hands due to a poorly-designed oven door (A. M. Kaplan
& Haenlein, 2010). These warnings have been written by users feeling that they had to share their (negative)
experiences and product knowledge. Other examples of collective intelligence are ask and answering sites,
demonstration videos, and product reviewing sites. Not only knowledge is shared with the rest of the world
through Web 2.0 solutions, people also are willing to share their opinions about various topics, their favourite
dining places, what they think about the new election candidate, etc. Especially these topics, that we can
position under the denominator of opinions and private events, are shared through means of social media.
3-2 Social Media
Social media allow users to connect and share content with each other through Web 2.0 based platforms. The
rise of social media in the ?rst decade of the 21
st
century is a natural consequence of Web 2.0. The relation
with Web 2.0 is highlighted by Kaplan & Haenlein’s (2010) de?nition of social media, stating that “social media
is a group of internet-based applications that build on the ideological and technological foundations of Web 2.0,
and that allow the creation and exchange of user-generated content”. However, with this de?nition we cannot
distinguish social media from Web 2.0 sites, since user-generated content is a fundamental element in Web 2.0
28 Research Domain
anyhow. Why is it that we refer to social media when we talk about Facebook or Twitter, but to Web 2.0 when
considering Wikipedia? Kietzmann et al. (2011) state that social media “allow individuals and communities to
share, co-create, discuss, and modify user-generated content”. Again, we see the importance of user-generated
content when de?ning social media. But Kietzmann et al.’s (2011) de?nition contains an important additional
component that allows us to distinguish social media from Web 2.0; communities.
Within social media – and especially in social networking sites – users can connect with other users, so that they
can share (personal) information. It is the aspect that allows people to connect with each other that distinguishes
social media from Web 2.0. The social element, connecting with other users, is much more existent in social
media than in Web 2.0. Whereas in Web 2.0, user-generated content is accessible to anyone, in social media
people can restrict this accessibility to people they have selected beforehand. Therefore, we de?ne social media
as Web 2.0 based applications that allow users to create and share user-generated content with pre-selected
users and communities.
3-2-1 Social Media Platforms
The web applications through which users can connect and share content with each other are called social media
platforms. There are many social media platforms available, and the range of social media platforms is vast and
growing (A. N. Smith et al., 2012; A. M. Kaplan & Haenlein, 2010; Hanna et al., 2011). These platforms di?er
in scope and functionality. In turn, there is variation in how people use these platforms. “Some sites are for
general masses, like Friendster, Hi5 and Facebook. Other sites, like LinkedIn, are more focused on professional
networks. Media sharing sites such as MySpace, YouTube, and Flickr concentrate on shared videos and photos
(Kietzmann et al., 2011).
In order to support managers in understanding social media, and to select the right platform for the ?rm’s
purpose, researchers have tried to classify the di?erences between the social media platforms. Weinberg and
Pehlivan (2011) distinguish social media platforms based on two dimensions; (i) half-life of information and (ii)
information depth. The half-life of information refers to the “longevity of the information in terms of availability
/ appearance on the screen and interest in a topic. Depth of information refers to the richness of the content,
and the number of diversity of perspectives” (Weinberg & Pehlivan, 2011). As such, Weinberg and Pehlivan
(2011) positioned popular social media platforms in their framework (?gure 3-1). Micro-blogs, such as Twitter,
allow users to use a limited number of character in each post and therefore have a shallow information depth. On
the other hand, community sites purposed to extensively discuss topics among users have a higher information
depth.
Blogs
(e.g. WordPress)
Communities
(e.g. MacRumors)
Micro-Blogs
(e.g. Twitter)
Social Networks
(e.g. Facebook)
Long
Short
Shallow Deep
Half-life of information
Information
Depth
Figure 3-1: Social media by information half-life and information depth (Weinberg & Pehlivan, 2011).
Kietzmann et al. (2011) created a framework that distinguishes social media platforms based on seven building
blocks. These blocks are “constructs that allow us to make sense of how di?erent levels of social media
functionality can be con?gured”. The functional building blocks are shortly discussed below, and are applied
3-2 Social Media 29
on Facebook, LinkedIn and Twitter in ?gure 3-2 (page 30) to illustrate di?erent focus points on di?erent social
media platforms.
1. Identity
This block represents the extent to which users reveal their identities in a social media setting. Especially
on self-branding platforms, such as LinkedIn, identity is a strong aspect.
2. Conversations
This block represents the extent to which users communicate with other users in a social media setting.
Some sites are much more intended to facilitate conversations – like Twitter – than others.
3. Sharing
Sharing represents the extent to which users exchange, distribute, and receive content. Especially on
Twitter, people share what they are doing, what they think of, etc.
4. Presence
The presence block represents the extent to which users can know if other users are accessible. It includes
knowing where others are, like ‘check-ins’ at Facebook or Foursquare.
5. Relationships
This block represents the “extent to which users can be related to other users”. Related implies “some
form of association that leads them to converse, share objects of sociality, meet up, or simply just list each
other as a friend or fan”.
6. Reputation
Reputation is the extent to which users can identify the standing of others.
7. Groups
The groups functional block represents the extent to which users can form communities and
sub-communities.
In the following sections we discuss three important social media platforms that are part of the analysis in this
thesis, Twitter, Facebook and Blogs.
Twitter
Twitter is recognised as being the site on which users ask information and complain. Twitter is a micro-blogging
site, designed to let people post short – 140 character – text updates called ‘tweets’ to others. Twitter prompts
users to answer the question ‘what are you doing?’, leading to a constantly updated timeline of short messages
that range from humor, opinions, musings on life to links and breaking news. Kietzmann et al. (2011) argues
that Twitter posts are “mostly short status updates of what users are doing, where they are, how they are
feeling, or links to other sites”. Participants choose Twitter accounts to ‘follow’ in their stream, and they each
have their own group of ‘followers’. Unlike social networks like Facebook and LinkedIn, where a connection is
bidirectional, Twitter has an asymmetric network infrastructure of followers. The site was launched in 2006,
and broke into the mainstream in 2008 – 2009, when accounts and media attention grew exponentially (Marwick
& Boyd, 2011)”. In February 2012, Dugan (2012) announced that Twitter had over 500 million users registered.
Twitter is an important phenomenon from the standpoint of its incredibly high number of users.
According to Jansen et al. (2009), around 20% of all tweets contain mention of a brand. Of these brand-related
tweets, nearly 20% express a brand sentiment, of which 50% were positive, and 33% were critically. In 2010, the
number of Twitter followers per ?rm increased by 241% over the year (Kirtis & Karahanb, 2011). Acknowledged
by A. N. Smith et al. (2012), Twitter posts contain more brand-related information than Facebook and YouTube.
Since the purpose of this thesis is to contribute to the development of social business intelligence in ?rms, Twitter
is a social media platform that is part of the analysis.
Facebook
Facebook is the absolute number one social networking site. Though it was only founded in 2004, it is ranking
second in the most popular websites in the world these days. In July 2012, the website reported to have 955
million monthly active users (Sloan, 2012) who log on at least once every 30 days. Half of these active users
30 Research Domain
Facebook
PRESENCE
RELATIONSHIPS SHARING
IDENTITY
REPUTATION CONVERSATIONS
GROUPS
LinkedIn
PRESENCE
RELATIONSHIPS SHARING
IDENTITY
REPUTATION CONVERSATIONS
GROUPS
Twitter
PRESENCE
RELATIONSHIPS SHARING
IDENTITY
REPUTATION CONVERSATIONS
GROUPS
Figure 3-2: Di?erent social media serve di?erent purposes. Based on Kietzmann et al. (2011).
3-2 Social Media 31
log on every day (Laroche, Habibi, & Richard, 2012). Facebook users can “create pro?les featuring personal
information, interests, photos, and the like, and can “friend” other site users. They can also participate in a wide
range of activities such as writing on friends’ walls, commenting on links, participating in forum discussions,
and “liking” brands. Facebook allows people to build or maintain social capital, communicate with others, keep
up with other peoples’ lives, and discover rumours and gossip” (A. N. Smith et al., 2012).
Over 2010, the percentage of corporations active on Facebook increased by 13%, with the number of ‘likes’
per page rising by 115% globally (Kirtis & Karahanb, 2011). Where Twitter is considered as a platform for
companies to communicate instantly with stakeholders, Facebook is suitable for creating communities among
stakeholders.
Blogs
While Twitter and Facebook conversations are often unstructured and brief, blogs may be a source for
organisations to discover structured customer opinions. After a slow start in the late 1990s, weblogs (“blogs”)
– such as Blogspot and Wordpress based websites – have become very popular, because they are easy to create
and to maintain (Kietzmann et al., 2011). Blogs are often designed as product review sites, where customers
can share their experiences with their products. Generally these blogs are publicly accessible and there are no
restrictions to the amount of characters per post.
3-2-2 User-Generated Content on Social Media Platforms
Social media platforms exist by the virtue of user-generated content (“UGC”). UGC is content that is “publicly
accessible, created outside of professional practices and shows a certain amount of creative e?ort” (A. M. Kaplan
& Haenlein, 2010). UGC can take many forms, such as pictures on Facebook, videos on YouTube, statements
on Twitter, product experiences on blogs, etc. Without users that create and share content, social media sites
are like an empty shell. Therefore, UGC is considered as the fundamental element underlying social media
(A. N. Smith et al., 2012; Boyd & Ellison, 2007). Enabled by Web 2.0, user-generated content has become
increasingly popular on the internet since the early 2000s: more and more users participate in content creation,
rather than just consumption (Agichtein et al., 2008). In China, the percentage of internet content that is
user-produced now exceeds that what is professionally produced (A. N. Smith et al., 2012). According to
A. M. Kaplan and Haenlein (2010) it are not only the technical developments of Web 2.0 and an “increased
broadband availability and hardware capacity” that has contributed to the popularity of UGC on social media
these days, but also “the rise of a generation of ‘digital natives’ and ‘screenagers’ with substantial technical
knowledge and willingness to engage online”.
Though UGC varies in nature between di?erent social media platforms (Kietzmann et al., 2011), A. N. Smith
et al. (2012) and Jansen et al. (2009) indicate that much user-generated content – around 20% – on the internet
contains a brand name. It is here where the opportunities for ?rms materialise, ?rms can inspect the posts that
contain a brand name to discover customer opinions related to their brands. The internet’s accessibility, reach,
and transparency have empowered ?rms that are interested in consumers opinions (Kozinets et al., 2010). User
opinions were not that easy to be gathered before the social media era, while they are now accessible at low
costs (Kirtis & Karahanb, 2011).
As illustrated in the previous paragraph, UGC on social media contains opportunities for ?rms. However,
user-generated content also contains disadvantages. Especially issues related to variance, cohesion and
veri?cation are at stake when processing user-generated content from social media sites. The three issues
are discussed in the following paragraphs.
Variance Firstly, the variance in the quality of UGC is high, “any data can contain information ranging from
excellent to spam” (Agichtein et al., 2008). This makes the tasks of ?ltering and ranking the importance of
social media posts more complex than non user-generated content.
Cohesion Secondly, professionals that base decisions on social media content should be aware of the negative
e?ects arising from cohesion; one negative message about a ?rm – which may not even be true – can snowball
over the internet, reaching many people, and may eventually harm the performance of the company. “Cohesion
describes the phenomenon that evaluations of cost and bene?t associated with prospective behaviour are
aligned via strong communication relationships. People thus become more homogeneous as a result of direct
32 Research Domain
contact via social networking links from node to node. This type of social contagion is typically referred to as
word-of-mouth” (Takac, Hinz, & Spann, 2011).
Cohesion can work out positive as well as negative for an organisation. The negative characteristic materialises
whenever a negative message circulates along the social network. This message is likely to in?uence the
perceptions of the organisation in a negative manner. On the other hand, cohesion may o?er opportunities.
Whenever an organisation intentionally in?uences the discussions on social networks in a positive manner, the
message is likely to be adopted by a large community.
Veri?cation Thirdly, the providers of the information in the social media world generally spread information
without veri?cation, unlike the traditional mass media. Dong-Hun (2010) argues that social media is not yet
capable of replacing the traditional media because of the credibility problem. However, Wikipedia is a successful
example of a website that is based on trust, and established and maintained by the crowd. Information posted
on the website is veri?ed among other users, that directly renovate incorrect information.
The identi?ed threats related to social media data should be considered when developing and implementing a
social business intelligence system.
3-2-3 Current Applications of Social Media in Firms
Both large and small organisations are increasingly visible on social media platforms. In addition, managers
“sense that social media is and will remain an important fabric of commerce” (Weinberg & Pehlivan, 2011).
A Burson-Marsteller research investigated the application of the platforms Twitter, Facebook, YouTube and
Corporate blogs, and found that 25% of the ?rms actively use all four social media platforms, 84% uses at least
one of them (Kirtis & Karahanb, 2011). Of the Fortune 2000 companies, 69% currently use social networking
sites, while 37% planned to use more of them over the next ?ve years (McCorkindale, 2010).
What does it mean when ?rms ‘use’ social media? IBM (2011) researched for which activities ?rms applied social
media, the results are shown in ?gure 3-3. Although IBM (2011) provides detailed insight in the many social
media activities that ?rms employ, we can conclude that ?rms generally apply social media to communicate
with customers, promote activities, monitor the brand name and inspect customer ideas. These four activities
are discussed in the following paragraphs in more detail.
27%
35%
37%
38%
40%
40%
41%
43%
43%
46%
46%
48%
50%
52%
60%
65%
74%
Vendor or partner communications
Customer-to-customer interactions
Training/education
Experts insights/though leadership
Solicit customer ideas
Provide support
Employee-to-employee interactions
Customer research
Recruit employees
Capture customer data
Brand monitoring
Solicit customer reviews
Sell products/services
Generate sales leads
Promote events
Respond to customers questions
Communicate with customers
27%
35%
0% 10% 20% 30% 40% 50% 60% 70% 80%
Vendor or partner communications
Customer-to-customer interactions
% of Respondents applying the social media activity n =351
Figure 3-3: Applications of social media (IBM, 2011)
Marketing
Web 2.0, and especially social media, has empowered the ‘voice of the customer’. Consumers are no longer merely
passive recipients in the marketing exchange process (Hanna et al., 2011). In the past, marketing campaigns
3-2 Social Media 33
were typically developed by companies in-house, without interference of (potential) consumers. Campaigns had
the character of ‘here is the advert, please absorb it’, or ‘here is the product, we hope you like it’. This ‘we talk,
you listen’ approach has been replaced by ‘you talk, we listen’ as a result of the possibilities o?ered by social
media (e.g., Patterson, 2012; Klassen, 2009).
Research indicates that marketing through social media is e?ective for ?rms. “70% of the consumers that have
used social media websites to take product or brand information, 49% of these consumers made a purchase
decision based on the information they pound through social media sites” (Kirtis & Karahanb, 2011).
Firms are also turning into social media marketing to lower the ?rms’ expenditures. The cost reduction aspect of
online marketing as compared to traditional marketing is one of the main reasons why companies are nowadays
applying social media for marketing purposes (Kirtis & Karahanb, 2011). Cost reduction is mainly achieved by
the elimination of the distribution phase, which is required in traditional mass media. In addition, marketing
through means of social media is less expensive than the regular channels because most social media applications
are free of charge. As such, the biggest expenditures related to the execution of a social media strategy represents
the time employees spend on posting messages, responding to comments and blogging. Social media allows
marketers to speci?cally target on client groups, and distinguish between products / services case by case. In
comparison with traditional marketing channels, social media shows also on this aspect lower costs. Driven by
the global recession, many ?rms are in a cost-reducing mode. Because of the economic turmoil social media is
applied as a survival tool by many ?rms, so the economic recession has increased the rate of shift change from
traditional media to social media (Kietzmann et al., 2011).
For the bene?ciary reasons of marketing through social media described in the previous, A. N. Smith et al.
(2012) estimate that the percentage of companies using social media for marketing is expected to reach 88% by
2012, up from 42% in 2008.
Customer Relations Management
Firms also use social media for customer relations management (“CRM”), also referred to as social CRM. Thanks
to social media, “the nature of public relations and how organisations engage their public has changed a great
deal in the past several years” (McCorkindale, 2010). “An environment in which control of the relationship
has shifted to the customer, who has the power to in?uence his or her social network” (IBM, 2011) drives
organisations willing to participate in the online conversations.
Social media platforms hold unprecedented opportunities for companies to get closer to customers, allowing ?rms
to communicate directly with customers, for instance to provide support when customers encounter problems
with products / services. According to Patterson (2012), ?rms “have made progress in conversing with their
customers”.
A recent study by Laroche et al. (2012) indicates that it is bene?ciary for ?rms to establish online communities
in which both ?rms and customers communicate with each other, “brand communities established on social
media have e?ects on customer/product, customer/brand, customer/company and customer/other customer
relationships, which in turn have positive e?ects on brand trust, and trust has positive e?ects on brand loyalty”.
Reputation Management
A failing social media engagement strategy can signi?cantly impact a ?rm’s reputation and sales (Kietzmann
et al., 2011). The increased application of social media has serious consequences for an organisation’s exposure
to its environment. It seems that the power has been taken from the corporate marketing departments by
individual consumers that create, share and discuss online blogs, tweets, Facebook entries, movies, pictures, etc.
With or without permission, communication about brands will happen. In an environment where customers gain
more and more power, organisations need to carefully tread their actions and control its’ exposure. Therefore,
companies more and more empower employees to talk, listen, and respond to what consumers post on social
media (A. N. Smith et al., 2012) in order to control the ?rm’s (online) reputation.
One negative message about an organisation – created by one single person – can snowball over the internet,
reaching many people, and may eventually harm the performance of the company. In 2008, Canadian singer
Dave Caroll wrote a song about United Airlines’ luggage handling employees recklessly throwing his guitar,
which caused a break in his guitar. Frustrated by bad customer experience, he uploaded his ‘United breaks
guitars’ song on YouTube. Consequently, United Airlines experienced a 10% drop (Patterson, 2012) in its share
value and su?ered damage to its reputation. The YouTube clip has been viewed over 12 million times. This is
34 Research Domain
one of the examples that show how powerful the force of social media can be, when a company does not act
according to the preferences of the community. As such, social media platforms may be a source of both threats
and opportunities for brands experiencing unfavourable exposure (A. N. Smith et al., 2012).
Co-creation & Pro-sumers
Today, consumers “are taking an increasingly active role in co-creating everything from product design to
promotional messages” (Berthon, Pitt, McCarthy, & Kates, 2007). This phenomenon is known as co-creation,
and more recently termed as “prosuming” (DesAutels, 2011), illustrating that people are not only consumers
but at the same time producers. Firms are much more required to perceive consumers as partners in the process
of creating products, whereas this was – before the social media era – formerly an activity for solely the ?rm.
An example of such a process is Lay’s recent campaign to decide the new ?avour of their potato crisps. In
Lay’s campaign, consumers were stimulated to contemplate new ?avours and to post these ideas on the web.
Other users consequently rated the ideas that were send it. The winning ?avours have actually been brought to
production. Another example related to co-creation is Samsung, which ‘listened’ closely to the user-generated
content on blogs, and, after hearing complaints that the speakers on the side of the TV were too wide for many
customers’ entertainment cabinets, it redesigned the product (Klassen, 2009).
The co-creation opportunities for ?rms o?ered by social media reach even further. A growing number
of organisations, among them 3M, AEGON, HCL Technologies, Red Hat and Rite-Solutions have recently
experienced with crowdsourcing their strategies (Gast & Zanini, 2012). The organisations o?ered the public the
possibility to provide input in the form of proposals for the company’s future directions. The e?ects resulting
from strategy crowdsourcing is twofold. In the ?rst place, the company gathers information from the external
environment, including perceptions from important actors that would normally be overlooked. An organisation
can consequently craft its strategy with a higher quality. Secondly, the organisation creates “enthusiasm and
alignment behind a company’s direction” (Gast & Zanini, 2012).
Though the previous sections illustrate that social media is widely applied for di?erent purposes in organisations,
many executives eschew or ignore this form of media because they “don’t understand what it is, the various
forms it can take, and how to engage with it and learn” (Kietzmann et al., 2011). Also A. M. Kaplan and
Haenlein (2010) state that the reluctant attitude of some managers towards social media is due to “a lack
of understanding regarding what social media are”. Many organisations acknowledge the opportunities in the
application of social media, while, on the other hand, there also exists a fair degree of uncertainty with respect
to allocating e?ort and budget to social media, and “limited understanding of the distinctions between various
social media platforms” (Weinberg & Pehlivan, 2011).
3-3 Social Business Intelligence
Firms should measure the e?ects of social media activities on organisational performance. As illustrated in
chapter 2, the process of business intelligence requires key-performance indicators to be de?ned so that the
performance of the ?rm can be measured against its strategy. This value-based management approach is
generally applied within ?rms, implying that when a ?rm pursues to perform social media activities, it should
measure the e?ects of these activities in relation with organisational performance.
Existing social media monitoring tools mainly reveal the performance of the organisation on social media
(number of mentions, number of likes, % of positive mentions), and treat the social media component of a ?rm
as a separate business unit executing its own strategy. However, the purpose of business intelligence is to reveal
the underlying parameters that determine the performance of the organisation, that is, not limited to solely
social media performance. In order to understand the in?uence of social media content on an organisation’s
performance, a link between the company’s key-performance indicators and clear social media parameters is
required.
It is argued that the possibilities of social media for business intelligence purposes reaches further than what is
currently o?ered by the social media analytics tools. The key bene?ts will be gained whenever the KPIs of an
organisation are linked to the parameters that are measured by social media tools. Only in that case, one can
speak about ‘social business intelligence’. In social business intelligence, the social media activities related to a
?rm are translated to organisational performance.
3-3 Social Business Intelligence 35
3-3-1 The Current State of Social Business Intelligence: Early Adoption
Software developers acknowledge the opportunities generated on social media platforms for ?rms. With the
rise of social media, and the popularity of BI within organisations, software solutions o?ering social media
‘intelligence’ are emerging rapidly. As a result, tools for analysing information become widely available at
ever-lower prices (Bughin, Chui, & Manyika, 2010), some are even o?ered for free.
Auditore (2012a) – the former head of SAP’s Business In?uencer Group and now researcher at Asterias research
– investigated the market for social business intelligence and found that the top four emerging SBI platforms
consists of Radian 6, Kapow, evolve24 and NetBase. According to Kapow (2009), a provider of business
intelligence software, we are at a point in time where social media can be integrated into enterprise business
intelligence platforms. Not only small software development ?rms are on a discovery journey, well-established
companies o?ering total business intelligence solutions are also embracing social media data. For example, SAP
collaborates with NetBase to o?er social media analytics. IBM’s Cognos provides social network capabilities.
Oracle recently acquired Involver, Vitrue and Collective Intellect to add social media analytics to their portfolio
of services. SAS incorporated social media analytics in its platform, and QVSource allows QlikView users to
extend their BI platform with social media intelligence. Table 3-1 lists the top existing, new and emerging
vendors of (social) business intelligence solutions.
Table 3-1: Top (Social) Business Intelligence Vendors (Auditore, 2012a).
Legacy BI vendors New social media BI vendors Emerging social media BI vendors
1. IBM 1. Google 1. Radian6
2. Oracle 2. SAS 2. Kapow
3. SAS Institute 3. IBM 3. evolve 24
4. SAP 4. NetBase
Emerging Social Media Business Intelligence Vendors
Companies that apply social media in their organisation generally apply a cycle consisting of three steps;
(i) monitoring, (ii) analysing, and (iii) engaging (Kapow, 2009; Bryant, 2011) using social media monitoring
platforms. The objective of these platforms is to ‘listen’, in order to monitor the brand(s). Generally,
“automated scripts monitor a handful of keywords from targeted web sites” (Kapow, 2009). The gathered
data in the listening phase is generally continued by mapping customer perceptions, sentiment measuring and
an indication of the company’s reach respecting social media. In general, these tools are solely based on “simple
quantitative counts of how many times a brand has been mentioned” (Patterson, 2012). Some exceptions exist
that provide a general mood of the brand, often based on large datasets. These functions are referred to as
analytics by the software o?erers. Clients receive weekly or daily reports containing ?gures representing the
amount of last week mentions on social media platforms, the number of likes, the sentiment related to that, the
number of shares, retweets, a distribution of the locations, gender distribution, etc. In addition, the software
platforms generally scan all social media platforms continuously and present all relevant content to the user(s),
via dashboards and/or automatic generated reports. Companies’ managers can consequently engage with the
social media users via one portal. The nature of the engagement of companies is often related to customer
relationship management, e.g. a customer-service department explaining to an individual why his or her credit
card is not functioning, or why the company’s website is not presented properly in the customer’s browser.
Other social media posts made by organisations are often marketing related, e.g. an announcement of a new
product release or an o?er.
The emerging social media intelligence tools – Radian6, Kapow, evolve24 and Netbase – and their features are
presented in the following section.
• Radian 6
Salesforce’s Radian6 provides social media monitoring tools, social media engagement software and social
customer relationship management and marketing software. It provides companies with social analytics
comprising of social media metrics and sentiment analysis. Radian6 provides ?rms with a dashboard
illustrating their performance on social media. Consequently, ?rms can engage in online discussions.
Clients of Radian6 include Fuji?lm, Commerce Bank, KLM, Pepsi, L’Oreal, Baker Tilly and Activision
(Radian6, 2012). Radian6 o?ers clients di?erent packages with di?erent features, ranging from EUR 750
per month to EUR 12,000 per month.
36 Research Domain
• Kapow
Kapow o?ers solutions for accessing, extracting and enriching web data (Kapow, 2009). The software
developer illustrates the applicability of public web data for business intelligence, it explicitly mentions
that it o?ers software that structures social media data and turns it into interpretable information. One of
the many tools o?ered by Kapow is the monitoring of social media platforms. Users of Kapow’s software
include AT&T, Intel, Cisco, Vodafone, Morgan Stanley, P&G, DHL, Barclays, Lenovo and Audi.
• evolve24
evolve24 mines, priorities and scores online conversations so that relevant intelligence is provided to the
manager of a ?rm. The software of evolve24 allows users to create custom dashboards to present those
social media metrics that are relevant to the ?rm. Next, it allows predictive modelling to predict the
impact of certain issues, so that the decision process in the organisation is supported by this information
(evolve24, 2012).
• NetBase
NetBase allows users to track social media issues related to the topics of interest. It processes billions
of social media posts to extract structured insights that enterprises can use to quickly discover market
needs and trends, quantify market perceptions about products, services, and companies (Netbase, 2012).
SAP collaborates with NetBase to for social media analytics solutions. Hence, SAP users can easily
integrate the social media analytics provided by NetBase in their existing BI platforms. Amongst others,
Tupperware, Hewlett Packard, Coca-Cola and Kraft Foods are users of NetBase.
Intelligence Provided by Social Media Monitoring Tools
Social media tools, whether they refer to themselves as monitoring, analysis or intelligence tools, o?er a variety of
insights in the performance of ?rms on social media. The novelty of social media and the even more unexplored
applicability of business intelligence on the new phenomenon makes in that there is little scienti?c literature
available in this ?eld. Instead, both large ?rms as IBM and small organisations present white papers and blogs
in which they describe their view on social media metrics. Many of these documents refer to the same variables
that they measure, though they generally adhere to their own ‘invented’ name. Common social media metrics
and intelligence that are provided by the social media monitoring tools are presented below.
• Volume of Posts
The volume of posts measures number of messages or articles that have been created on social media for
a speci?c topic over a given period. The volume of created social media posts containing a ?rm’s name
(or product / service name) illustrates to what extent a company is subject of discussion of social media.
The volume of social media posts can vary from day to day or even from hour to hour. A sudden upwards
deviation in the average volume is a signal for a ?rm that people are paying attention to the ?rm, whether
positive or negative. Firms can relate this ?gure to marketing campaigns or other organised events to
determine the success of their reach.
• Engagement
Engagement represents the involvement of users in the brand. Often, companies measure engagement by
the amount of likes, followers, shares, retweets, etc. However, solely looking at this ?gure is not enough.
After all, it is relatively easy to in?uence this ?gure, for instance by organising a lottery in which users can
win an iPad. Doeland (2012) distinguishes engagement metrics into distribution metrics and interaction
metrics. Distribution metrics describe how well the organisation is visible to the social media public, while
interaction metrics represent how well the public engages in the brand.
• Sentiment
Almost all software tools – even those available for free – o?er sentiment analysis, a measure that represents
the attitude of the content generated by the social media users. Generally, social media posts are classi?ed
as either positive, neutral or negative by linguistic algorithms. These algorithms ‘simply’ textmine each
post associated to the organisation and consequently connect words and phrases like ‘great’, ‘wow’, ‘good’,
‘
’, ‘super’, etc. with a positive attitude. Posts containing words like ‘bad’, ‘dumb’, ‘worthless’, etc. are
classi?ed as negative posts. As such, an indication of the sentiment under the social media users is
generated. Figure 3-4 illustrates the output of a sentiment analysis as it is provided to a user of a social
media monitoring tool, in this case uberVU, one of the popular social media monitoring tools.
Sentiment analysis is a complex activity. Not only because each language requires its own meta data to
classify words and phrases in di?erent languages, but also because most of the sentiment analysis tools
3-3 Social Business Intelligence 37
Table 3-2: Examples of Engagement Metrics
Distribution Metrics Interaction Metrics
Followers Retweets
Fans Forwarding
Mentions Sharing
Reach Comments
Bookmarks Likes
Inbound links Rates
Blog subscribers Reviews
Contributors
Tra?c generated
Time spent on site
Response time
Sentiment
Sentiment refers to whether the tone of the conversation where the keyword was mentioned was
positive, neutral or negative. As a simplified example, "I love Apple" is considered positive towards
"Apple" whereas "I bought an IPad yesterday" is neutral. We use one of the most powerful fully
automated sentiment engines on the market.
Daily Sentiment breakdown shows the number of positive, neutral and negative mentions each day.
Main Negative Themes are the topics that people talk about negatively when mentioning the
keyword.
So for the mention "I hate Apple support" a negative theme is "support"
AVERAGE SENTIMENT
Slightly POSITIVE
27% positive
SENTIMENT BREAK-DOWN
43.3 43.3% positive % positive 40.3 40.3% neutral % neutral 16.2 16.2% negative % negative
DAILY SENTIMENT BREAKDOWN
Jul 27 Jul 28 Jul 29 Jul 30 Jul 31 Aug 01
2k
4k
Figure 3-4: Sentiment Analysis Example
use Natural Language Processing techniques. These techniques assume that the underlying text is “clean
and correct” (Dey & Haque, 2008), a requirement that is not always present in social media posts. Social
media posts comprise spelling errors, ad-hoc abbreviations and improper casing, incorrect punctuation
and malformed sentences. These features pollute the outcome of the algorithms. However, “interest in
noisy text analytics has increased signi?cantly in the recent past” (Dey & Haque, 2008). The systems
that are currently developed also take phrases into account (Agichtein et al., 2008), turning “Wow, the
new product of ABC is really great.. NOT!” into a negative sentiment post. As such, the accuracy of
sentiment analysis is expected to increase by new methods that are currently developed. Most platforms
are commercial and do not disclose full details of their internal feature set.
• Geography
Whenever a person registers itself for a social media platform, he or she is required to ?ll up some personal
information, including the person’s residence. Though it is not guaranteed that users provide legitimate
personal details, social media monitoring tools use this information to determine the location of where
the posts has been made. Next, mobile devices including a GPS component can – if allowed by the
user – provide the social media post with more accurate geographic information. As such, social media
monitoring tools provide details about the geography of the social media posts of a ?rm in a given period.
Figure 3-5 shows an example of the output of a social media geography analysis.
• Topic and theme detection
Social media monitoring tools provide details in the primary topics and themes that consist in the dataset
related to the ?rm. Generally, a list of the ten most ‘trending topics’ is presented. Topic and theme
detection allows ?rms to quickly grasp an understanding of the most discussed topics that consist in the
social media posts related to the ?rm.
• In?uencer ranking
38 Research Domain
Geolocation
Geolocation represents the countries where people talked about the keyword during the selected
time period and the respective share of the conversation. The location of a person is determined by
using mostly Twitter and Facebook data and other profile or location data where available. The
darker the green, the more conversations have taken place about this keyword in that region of the
world.
GEOLOCATION HEAT MAP
TOP LOCATIONS
JAPAN 17%
UNITED STATES 11%
BRAZIL 8%
INDONESIA 7%
ITALY 6%
TURKEY 4%
SPAIN 4%
MEXICO 3%
SOUTH KOREA 3%
UNITED KINGDOM 3%
TOP LANGUAGES
SPANISH 35%
ENGLISH 22%
PORTUGUESE 10%
JAPANESE 8%
ITALIAN 4%
1 6637
Figure 3-5: Geographic Analysis Example
Almost all social media platforms – and especially social networking sites – provide the possibility to
follow other users. As a consequence, the messages that have been created by people with many followers
will reach many other users. Social media monitoring tools provide insight in the amount of followers
of the people that posted a message containing the ?rms names. The users with the most followers are
considered the strongest in?uencers.
It would be wise to combine the sentiment of the posts with the posts made by the strongest in?uencers.
A negative message created by a strong in?uencer is likely to reach many people, which may result in an
overall decrease of the sentiment towards the ?rm. On the contrary, strong in?uencers posting positive
messages may increase the overall sentiment. With this information, web care teams can focus on the
people with many followers in order to have the strongest e?ect on the desired result, which may be an
increase in the overall sentiment.
• Channel distribution
In order to understand on which social media platforms ?rms are subject of discussion, social media
monitoring tools provide insight in the distribution of the posts related to the ?rm across the di?erent
platforms. With this insight, ?rms can decide to focus on those platforms where their ?rm is subject
of discussion. Figure 3-6 shows an example of the output provided by a social media monitoring tool
(uberVU) illustrating the distribution of social media posts in a certain period across various channels.
A 2012 research by IBM and SHARE-Unisphere amongst 711 business and IT managers from across the
world revealed that “72% of the respondents – ?rms – are monitoring social media networks, re?ecting great
awareness of the importance of understanding information ?ow and engaging social media networks”. The
most mentioned business functions employing social media include “sales and marketing (64%), public relations
and communications (38%), IT (37%) and customer service (37%)” (Auditore, 2012b). Table 3-3 lists the top
business initiatives and the parameters that were measured (Auditore, 2012a). The research also indicated that
investments in the area of social business intelligence “continue to trend upward ... 60% of the respondents
indicated that they expect to increase social media monitoring over the next 1–2 years, while 21% indicated it
would be 3–5 years. However, the study shows, managers are unclear about the ultimate usefulness of social
media. This re?ects that social business intelligence is still in an ‘early adopter’ phase. The study concluded
that “social media based business intelligence represents the next great frontier of data management, promising
decision makers vast vistas of new knowledge gleaned from exabytes of data generated by customers, employees,
and business partners” (Auditore, 2012b).
3-4 EU Legislation on Social Media Data Processing 39
Share of voice
This metric represents the breakdown of mentions about the keyword by specific platforms. The
breakdown is based on total number of mentions per platform. This is important when you're trying
to figure out where most of the conversation is happening and where you should focus your
listening and engagement efforts.
Top stories on the top platforms provides a sense of what people are mostly talking about regarding
the keyword on each individual platform.
PLATFORM DISTRIBUTION
Flickr
3407 mentions
Twitter
1439 mentions
Blogs
60 mentions
Facebook
59 mentions
News
9 mentions
Youtube
3 mentions
Boards
1 mention
Vimeo
1 mention
Figure 3-6: Channel Distribution Example
Table 3-3: Top Business Initiatives for Social Media and Measured Parameters (Auditore, 2012b).
Top business initiatives Top metrics employed
1. Brand-reputation management 1. Customer satisfaction
2. Marketing communications 2. Overall chatter
3. Customer service 3. Brand experience
4. Customer experience management 4. Advertising campaign performance
5. Sales
6. CRM
3-4 EU Legislation on Social Media Data Processing
Social networks have obtained a “poor reputation for protection users’ privacy due to a continual ?ow of media
stories discussing privacy problems” (Bonneau & Preibusch, 2010). Examples of such stories include “disclosure
of embarrassing personal information to employers and universities, blackmail using photos found online and
social scams” (Bonneau & Preibusch, 2010). The European Commission is of the opinion that social networks
are a useful tool for staying in touch with friends, family and colleagues, but that these networks also present
a risk that personal information, photos and comments might be viewed more widely than people realise. The
Commission also states that in some cases this can have ?nancial, reputational, and psychological consequences.
Currently, legislation in the European Union’s member states on data privacy is based on the Data Protection
Directive 95/46/EC. This Directive has been established in 1995, a period in which Web 2.0 and social networks
did not exist. The technological developments and the scale of data sharing and collecting have increased in
recent years. Given the advances in IT, the Commission deems Directive 95/46/EC outdated. In addition,
as with any Directive, all member states have composed national legislation based on the Data Protection
Directive, implying that each member state applies its own Data Privacy Policy. E.g, in the Netherlands this
resulted in the Personal Data Protection Act
1
in 2001. It is therefore that the Commission drafted a proposal
in January 2012 for Regulation on the protection of individuals with regard to the processing of personal data.
This new legislation takes the social media era into consideration, and is directly applicable in all EU member
states.
Directive 95/46/EC provides the basis for the de?nition of personal data, which may be contained in social
media messages. Personal data are de?ned as “any information relating to an identi?ed or identi?able
1
In Dutch: Wet Bescherming Persoonsgegevens.
40 Research Domain
natural person; an identi?able person is one who can be identi?ed, directly or indirectly, in particular by
reference to an identi?cation number or to one or more factors speci?c to his physical, physiological, mental,
economic, cultural or social identity”
2
. The processing of personal data is de?ned as “any operation or set
of operations which is performed upon personal data, whether or not by automatic means, such as collection,
recording, organisation, storage, adaption or alteration, retrieval, consultation, use, disclosure by transmission,
dissemination or otherwise making available, alignment or combination, blocking, erasure or destruction”
3
. The
Data Protection Directive is only applicable when the data can be marked as personal data.
The newly proposed Data Protection Regulation adheres to the personal data de?nition of Directive 95/46/EC.
Thus, any data that provides one the possibility to retrace a natural person from that data, is personal data.
The Commission introduces the ‘right to be forgotten’, implying that a social network user can request – if there
is to legitimate reason to store it – to remove all data related to the person from their system. Personal data
can only be collected after explicit consent of the person that provides the information. Furthermore, providers
of social media should adopt the principle of ‘privacy by default’, implying that the default settings should
be those that provide the most privacy. Social media sites should also inform users about how the personal
data will be used. The new legislation is expected to come into force in 2014, with penalties up to one million
Euro or 2% of the ?rm’s global revenue in case of a breach. In the following paragraph, we discuss how the
new legislation a?ects the possibilities o?ered by publicly accessible data for ?rms and what procedures are
necessary to be in compliance with the new legislation.
3-4-1 What Firms are allowed to do with Public Data
Firms are allowed to process data whenever these data are not personal data or whenever the creator of the
data has given prior consent to process the data. In order to avoid data to be legally labelled as personal, the
data should be pre-processed in a way that it is not possible to retrace a natural person from the data. Thus,
the data should be made anonymous. E.g., attributes containing the name of the users should be removed.
Though it is not guaranteed that persons actually use their o?cial name on social media, it is advised that ?rms
remove those attributes that may contain information allowing one to retrace a natural person from these data.
Furthermore, ?rms can aggregate the data to a level at which the individual message is not considered for their
analyses. The ‘right to be forgotten’ has consequences for the way in which social media data is stored and
distributed. With the new Regulation, any person can withdraw his or her information from a social media site.
However, social network sites distribute – by means of APIs or trough other ways – the social media messages
to third parties. It will be the social network providers that will become responsible to communicate to its third
parties that a certain user has requested to delete its content.
In order to avoid suspicions, it is advised that ?rms intending to process social media data carefully document
the steps that they undertake to make the data anonymous, and how the ‘right the be forgotten’ is enabled in
the processing of the social media data. Such procedures are to be designed so that privacy is embedded in the
procedure, known as the ‘privacy by design’ principle in the new Data Protection Regulation.
3-5 Sub Conclusion
Social media platforms are Web 2.0 based applications that allow users to create and share user-generated
content with pre-selected users and communities. There exists a variety of social media platforms, some are
aimed at relations between the users, while others are developed to share media like photos and videos. Research
indicates that of all the user-generated content on the internet, about 20% is brand-related. Users e.g. write
their opinion about a new product, complain about a service, discuss new ideas, etc. Therefore, it is interesting
to investigate the opportunities for ?rms to analyse the content that contains their brand name. However,
though user-generated social media content may be valuable for a ?rm, there also exist pitfalls in collecting,
analysing and drawing conclusions from these data. Firstly, the variance in the quality of UGC is high; any
data can contain information ranging from perfectly true to spam. Secondly, cohesion may lead to homogeneous
content; implying that one user adopts the opinion of another. Thirdly, one should be aware of the fact that
users post their messages on social media sites without veri?cation.
Firms are increasingly visible on social media. This trend is even ampli?ed by the current global recession,
bringing ?rms in cost-reduction mode. Firms generally apply social media for marketing e?orts, customer
2
Directive 95/46/EC, O?cial Journal of the European Communities. L 281/31, Article 2(a).
3
Directive 95/46/EC, O?cial Journal of the European Communities. L 281/31, Article 2(b).
3-5 Sub Conclusion 41
relations management, reputation management and to stimulate co-creation. For all these four aspects, social
media engagement is a more e?cient and inexpensive activity than the traditional channels.
G
D
A
Register
React
Process
STRATEGY
BUSINESS MODEL
KEY PERFORMANCE
INDICATORS
Values, mission,
vision, objectives,
goals, plans
Performance
metrics
Revise
strategy
Stakeholders beliefs,
perceptions, values
External factors
Business Intelligence Process
Strategy Alignment
S
h
o
r
t
-
t
e
r
m
f
i
n
a
n
c
i
a
l
r
e
s
u
l
t
s
C
u
s
t
o
m
e
r
r
e
l
a
t
i
o
n
s
E
m
p
l
o
y
e
e
r
e
l
a
t
i
o
n
s
O
p
e
r
a
t
i
o
n
a
l
p
e
r
f
o
r
m
a
n
c
e
P
r
o
d
u
c
t
a
n
d
s
e
r
v
i
c
e
q
u
a
l
i
t
y
A
l
l
i
a
n
c
e
s
S
u
p
p
l
i
e
r
r
e
l
a
t
i
o
n
s
E
n
v
i
r
o
n
m
e
n
t
a
l
p
e
r
f
o
r
m
a
n
c
e
P
r
o
d
u
c
t
a
n
d
s
e
r
v
i
c
e
i
n
n
o
v
a
t
i
o
n
C
o
m
m
u
n
i
t
y
Raw messages
Volume of messages
Sentiment
Location of messages
Trending topics
Influencer ranking
Channel distribution
Multiple sources of social media data
Social media data
Unstructured, non-
personal, external data
Engage on
Social media
Personal Data Pre-Processing
Remove or make personal data
anonymous
Figure 3-7: Strategy, Business Intelligence Process and Social Media Data
As with any activity that is performed in a professional organisation, the performance of the activity and its
contributing value to the ?rm’s overall objective is to be measured. Current social media monitoring tools
– which are evolving rapidly – do not o?er insight in the e?ects on organisational performance due to the
organisation’s social media undertakings. Rather, these tools provide the number of brand mentions on di?erent
platforms, the locations of where the posts were made, gender classi?cation, language, (unreliable) sentiment
analysis, etc. and thereby treating the ?rms’ social media activities as a separate, isolated, business activity.
It is therefore argued that ?rms need to establish a clear link between social media metrics (such as number of
likes, shares, sentiment) and the ?rm’s key performance indicators. Figure 3-7 illustrates the concept of social
business intelligence, with the types of social media data ?owing in the business intelligence process.
Chapter 4
Content Analysis
This thesis examines the applicability of social business intelligence for ?rms in di?erent contexts. More speci?c,
the applicability is investigated for ?rms in (i) di?erent industries and (ii) for di?erent customer relation types.
The purpose of this chapter is to reveal di?erences in social media content related to di?erent ?rms. As such,
social media data related to di?erent ?rms will be collected from social media platforms, after which these
data are analysed on these two dimensions. As described in section 1-6-2, we will apply the content analysis
procedure developed by Bos and Tarnai (1999) as a guidance for this analysis. A content analysis “entails a
systematic reading of a body of texts” (Krippendor?, 2004), which is required to analyse the social media posts
related to di?erent ?rms. The structure of this chapter corresponds with the ?ve step procedure, as shown in
?gure 4-1.
Research outline, research questions,
formulation of hypotheses, material to
investigate
Operationalising the categories,
determining the sample, determining the
unit of analysis
Establishment of categories
Theoretical level
Determining reliability and validating the
categories
Pretest
Appropriate statistical analyses
Data collection and evaluation
Immanent interpretation of the results,
discussion of the results on the basis of
the problem
Interpretation of the results
Section 4-1
Section 4-2
Section 4-3
Section 4-4, 4-5
Section 4-6
Figure 4-1: Content Analysis (Bos & Tarnai, 1999) and Outline Chapter 4.
Section 4-1 formulates the research questions that are to be answered by the content analysis, and it describes
the material that will be investigated. Next, section 4-2 establishes the categories that are to be analysed, it
describes the sample ?rms and the sample period. Thirdly, in section 4-3 a pretest is performed to validate
whether it is possible to collect the data and classify it into one of the established categories. If necessary, the
categories will be adjusted. Furthermore, section 4-3 describes the categories in more detail and presents the
taxonomy – or coherence – of the categories. Fourth, in section 4-4 the data – social media messages related
to di?erent ?rms – is collected and the collection process is evaluated. In section 4-5, the data is analysed and
visualised. Section 4-6 interprets the results of the content analysis. Finally, section 4-7 concludes the ?ndings
of this chapter.
4-1 Theoretical Level 43
4-1 Theoretical Level
The ?rst step in the content analysis contains theoretic notions. Hypotheses and the material to be investigated
are determined. Hypotheses are established to explicitly specify what will be researched in the analysis. The
hypotheses and the material to investigate are described in this section.
4-1-1 Hypotheses Formulation
This thesis answers three sub research questions (see page 4), of which the second question will be answered in
this chapter. From the research questions that are to be answered by means of the content analysis, hypotheses
are formulated. These hypotheses explicitly state what will be investigated. In total, four hypotheses are
formulated that correspond to the second research question of this thesis, i.e. In which contexts are ?rms able
to acquire social media data for business intelligence? As discussed in chapter 1 we specify a ?rm’s context on
two dimensions. Firstly, a ?rm’s context is described based on its customer relation type, consisting of either
B2C or B2B. Secondly, a ?rm’s context is speci?ed by the industry in which it is active. The speci?cation of the
?rm’s context is incorporated in the hypotheses that are established in this section. Two groups of hypotheses
are established; volume-related and subject-related hypotheses.
Hypotheses related to the Volume of Social Media Messages
The ?rst topic of investigation relates to the volume of ?rm-related social media messages. Particularly, the
volume of messages related to ?rms performing di?erent customer relations are compared. Throughout this
thesis, we distinguish two types of customer relations; B2C and B2B. As illustrated in section 1-2, it is expected
that ?rms performing B2C relations are more often subject of discussion on social media than B2B ?rms.
Firms performing B2C relations generally have more customers than B2B ?rms and are more visible to the
end-consumer than B2C ?rms. We expect that these aspects in?uence the amount of social media messages
related to a ?rm. The reason why the volume of social media messages related to di?erent ?rms is important
to investigate is the fact that social business intelligence will only be possible for a ?rm in case that there are
actually messages created that are related to the ?rm. Hence, the ?rst hypothesis is formulated as:
H
1
: The volume of ?rm-related social media messages is higher for B2C ?rms than for B2B ?rms.
Secondly, it is expected that the volume of ?rm-related messages di?ers among ?rms active in di?erent industries.
As illustrated in section 1-2, the nature of the products and/or services traded in di?erent industries a?ects
the rate at which products/services are sold. Therefore, we expect that ?rms in some industries are more often
subject of discussion on social media than ?rms in other industries. For instance, retail products are more
frequently bought by people than houses. Taken into account this rational reasoning, the second hypothesis is
formulated as:
H
2
: The volume of ?rm-related social media messages di?ers between industries.
Whereas the customer relation types have yet been operationalised by two groups (B2B or B2C), the industry
types have not yet been established. In section 4-2, the industry categories will be established based on a generic
classi?cation.
Hypotheses related to the Subjects of Social Media Messages
The third aspect that is investigated in the content analysis relates to the subjects of social media messages.
Whereas the ?rst two hypotheses provide insight in the existence of ?rm-related social media messages, this
insight is not su?cient to draw conclusions on the applicability of social business intelligence. The subjects
of social media messages are also to be included in the analysis. It is important to research the subjects of
social media messages related to ?rms since the subjects of the messages are used to assign the messages to
key-performance indicators. As such, the subjects contained in the messages determine – in combination with
the volume – the applicability of social business intelligence for ?rms.
The subjects of ?rm-related social media messages are investigated on the same two dimensions as the volume
of the messages. The contexts of B2C ?rms di?er from B2B contexts. It is therefore likely that the subjects
44 Content Analysis
of the messages related to B2C ?rms di?er from the subjects of B2B related messages. So far, we do not
have strong signals that certain subjects are more often discussed in one customer category than in the other.
In order to understand which type of ?rms can ?nd messages related to di?erent KPIs, the third hypothesis
investigates whether or not the subjects of ?rm-related social media messages di?er between ?rms performing
di?erent customer relation types. Hence, the third hypothesis is formulated as:
H
3
: The subjects of ?rm-related social media messages di?er between ?rms performing B2B and
B2C relations.
For similar reasons that are concerned with the second hypothesis, the subjects of messages related to ?rms
in di?erent industries are expected to vary. For example, it is likely that messages related to user experiences
are more frequent created in an industry that creates electronic consumer products as compared to an industry
that delivers consulting services. The subjects contained in social media messages – combined with the volume
of these messages – a?ect the opportunities for social business intelligence for ?rms. Since this thesis examines
the opportunities for social business intelligence for di?erent ?rms, it is necessary to investigate variations in
the subjects related to ?rms in di?erent industries. Consequently, the last hypothesis is formulated as:
H
4
: The subjects of ?rm-related social media messages di?er between industries.
The content analysis will be designed according to these hypotheses, and the results of the content analysis
allow us to con?rm or reject the four hypotheses. As such, we can draw conclusions on the applicability of social
business intelligence for ?rms in di?erent contexts.
4-1-2 Material to Investigate
The second step in the theoretical level consists of a description of the material to be investigated. In this
thesis, we investigate social media posts that are related to ?rms. As described in chapter 3, di?erent social
media platforms serve di?erent purposes, leading to di?erent type of posts. A 140 character tweet has a lower
information depth than e.g. a product review site. Since this research is exploratory in nature, it is valuable to
gain as much understanding as possible from the content posted on di?erent social media platforms. Therefore,
the material to investigate is sourced from popular social media sites in Western Europe. The content in the
dataset is sourced from Twitter, Facebook’s public pages, Flickr, Newssites, Google+ public pages, (Wordpress)
Blogs, Picasa, YouTube and Friendfeed.
4-2 Establishment of Categories
In the second step of the content analysis, the categories to be analysed are established and the sample set is
determined. Depending on the research, the categories will di?er. In this thesis, the categories that are to be
analysed consist of ?rms in di?erent industries and with di?erent customer relations. Next, the content analysis
of this thesis requires categories of social media posts to draw conclusions on the applicability of social media
posts for business intelligence.
4-2-1 Operationalising the Categories
We examine di?erences in the volume and subjects of social media messages related to ?rms on two nominal
dimensions; (i) industries and (ii) the customer relation type. The industry dimension is operationalised by
categorising ?rms in di?erent industries. The relation with end-users dimension is operationalised through
means of a distinction between either Business-To-Business ?rms or Business-To-Consumer ?rms.
(i) Industry Classi?cation
CBS (2012) – Statistics Netherlands – provides a hierarchical classi?cation of economic activities, called SBI
1
.
The European Union also has a classi?cation, called NACE
2
, on which SBI is based. SBI allows to classify
?rms based on their economic activities. SBI distinguishes multiple levels, of which the most aggregate level
distinguishes twenty main activities. These activities are listed in table 4-1, and are engaged to classify the
?rms that have been selected in the analysis of this thesis.
1
Standard Industry Classi?cation, in Dutch Standaard Bedrijfsindeling.
2
Statistical Classi?cation of Economic Activities in the European Community, in French Nomenclature statistique des Activités
économiques dans la Communauté Européenne.
4-2 Establishment of Categories 45
Table 4-1: General Classi?cation of Firms (CBS, 2012)
Industry
A Agriculture, forestry and ?shery
B Mining and quarrying
C Industry
D Production and distribution of and trade in electricity, gas, steam and air
E Extraction and distribution of water, sewerage, waste management and remediation
F Construction
G Wholesale and retail
H Transport and storage
I Accommodation, meals and drink provision
J Information and communication
K Financial institutions
L Real estate
M Consultancy, research and other specialised business services
O Public administration
P Education
Q Health and welfare
R Culture, sport and recreation
S Other services
T Households as employers
U Extraterritorial organisations and bodies
(ii) Customer Relation Type Classi?cation
As illustrated in chapter 1, the type of customer relations is likely to have an e?ect on the availability of social
media data related to a ?rm. Therefore, a category describing the type of customer relation is established.
Based on Turban et al.’s (1999) classi?cation of e-commerce, we classify ?rms in either Business-To-Business
(“B2B”) or Business-To-Consumer (“B2C”). This classi?cation will provide insight in the availability of social
media data related to ?rms based on the customer relation type.
Categories of Key-Performance Indicators
The key purpose of this thesis is to link social media posts to organisational performance. As described in
chapter 2, ?rms measure organisational performance based on key-performance indicators. In order to draw
generic conclusions of the applicability of social media data for the purpose of organisational performance,
a generic classi?cation of key-performance indicators is required. Section 2-2-5 illustrated that KPIs can be
classi?ed into ten categories. We will use these categories for the classi?cation of social media posts based on
the subjects of the messages. Table 4-2 shows the categories that are pursued in this research. In section 4-3-1
these categories are described.
4-2-2 Determining the Sample
One of the objectives of this research is to investigate possible di?erences in the user-generated content on
social media related to di?erent ?rms. In order to spot di?erences, our sample exists of ?rms that are active in
di?erent industries and take di?erent positions regarding end-users. This section describes the sample and the
industries that are part of the sample.
Selection of Firms
Based on the industry classi?cation presented in table 4-1, eighteen ?rms have been selected. The starting point
of the sample selection has been the list of ?rms that are part of the Amsterdam stock Exchange (“AEX”). The
main reason for this selection criterion is the fact that these ?rms are stock listed, and hence publicise annual
reports containing information about strategic initiatives, ?nancial ?gures, etc. In case the analysis shows inter
industry di?erences – e.g. between two comparable ?nancial institutions – the annual reports may provide
company speci?c information (e.g. amount of employees, attitude towards social media, etc.) clarifying these
46 Content Analysis
Table 4-2: Categories of Social Media Posts
Category Social media posts . . .
Short-term ?nancial results related to the ?rm’s ?nancial performance
Customer relations from individuals purposed to contact the ?rm, or from the ?rm purposed
to contact an individual
Employee relations related to employees of the ?rm
Operational performance related to the ?rm’s productivity, fact-based statements
Product and service quality related to the experience of products and services
Alliances related to joint-ventures or other cooperations
Supplier relations related to the suppliers of the ?rm
Environmental performance related to environmental / sustainability compliance
Product and service innovation related to innovation
Community revealing the community’s perception of the ?rm (not purposed to
contact the ?rm), chatter
Unde?ned that could not be de?ned in one of the categories
Spam that are not related to the ?rm
di?erences. If a sample containing privately owned companies would have been selected, access to additional
information would have been limited. In addition, ?rms that are listed in the AEX are generally well-established,
visible to the public and regularly subject to news articles. It is therefore expected that these ?rms are subject
of discussion on social media. Furthermore, the ?rm that sponsors this research – KPMG – requested to apply
the analysis on this list of corporations. Table 4-3 lists the selected ?rms, their corresponding industries and
main customer relation. For a description of the individual ?rms and their activities, see appendix B.
In order to design a uniform sample, ?rms of over-represented industries have been replaced by non AEX ?rms
which are also well-established corporations. The distribution of ?rms in the di?erent industries is shown in the
third column of table 4-3. The ?rms have also been classi?ed based on their type of customer relation. Though
the split between B2B and B2C is hard to make for some ?rms because they have B2C as well as B2B relations,
the motivation for the classi?cation of the ?rms is based on its main activities. The main activities, on which
the industry classi?cation as well as the customer relation classi?cation is based, of each ?rm are described in
appendix B. The ?nal column of table 4-3 shows the distribution of B2B versus B2C ?rms in the sample.
Table 4-3: Sample
Firm Industry CBS Relation
1 Akzo Nobel Mining and quarrying B B2B
2 ArcelorMittal Mining and quarrying B B2B
3 Unibail-Rodamco Financial institutions K B2B
4 Arcadis Consultancy, research and other specialised business services M B2B
5 Fugro Consultancy, research and other specialised business services M B2B
6 Coca-Cola Industry C B2C
7 Heineken Industry C B2C
8 Philips Industry C B2C
9 Albert Heijn Wholesale and retail G B2C
10 Blokker Wholesale and retail G B2C
11 C-1000 Wholesale and retail G B2C
12 KLM Transport and storage H B2C
13 NS Transport and storage H B2C
14 PostNL Transport and storage H B2C
15 Bol.com Information and communication J B2C
16 TomTom Information and communication J B2C
17 ABN AMRO Financial institutions K B2C
18 Aegon Financial institutions K B2C
4-2 Establishment of Categories 47
Description of the Industries
The sample consists of ?rms active in seven di?erent industries. As indicated, the industry classi?cation is
based on CBS’ (2012) Standard Industry Classi?cation. A description of the industries that are part of the
sample is presented in this section.
1. Mining and Quarrying
The activities of ?rms in the mining and quarrying industry are concerned with the extraction of oil, gas
and/or minerals such as sand, gravel and clay.
2. Industry
Industry ?rms are producers of food, beverages, tobacco, textile, chemical products, pharmaceutical raw
materials, metal, electric products, machines, cars, other transport modes, furniture and other products.
3. Wholesale and Retail
The industry wholesale and retail consists of ?rms trading in cars, food, machinery, agricultural products,
textile, books, and other consumer products. In addition, ?rms in this industry operate shops in which
consumers can buy their products.
4. Transport and Storage
The activities of ?rms that are active in the transport and storage industry transport persons or products
across land, water, air or other transport modes. Next, ?rms in this industry store products. Also, mail
related activities belong to the transport and storage industry.
5. Information and Communication
Firms in the information and communication industry are publishers of books, papers, magazines, software
and computer games. In addition, the production and distribution of ?lms, music and television shows
are assigned to the information and communication industry. Also telecommunication activities, whether
through wires, wireless, satellite or other mediums are assigned to the information and communication
industry. All software related activities required for telecommunications are part of the information and
communication industry.
6. Financial Institutions
The industry ?nancial institutions consists of banks, ?nancial holdings, investment institutions, insurance
companies, pension companies, asset management companies and other ?nancial ?rms.
7. Consultancy, Research and Other Specialised Business Services
The activities of ?rms in the consultancy, research and other specialised business services relate to advisory
services on di?erent domains. Examples of ?rm types in this industry are law ?rms, accountancy ?rms,
engineering ?rms, architects, marketing ?rms, research ?rms, etcetera.
4-2-3 Description of the Measuring Period
uberVU, one of the emerging social media monitoring and analysis tools has granted access to their tool for a
period of 14 consecutive days, i.e. from Friday 20 July to Thursday 2 August. This tool is further described in
section 4-4-2. During the measuring period the eighteen ?rms have been monitored, resulting in the collection
of 224.687 social media posts related to di?erent ?rms. This amount of messages is deemed su?cient to analyse
what the subjects of social media messages related to ?rms are, and how these subjects di?er in volume from
each other, which is the purpose of this chapter. During 7 of the 14 days in the period, the 2012 Olympic Games
took place. As a consequence, the social media posts of ?rms that are for some reason – e.g. as a sponsor –
involved in the Olympic Games often have the Olympic Games as a subject of the post. It is common for ?rms
to sponsor events. In case the social media messages would have been collected during another period, it is likely
that some ?rms were sponsoring an event as well during the measurement period. Furthermore, the summer
holidays took place during the measurement period. Though people undertake other activities during their
holidays, and hence may show di?erent activities on social media as well, it is not likely that large corporations
– i.e. the ones in our sample – are not subject of discussion during this period. On the contrary, some ?rms will
precisely be mentioned during this period. However, when interpreting the results after the analyses, one should
be aware of the fact that the social media messages at which the conclusions are based have been created during
a holiday period. Furthermore, conclusions related to the procedure of social business intelligence will not be
a?ected by the fact that the data are created during a holiday period, since the way of collecting, processing
and analysing the data will be the same in any period.
48 Content Analysis
4-3 Pretest
In step 3 of the content analysis – see ?gure 4-1 –, the data collection is tested to ensure that the established
categories can be ?lled with data from the selected sample. To validate the categories that were established to
classify social media posts, we pretest the categories that are presented in table 4-2 by classifying the ?rst 100
messages of each respondent. Each of the pretest social media posts have been read and consequently assigned
to one of the categories. The sample illustrated that some of the categories were too generic. Therefore, we
added sub categories to some of main categories to gain more insight in the nature of the social media posts.
The categories of social media posts are discussed in the following section.
4-3-1 Categories of Social Media Posts
The purpose of this thesis is to assign social media posts to key-performance indicators. To achieve this, we
apply the classi?cation scheme of Ittner et al. (2003) to distinguish key-performance indicators from each other.
This classi?cation scheme distinguishes ten key-performance indicator categories. Basically, we adhere to these
ten categories. However, as we have experienced in the pretest, social media posts within one category are too
heterogeneous to simply assign the social media post to the high-level classi?cation that distinguishes between
ten categories. Therefore, an additional level of detail has been assigned to some of the main categories. This
additional level of detail has been established based on the pretest of the categories. Thus, the empirical data
has driven the establishment of these categories. The naming of these – more detailed – sub categories represent
the nature of the social media posts as well as possible. In the following section we discuss the categories of the
social media posts, which are based on the the KPI classi?cation of Ittner et al. (2003).
1. Short-term ?nancial results
The ?rst KPI category consists of indicators representing the (short-term) ?nancial performance of the ?rm.
Financial indicators are typically measured by ?rms using internal systems. In other words, there is no external
in?uence required to measure these metrics. Though the added value of the information of social media posts
related to ?nancial results may be of little value for the ?rm (management has ?nancial results earlier available
than the ?rm’s environment), it is wise to classify these social media posts nevertheless to provide an as
complete overview as possible of the type of social media posts that are available for ?rms. Typical examples of
?nancial indicators are the number of sales in a certain period, amount of debt on a certain moment, operating
expenditures in a certain period, etc. Social media posts that can be classi?ed into this category relate to the
?rm’s ?nancial performance. As we have experienced in the pretest, social media posts related to ?nancial
aspects of a ?rm are either related to discussions of the performance of a ?rm, or related to the ?rm’s share
prices. Accordingly, these two categories have been added as sub categories. These two sub categories are
discussed in the following sections.
1.1 Financial performance discussions
Social media are used to discuss the ?nancial performance of a ?rm. Often, posts related to ?nancial performance
contain a factual statement of the ?rm’s performance, which is sometimes followed by an opinion of the creator
of the post. In addition, these type of social media posts often contain a hyperlink to a website at which the
?nancial performance is further analysed. The following two example posts that consist in our sample illustrate
the type of posts that are classi?ed as ?nancial performance discussions posts:
“Akzo Nobel Q2 Pro?t Takes 21.5% Hit on Restructuring Charges http://t.co/b1518IDQ”.
“STEEL RESULTS: #ArcelorMittal Flat Carbon #Europe reports Q2 earnings fall http://t.co/0nPq7VjM
#steel”.
The existence of ?nancial performance related discussions in our sample data is due to the selection of the
?rms in our sample. As discussed in section 4-2-2, the sample ?rms are based on stock listed companies.
These companies are public limited liability ?rms, hence required to publicise their ?nancial performance on a
regular basis by law. Since the ?nancial performance is publicly available, it makes that these ?rms’ ?nancial
performance are subject of discussion on social media. In case that our sample would have consisted of limited
companies, which are not required to publicise their ?nancial position, we would probably not have found social
media posts related to ?nancial performance in the dataset.
4-3 Pretest 49
1.2 Shares related discussions
Another substantial part of the social media posts related to the ?nancial aspects of a ?rm has the ?rm’s shares
(prices) as subject. Often, these posts are made by analysts specialised in stock markets. The following two
social media posts illustrate discussions related to the shares of a ?rm:
“TomTom: Ster van de week: TomTom maakte vorige week bekend navigatieproducten en -diensten te gaan
leveren... http://t.co/8QRPxJmj #beleggen”.
“Stijgers 1: Wereldhave (4,05%) ; 2: TomTom (2,43%) ; 3: Bal last Nedam (2,25%) ; 4: VastNed
Retail (1,94%) ; 5: Heineken (1,76%)”.
The character of the second example post is typically found in the dataset, it shows the top ?ve funds of that day.
These type of posts are generated automatically by computers, often referred to as ‘bots’. For similar reasons
with the ?nancial performance discussions posts, we expect that shares related discussions are particularly found
in our dataset because the ?rms in our sample are stock listed.
2. Customer relations
The second KPI category de?ned by Ittner et al. (2003) consists of performance indicators related to customer
relations. As indicated in section 1-2, many researches acknowledge the opportunities for customer relationship
management using social media. Not surprisingly, social media posts in our test sample could be assigned under
the umbrella of customer relations. In our opinion, this category required a higher level of detail to distinguish
the social media posts made by the ?rm’s own web team from the posts that were directed towards the ?rm’s
web team. Seven categories have been added under the customer relations category to distinguish the nature
of the social media posts made by or directed to the web care teams. From a BI perspective it is valuable to
gain insight in the type of social media posts that are made by web care teams, since these posts may in?uence
other factors like customer satisfaction, sales, or the costs related to customer relations.
The social media posts related to customer relations typically show a conversation between a ?rm and a customer.
Social media posts related to customer relations represent a direction that is either from the customer to the
?rm, or from the ?rm to the customer. Social media posts from customer to customer containing a ?rm’s name
are categorised in a di?erent category, which will be discussed later. The identi?cation of social media posts
from customer to ?rm, or the other way around, is relatively easy because people begin their message with the
name of the receiver, preceded with an “@”. As such, social media posts starting with e.g. @ABN AMRO have
been classi?ed as a customer to ?rm post. Social media posts that were made by a webcare team could also be
easily recognised, because the creator of the message generally contains the name of the ?rm. To illustrate the
direction of the customer relations posts, the naming of the categories represent the direction of the post.
2.1 Customer questioning the Firm
Social media are used by customers to ask questions. The nature of the questions di?er, some people ask speci?c
questions about a product or service, how to use it or how it’s made while other questions are very broad and
relate to the company’s strategy or position in the market. Social media posts that were made by customers,
directed towards a ?rm and illustrating a question have been classi?ed under the category questioning customer.
Figure 4-2 illustrates the direction of these posts; from customer to ?rm.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-2: Questioning customer
Two example posts of the category questioning customer that have been found are illustratively shown below.
“@albertheijn Wat zijn de ramadanproducten? Ken je me die ? tweeten? alvast bedankt.”
“@KLM Hi, booked ?ights with you via @lastminute.com, wondering how we check in online? Saying that
option isn’t available?”
50 Content Analysis
Obviously, customer relationship teams that are active on the Web monitor social media posts in which customer
ask questions that are related to the ?rm, and consequently respond to these questions. Often, a questioning
customer post is followed by an explaining ?rm post.
2.2 Firm explaining the Customer
Clearly, one of the purposes of a web care team is to help customers with problems they experience. Many social
media posts that are made by web care teams contain an explanation of problems or questions that customers
posted on social media platforms. These posts have been classi?ed as explaining ?rm, illustrating that the social
media post has been written by a ?rm’s web care team to explain a certain customer something in response to
an earlier post made by the customer. Figure 4-3 illustrates the direction of the social media posts that have
been categorised as explaining ?rm.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-3: Explaining ?rm
Two example posts of the category explaining ?rm are shown below:
“@kiimberley94 Dag Kim. Als een bedrag dmv een automische incasso is afgeschreven, wordt het bedrag met
max 2 werkdagen teruggeboekt. Elvira.”
“@mepe176 Bij geldautomaten met een Maestro-logo is dit zeker mogelijk. Sommige winkels bieden
ook deze mogelijkheid. Suzanne.”
2.3 Customer complaining to the Firm
Customers employ social media as a means to complain. Plenty examples that have reached the newspapers
in recent years exist. It is therefore not surprising that our data set shows social media posts that represent
a complaint. Customers complain about product experiences, how they have been treated in their complaints
procedure, etc. Social media posts that have been made by customers and illustrating a complaint have been
classi?ed into the category complaining customer. Figure 4-4 illustrates the direction of these social media
posts; from customer to ?rm.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-4: Complaining customer
Two example posts illustrating complaining customer posts are shown below:
“@ABNAMRO Blijkbaar is de enige manier om met jul lie een probleem op te lossen om NIET TE
BETALEN. Want telefonisch sta je in de kou! HELP”
“@PostNL @PostNLWebcare de zoveelste keer dus dat de bezorgers de aangetekende stukken niet
laten tekenen ..........”
It is the purpose of a ?rm’s web care team to respond to the complaining posts. Therefore, a complaining
customer post is often followed by an understanding ?rm or complaining ?rm post.
2.4 Firm showing feeling of understanding to the Customer
Next, as the sample data shows, a ?rm’s web care team is also purposed to show a customer a feeling of
understanding of his or her experienced problem. The social media posts that were made by a ?rm’s web
4-3 Pretest 51
care team and represent a feeling of understanding with the customer’s complaint have been classi?ed into the
category understanding ?rm. Figure 4-5 schematically shows the direction of these posts. Understanding ?rm
posts di?er from explaining posts since understanding ?rm posts do not o?er the customer a solution to the
experienced problem, but rather show a feeling of understanding.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-5: Understanding ?rm
Two example posts that were found in the test sample and clearly illustrate a sense of understanding of the
customer’s experienced problems are illustrated below.
“@normanwil lems Dag Norman, vervelend te horen dat je reis niet doorgaat. Als je belt met 0900-0024
kunnen we je verder helpen. Margot.”
“@noni1967 Dag Nanette, dat is erg vervelend om te horen. Ik hoop dat het snel verwerkt wordt.
Margot.”
2.5 Customer thanking the Firm
As discussed, web care teams are amongst others purposed to help customers with problems that they experience,
thereby replacing the traditional telephone help desks. The social media posts in our sample clearly show a
conversation, where a ?rm replies to a social media post made by a customer. Once a customer has been assisted
by a company’s web care team, some customers take the e?ort to thank a ?rm for their assistance. Social media
posts that have been made by customers that are directed towards a ?rm and illustrating gratitude towards
the ?rm, have been classi?ed into the category named thanking customer. Figure 4-6 illustrates the direction
of thanking customer posts; from customer to ?rm.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-6: Thanking customer
Two example posts illustrating thanking customer posts are shown below:
“@ABNAMRO Ok, thanks voor de snelle reactie en ?jne dag!.”
“@albertheijn Oke bedankt! Dan ga ik Valkeniersplein proberen.”
2.6 Firm thanking the Customer
The sixth sub category that was added to the customer relations umbrella has been assigned the name thanking
?rm. As we experienced, web care teams often thank the customer for mentioning a de?ciency of a product /
service, or the web care teams thanks a customer for a compliment made on the side of the customer. Figure 4-7
illustrates the direction of these type of posts; from ?rm to customer.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-7: Thanking ?rm
Below, two example posts that were found in the sample and clearly present that the web care team is grateful
towards the customer’s earlier post are shown.
52 Content Analysis
“@Birdy_Fly Bedankt voor de tip Tom, ik zal deze door gaan zetten naar de betre?ende afdeling. Fijn
weekend. Martijn.”
“@LindaWestenberg Bedankt Linda! Ik wens je een ?jne dag toe. Elvira.”
2.7 Firm informing many Customers
Finally, ?rms also use social media to inform their customers on certain topics. Social media posts made by a
?rm and purposed to inform customers on a certain topic are classi?ed into the category named informing ?rm.
Whereas social media posts of the category explaining ?rm also inform customers, informing ?rm posts di?er
because they are directed to anyone. Figure 4-8 illustrates the “one-to-many” relation of the informing ?rm
posts. Thus, informing ?rm posts are not speci?cally directed towards an individual and hence – as opposed
to explaining ?rm posts – do not start with an “@”.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-8: Informing ?rm
Two example social media posts found in the sample and classi?ed as informing ?rm are shown below:
“#NS Deventer-Zutphen (overwegstoring) Tussen Zutphen-Deventer geen treinen door overwegstoring.Extra
reistijd 30/60 min.Tot +/- 14:30... .”
“#NS Zwolle-Amersfoort: defecte trein: Tussen Amersfoort en Zwol le minder treinen.. Extra reistijd 15 -
30 min. ( Tot +/- 22:00 ).”
3. Employee relations
The third category of key-performance indicators comprises indicators related to employee relations. In this
thesis we seek for social media posts that relate to KPIs. Social media posts that are related to employee
relations will be classi?ed under this umbrella. However, as the pretest illustrated, there exists a variety in the
social media posts that could be assigned to the employee relations category. Therefore, two sub categories have
been established; recruitment and employee posts. These ?ndings are in line with McCorkindale (2010).
3.1 Recruitment
The ?rst sub category related to employee relations contains social media messages that involve employee
recruitment processes. Our sample shows messages where people write about vacancies in a ?rm, students
asking companies for an internship place, human resource managers wishing new employees a good start at
their ?rst day of work in the company, etcetera. Social media messages that are related to a company’s
recruitment process have been classi?ed into the recruitment category. Illustratively, two example posts of the
category recruitment are shown below.
“#nieuwe #vacature: Medewerker Verkoopklaar / Rotterdam / Albert Heijn #VCW #banen
http://t.co/bEDnvCFw.”
“Job opportunity: Deputy Program Manager - Trenchless Tech at ARCADIS - Washington D.C.
Metro Area #jobs http://t.co/RRSw5RtV.”
3.2 Employee posts
The second sub category related to employee relations contains social media messages made by the ?rm’s
employees. As our dataset illustrates, employees use social media to share their work experiences or indicate
that they are working at the ?rm. Social media messages that have been made by employees have been classi?ed
under the category employee posts. Illustratively, two posts existing in our data set that have been classi?ed as
employee posts are shown below.
4-3 Pretest 53
“O?cial ly 10 years working @ ABN Amro Bank... Uno?cially 13 years.. .”
“@PostNL Overuren niet betaald, geen bevestiging van gevraagde vakantie en ?etsdeclaratie wordt
niet uitbetaald. Lekker motiverend! #postnl.”
4. Operational performance
The fourth main category of key-performance indicators relates to operational performance. Social media
messages that relate to the operational performance of a ?rm are classi?ed into this category. As discussed,
category 2.3 contains social media messages in which users complain about the ?rm’s product / or service. It
is possible that customers complain about the ?rm’s operational performance, for instance about the delivery
time of a product. However, messages that have been classi?ed into the operational performance category
re?ect facts, while customer-to-?rm posts classi?ed as complaining customer (category 2.3) are more subjective
in nature and directed towards a ?rm. Below, two posts in our sample that have been classi?ed as operational
performance posts are shown.
“RT @ Webwereld: Derde keer in korte tijd storing ABN Amro http://t.co/v8Y953kC.”
“Vrijdag 27-7 kaart verstuurd uit Assen, 27-7 gestempeld in Zwol le. 31-7 al aangekomen in Tynaarlo.
Bravo @postnl. Niet gek voor 15 km.”
5. Product and service quality
The ?fth KPI category that ?rms apply contains indicators related to product and service quality. As our
sample illustrates, customers share their product and/or service experiences through social media. Social media
messages that represent product and service experiences of customers have been classi?ed in the category product
and service quality. Two posts existing in our sample and classi?ed as product and service quality posts are
shown below.
“Ik haat die Heineken met draaidop, altijd snij ik m’n hand er mee open.”
“Ik had een albert heijn tas toen ik thuis kwam waren myn handen helemaal blauw.”
6. Alliances
The sixth main category of key-performance indicators relates to the ?rm’s alliances. Social media messages
that are related to the ?rm’s partnerships / alliances have been classi?ed into the alliances category. Two posts
that have been classi?ed as alliances posts are shown below.
“In what has to be one of the strangest collaborations ever, military scientists from the UK’s Defence
Science and Technology Laboratory (DSTL) have been working with global paint and coating company
AkzoNobel to develop an anti-chemical weapon paint that can absorb harmful chemicals from enemy...
http://inhabitat.com/uk-military-develops-paint-that-absorbs-fal lout-from-chemical-attacks.”
“Op weg naar #Atrium MC om te spreken met nieuw bestuur #vereniging #artsassistenten en samenwerking
met #ABNAMRO.”
7. Supplier relations
The seventh main category of KPIs involves indicators related to a ?rm’s supplier relationships. Since this
thesis seeks to link social media data to KPIs, social media messages related to this category of KPIs have been
classi?ed into the supplier relations category. Apparently, suppliers of a company post social media messages
indicating that they supply the ?rm with products / services. Illustratively, two social media messages existing
in our sample which have been classi?ed as supplier relations posts are shown below.
“Onze koks ?etsen door het Hol land Heineken House, al les loop op rol letjes! #hhh2012
http://t.co/DrbbbXwl”
“Bezig met een nieuwe klus, het #vormgeven van een #advertentie deze keer voor de #albertheijn
#AH Valkeniersplein te #Breda.”
54 Content Analysis
8. Environmental performance
The next main category of key-performance indicators involves indicators re?ecting the ?rm’s environmental
performance. Social media messages relating to the environmental performance of ?rms have been classi?ed
into this category. The dataset shows social media messages in which consumers discuss the environmental
responsibility of the company. Below, two posts existing in our dataset that have been classi?ed as environmental
performance posts are shown.
“We kunnen heel veel bijdragen aan ontwikkelingen op het gebied van duurzaamheid... aldus Albert Heijn.
Ze verkopen uien uit Australië #AH.”
“@GreenpeaceNL - olijfolie v @albertheijn zit tegenw. in plastic “samen meer doen voor het milieu”
-> is t idd beter? http://t.co/2vLBda61.”
9. Product and service innovation
The ninth main category of key-performance indicators involves indicators related to product and service
innovation. As illustrated in section 3-2-3, social media are used by consumers to share product experiences
and to suggest innovations. The process of co-creation and prosuming is shown in our dataset as well. Social
media messages that re?ect people’s attitude to new products or services or contain suggestions for innovations
have been classi?ed as product and service innovations, of which two posts are illustratively shown below.
“Best Reviews - Philips Sonicare HX6732/02 HealthyWhite R732 Rechargeable Electric Toothbrush -
http://maxtodaystore.info/today-p... .”
“Moe worden van #ABN-AMRO bank geld overmaken steeds weer die achterlijke reader nodig , neem een
voorbeeld aan tan- codes van #ING!!”
10. Community
The tenth and ?nal main category of key-performance indicators related to indicators related to the ?rm’s
community. Social media posts belonging to the category community reveal how the community, that is,
external actors, perceive the ?rm. Many social media posts in the pretest could be assigned under the community
category. However, to provide more detail in the type of social media posts related to the community category,
we established ?ve sub categories to the community category. These sub categories are discussed in the following
sections.
10.1 Promotion
Some social media posts in our sample were created by ?rms themselves, and are hence not perceived as
user-generated content from a ?rm point of view. Social media messages that were made by ?rms themselves
and were purposed to promote the ?rm to the environment, were assigned to the sub category named promotion.
The following two posts illustrate promotion activities of a ?rm:
“What do you want to do more of in retirement? Travel, spend time with family, pursue hobbies, or more
education? Check out the results from the quick pol l here!”
“Albert Heijn - Kom op 18,19,25 en 26 aug. naar de Open Dagen van onze boeren en telers.
http://t.co/Saiv2aWY http://t.co/bmP1CePw.”
10.2 News
The test sample illustrated that many social media posts are (simply) noti?cations of news articles. We
positioned these posts under the category community since news messages determine the ?rm’s exposure to
the ?rm’s community. Posts belonging to the sub category news are written by professionals, which are mostly
journalists promoting their news article. Below, two examples of news categorised social media posts are shown.
“En verder in de serie vakantiebaantjes vandaag: Gerrit Zalm, waarmee verdiende de baas van #abnamro
zijn eerste centjes? #BNR.”
“@huizenprijzen: Han de Jong (Chief Economist ABN AMRO) : "Er is weinig in dit leven zo gevaarlijk als
schuld" : http://t.co/VOTreLdg via @youtube #schuld.”
4-3 Pretest 55
10.3 Public image
Thirdly, the sample test illustrated that individuals share their attitude towards a ?rm through means of social
media. We classify these posts into a sub category named public image. Social media posts classi?ed into the
public image category are not directed towards a ?rm, or, not purposed to get in contact with the ?rm. Rather,
social media messages assigned in the public image category represent discussions and “chatter” amongst the
social media users, in which the topic of discussion is the ?rm or its products / services. Public image posts are
written by non-professionals, while – as we will see in the next sub category – posts created by professionals
are assigned into a separate sub category. Below, two example posts that were found in the sample set and
represent the public’s image towards the ?rm are shown.
“Even kijken of blokker kruimeltje de ?lm heeft liggen want ik heb m al leen nog maar op video band!”
“Je moet staatsbanken ABNAMRO en ING 15% betalen als je rood staat op je betaalrekening en je
krijgt 2% als je + staat #Schurkenbanken.”
10.4 Professionals
Fourth, our pretest indicated that there are social media messages created by professionals. Messages that have
been created on social media by external professionals – not from the company – and talking about the ?rm
have been classi?ed into the category labelled professionals. Two example posts of this category are:
“Presentatie @ jaccooudhof van #KPMGmkb op de "Kengetal lenbijeenkomst" van @ Ful lFinance @
ABNAMRO en NOVAK”
“Sarah Harding interviews Arcadis at the A&WMA conference in San Antonio.”
10.5 Distributors
Finally, the data showed that social media are also used to promote products. However, the ?rm that produces
does not have to be necessarily the one that promotes the product. Our dataset contains social media posts
made by distributors of the product. These posts have been classi?ed into the category labelled distributors.
Two example posts of distributors are shown below.
“Macco Akzo Nobel Pai DWP24 Liquid Nails Drywal l Construction Adhesive: Special ly formulated latex
product for in... http://t.co/Vk3PdXLR”
“Best O?er - Philips Norelco AT830 PowerTouch Rechargeable Cordless Razor, Gray/Silver/Black -
http://maxtodaystore.info/weekly-...”
Unde?ned
Based on the 26 social media post categories (including main categories) that have been established in the
previous sections, we can not classify all social media posts. As discussed before, social media data is
unstructured and the interpretation of a social media post is not always easy. Therefore, messages that –
despite of the mentioning of the ?rm’s name in the post – could not be assigned to one of the categories have
been assigned into the unde?ned category. Two examples of posts that were unde?ned are:
“Volg ons (Unicum) @ ABN AMRO bij zuidplein om 13 : 00 !!!!! RT RT.”
“Nu naar ?etsenwinkel, blokker en c1000 met pap en mam!”
Spam
Unfortunately, the search queries that were used to scrape the social media content did still result in the
collection of data that is totally unrelated to the ?rm. This is due to the fact that people’s names or IDs are
similar to the ?rm’s name. Social media posts that were totally unrelated to the ?rm have been classi?ed as
spam, of which two examples are shown below:
“@Klm_babe Okay wel l maybe sometime next week then”
“@Jack_Heineken meen je dat nou? -_-
een korte broek aan naar de zaak ”
56 Content Analysis
4-3-2 Revised Taxonomy of Categories
The categories of the social media posts are based on the KPI classi?cation scheme of Ittner et al. (2003), which
allows classi?cation of performance metrics into one of the ten categories. Our addition of sub categories does
not a?ect the structure of Ittner et al.’s (2003) classi?cation scheme, but rather adds a layer of detail to the
categories. An overview of the revised taxonomy – after addition of the sub categories – and a short description
of the social media posts of the corresponding categories is presented in table 4-4. Figure 4-9 schematically
shows the taxonomy of the key-performance categories and the social media post categories.
Table 4-4: Taxonomy of Categories of Social Media Posts
KPI Category Social media posts . . .
1. Short-term ?nancial results related to the ?rm’s ?nancial performance
1.1 Financial performance discussions related to the ?rm’s ?nancial performance
1.2 Stock related discussions made by professionals/individuals analysing the ?rm’s stock price
2. Customer relations from individuals purposed to contact the ?rm
2.1 Questioning customer posts from a customer asking a question to the ?rm
2.2 Explaining ?rm from the ?rm purposed to explain the customer something
2.3 Complaining customer from a customer complaining about the ?rm / ?rm’s products or
services
2.4 Understanding ?rm from the ?rm purposed to show the customer shared
understanding
2.5 Thanking customer from individuals purposed to thank the ?rm
2.6 Thanking ?rm from the ?rm purposed to thank the customer for an earlier post
2.7 Informing ?rm from the ?rm purposed to inform customers (not responding to
an individual)
3. Employee relations related to employees of the ?rm
3.1 Recruitment related to recruitment of new employees
3.2 Employee posts made the ?rm’s employees
4. Operational performance related to the ?rm’s productivity
5. Product and service quality related to the experience of products and services
6. Alliances related to joint-ventures or other cooperations
7. Supplier relations related to the suppliers of the ?rm
8. Environmental performance related to environmental compliance
9. Product and service
innovation
related to innovation
10. Community revealing the community’s perception of the ?rm (not
purposed to contact the ?rm)
10.1 Promotion made by the ?rm for promotion activities
10.2 News made by external professionals (journalism)
10.3 Public image made by non-professionals, individuals (‘chatter’)
10.4 Professionals made by professionals talking about the ?rm
10.5 Distributors made by distributors of the ?rm’s product/service
Unde?ned that could not be de?ned in one of the categories
Spam that are not related to the ?rm
4-4 Data Collection and Evaluation
The fourth step of the content analysis comprises the data collection and analysis. The purpose of this section is
to collect social media posts related to the ?rms of the sample, and to analyse these data to identify di?erences
in the content related to di?erent ?rms. Moreover, the experiences that we encounter in the data collection
4-4 Data Collection and Evaluation 57
1. Short-term financial results
4. Operational performance
5. Product and service quality
6. Aliances
7. Supplier relations
8. Environmental performance
9. Product and service innovation
10. Community
10.2 News
10.1 Promotion
1.1 Financial performance discussions
10.3 Public image
2. Customer relations
Customer to Firm
Firm to Customer
2.3 Complaining customer
2.1 Questioning customer
2.5 Thanking customer
2.6 Thanking firm
2.4 Understanding firm
2.7 Informing firm
Main KPI Categories Social Media Post Categories
3. Employee relations
3.1 Recruitment
3.2 Employee posts
10.4 Professionals
10.5 Distributors
2.2 Explaining firm
1.2 Stock related discussions
Figure 4-9: Taxonomy of Social Media Post Categories
The ?gure illustrates the ten main categories of key-performance indicators that have been found in the literature.
Additionally, sub categories have been established at which the social media messages could be assigned. These
additional categories have been constructed based on the empirical data.
phase as well in the data analysis phase will serve as a baseline in formulating requirements for a social business
intelligence procedure that we develop in a later stage of this thesis.
Watson and Wixom (2007) illustrate that data collection requires about 80% of the “time and e?ort” related to
business intelligence, and that data collection is responsible for “50% of the unexpected costs” in BI projects.
Social media platforms are new sources for ?rms to collect data. The experiences from the collection of social
media data for this research contain valuable lessons learned for ?rms willing to utilise social media data for
58 Content Analysis
business intelligence. Therefore, we pay much attention to describing the steps that are necessary to collect
social media data.
In section 4-4-1 we discuss the search queries that are used to ?lter out those social media posts that relate to
the ?rms in our sample. Section 4-4-2 describes how the social media posts have been extracted from the web,
and how these posts have been placed in a database allowing to be analysed. Although we will use proper search
terms, it is expected that many social media posts contain unrelated information. Therefore, the data will be
cleaned in section 4-4-3. Once the data is cleaned, section 4-5 provides descriptive statistics about the amount
of social media posts available for ?rms in di?erent industries and for di?erent positions regarding end-users.
4-4-1 Search Terms
As on the regular web, search terms are used on social media to ?lter out information that is the subject of
interest. In social media, and especially on Twitter, users place a # (‘hashtag’) before a word to indicate the
subject of the particular social media post. Hashtags can be considered as meta data tags indicating the subject
of the social media post. When each user uses the same hashtags about a certain topic, it becomes easy to track
the stream of social media posts related to that subject. We will use the strength of hashtags to ?lter out the
social media posts that are related to the ?rms in our sample. Another widely used symbol to indicate that a
social media post is direct to a person, or a ?rm, is the @ (‘at symbol’). As with the hashtag, this symbol is
positioned before one’s (nick)name to illustrate that a post is direct towards this person or organisation. We
will use the at symbol in our search terms, because social media posts containing this symbol are directed to a
receiver, the ?rm.
Additionally, because not everyone and not each social media platform adheres accurately to the usage of
hashtags, we add a search term containing the name of the ?rm without a hashtag to the search terms. Next,
because some ?rms have name that can be written in multiple forms, we also search on di?erent names. An
overview of the search terms used to ?lter out the social media posts that are related to the ?rms in our sample
is presented in table 4-5.
Table 4-5: Firms and Corresponding Search Terms
Firm Search Terms
1 ABN AMRO #abnamro, #abn amro, @abnamro, @abn amro, abnamro, abn amro
2 Aegon #aegon, @aegon, aegon
3 Akzo Nobel #akzonobel, #akzo nobel, @akzonobel, @akzo nobel, akzo nobel, akzonobel
4 Albert Heijn #albertheijn, #albert heijn, @albertheijn, @albert heijn, albertheijn, albert heijn
5 Arcadis #arcadis, @arcadis, arcadis
6 ArcelorMittal #arcelor mittal, #arcelormittal, @arcelor mittal, @arcelormittal, arcelor mittal,
arcelormittal
7 Blokker #blokker, @blokker, blokker
8 Bol.com #bol.com, @bol.com, bol.com
9 C-1000 #c1000, @c1000, c1000
10 Coca-Cola #coca-cola, #cocacola, @coca-cola, @cocacola, coca-cola, cocacola
11 Fugro #fugro, @fugro, fugro
12 Heineken #heineken, @heineken, heineken
13 KLM #klm, @klm, klm
14 NS #ns, @ns, ns
15 Philips #philips, @philips, philips
16 PostNL #postNL, @postNL, postNL
17 TomTom #tomtom, @tomtom, tomtom
18 Unibail-Rodamco #unibail-rodamco, @unibail-rodamco, unibail-rodamco
4-4-2 Scraping Social Media Content
Web scraping – also known as web crawling – is the excavation of data from web pages into a local structured
database, so that these data can be analysed (Huang, Li, Li, & Yan, 2012). Figure 4-10 visualises this process.
The web scraper is provided with keywords, so that it can detect those particular web pages or social media
posts related to the topic of interest. All content that ful?ls the keywords are consequently stored into whatever
4-4 Data Collection and Evaluation 59
form the person prefers, which is often a database, a web page, another application or – as in this thesis – a
spreadsheet. Scraping web pages allows one to extract that particular information from the web that one is the
topic of interest, and consequently process the data for its own purpose.
Web
Spreadsheet
Browser
Database
Application
Scraper
Figure 4-10: Web Scraper
Content from the web (e.g. a social network site) is ?ltered, extracted and stored to into di?erent types of media,
e.g. in a database, or a spreadsheet.
Software tools purposed to monitor social media are emerging rapidly. uberVU is one of such tools that
are available on the market. In addition to social media monitoring uberVU allows the extraction of social
media posts into comma-separated value (“CSV”) format, and is therefore regarded as a social media platform
scraper. It is this aspect of the tool that was decisive for the selection of a social media extraction tool that we
used to scrape the content. uberVU was established in 2008, their software is amongst others used by NBC,
Microsoft, Audi, Nestle, T-Mobile, Thomas Cook, 3M, PayPal, BASF and The World Bank. The software
indexes multiple social network platforms, including Facebook, Twitter, YouTube, Flickr, Vimeo, Picasa. In
addition, traditional media like news sites and (Wordpress) blogs are monitored. Consequently, the software
presents metrics including number of mentions over the last period, number of likes, number of shares, platform
distribution, sentiment of the online posts, gender distribution, language of the posts and the countries where
the posts originated.
The eighteen selected ?rms have been monitored for a period of 14 days using search queries based on keywords
containing the name of the ?rms. Please see table 4-5 for the list of keywords used to ?lter out content from the
social media platforms. As such, all posts that were publicly available have been scraped from the social media
websites. uberVU allowed the exportation of maximum 10.000 posts in CSV format per request. Therefore,
the ?rms were subjected to a request on a daily basis. As a consequence, the search request on day t contained
content that existed yet in the search request of day t ? 1. Figure 4-11 illustrates the overlap in the scraped
content. Before the individual daily search requests were consolidated, the records that existed were removed.
The CSV exportation has been executed on a daily basis, and each addition to the database (except for the
?rst) resulted in the notion that there existed yet certain posts in the database. Therefore, we can conclude
that the social media messages in our database provide a complete overview of the messages created in the
measuring period and related to the sample ?rms.
Day Scraped content
1
2
3
4
5
…
Overlap
Overlap
Overlap
Figure 4-11: Overlap in Scraped Content
The ?gure illustrates that the daily search runs for new social media messages resulted in content that did yet exist
in the database. Messages that existed in the database were removed.
When scraping the social media posts in the database, the following attributes of the posts were recorded: date,
platform, username, content, language, sentiment, gender, followers, pro?le, country, region, city and URL to
the post. The URL is the only attribute that is unique for a post, and is used to determine whether or not a
post existed yet in the database, before it was recorded. It is common that social media messages are copied
from one platform to the other. With our scraping method, identical social media messages created on two
60 Content Analysis
platforms are regarded as two individual posts, and therefore consist two times in our database. Table 4-6
shows a cross-cut of one of the monitored ?rms, showing one scraped post and the corresponding attributes.
As illustrated, the scraper is not able to determine all attributes of a posts. For the example post presented in
table 4-6, the tool was unable to determine the sentiment while is it very easy for a Dutch speaking person to
determine that the content is obviously negative. Probably, this is due to the fact that the content is written
in Dutch, while meta data required to determine the sentiment of the post about this language is not (yet)
available. Also the country and the region are unknown, which is due to the fact that the user that has written
the posts did not agree to share his or her location with to the social media platform.
Table 4-6: Example of Scraped Data
Attribute Example
Date 23-7-2012
Platform twitter
Username Anne
Content @ABNAMRO De internet site doet het nog steeds niet... Ik kan
dus geen geld overmaken nu. Dat is mijn probleem nu.
Language dutch
Sentiment unknown
Gender f
Followers S
Country unknown
Region unknown
Pro?le http://twitter.com/AnneXD_
URL http://twitter.com/AnneXD_/statuses/227776320254382080
4-4-3 Data Cleaning
Before the social media data is ready for analysis, it requires cleaning. Though we used proper search terms, not
all these posts actually relate to the ?rms. The search terms used to monitor the selected companies resulted
in posts that did not have any relation with the selected companies. For instance, Twitter user names like
@klm_klm_klm, @KLM_350, @KLM_2013, @KLM_babe and @klm_luvsya existed in the dataset belonging
to KLM, though these Twitter accounts do not have any relation with KLM (the company). This so-called
‘noise’ has been classi?ed as spam. The existence of noise in datasets is especially applicable on social media
data. Therefore, any organisation that using social media data should ?lter out the valuable content from the
noise.
4-5 Descriptive Statistical Analysis
This section describes the statistics that are acquired by the collection of the social media messages. More
speci?cally, three topics are discussed in the following section. First, the distribution of the sources of the social
media messages are presented (section 4-5-1). This distribution will reveal – taken into account the publicly
accessible social media posts – which social media platforms are mostly used by customers to discuss ?rms
and ?rms’ products and services, hence providing ?rms insights in ‘where to look’ for ?rm-related social media
content. Second, the volume of ?rm-related social media posts are examined (section 4-5-2). The volume of
these messages is analysed per industry. As such, we gain insight in the amount of user-generated content that
is created in di?erent industries. The average daily mentions are also analysed per customer relation type.
The ?nal topic that is discussed in this section describes the statistics of the classi?ed social media posts into
categories (section 4-5-3). These categories are related to di?erent sort of KPIs. Therefore this analysis will
reveal which sort of KPIs are likely to be in?uenced by social media activities.
4-5-1 Channel Distribution
The social media messages have been collected from a variety of social media platforms. Figure 4-12 presents
the distribution of all collected posts along the social media channels. As can be concluded, the largest share
4-5 Descriptive Statistical Analysis 61
(83%) of the collected posts have been created on Twitter. These ?ndings are in line with A. N. Smith et al.
(2012). While Facebook and other social media channels are responsible for a much smaller portion of the
posts according to this dataset, one should place a note to these data. Facebook pro?les may namely be set
unattainable for non-friends. Therefore, the scraper – like any other web scraper – was unable to extract data
from private pro?les. Although one can argue that this distribution may provide an incomplete view of the
situation, it is representative for a real situation in which a ?rm would collect social media posts from the
popular sites because it holds also for a ?rm that it cannot access private social media pro?les of people.
8% 3%
5%
n=224.687
83%
Facebook Twitter Blogs News Other Platforms
Figure 4-12: Social Media Channel Distribution
Illustrates the sources of the social media messages in our sample.
The channel distribution di?ers per ?rm, as shown in table 4-7. For each ?rm, the table shows from which
channels the messages have been collected. As can be concluded, it holds for all ?rms that the majority of the
publicly accessible messages are created on Twitter. The table shows one remarkable value, the collected social
media posts of ABN AMRO are for 55% created from Picasa, a photo sharing platform. This ?gure can be
perceived as a one-time event, because ABN AMRO has uploaded pictures from marketing events (KLM Open,
World Tennis Tournament) that the ?rm organises to their Picasa pro?le. The web scraper recognised each
individual picture as a separate social media post. Appendix C shows the channel distribution on a ?rm to ?rm
basis graphically.
Table 4-7: Social Media Channel Distribution
Shows the absolute number of messages that have been collected from the various platforms. Furthermore, it shows
the percentages of the platforms from which the messages have been collected.
Platform Facebook Twitter Blogs News Other Total
Abs % Abs % Abs % Abs % Abs % Abs
ABN AMRO 124 2% 3.000 42% 70 1% 15 0% 3.858 55% 7.067
Aegon 110 8% 1.173 81% 79 5% 20 1% 67 5% 1.449
Akzo Nobel 30 3% 806 87% 43 5% 25 3% 18 2% 922
Albert Heijn 328 3% 11.116 96% 77 1% 1 0% 59 1% 11.581
Arcadis 8 2% 422 93% 9 2% 10 2% 6 1% 455
ArcelorMittal 439 8% 4.569 83% 296 5% 89 2% 139 3% 5.532
Blokker 155 6% 2.526 91% 71 3% 3 0% 14 1% 2.769
Bol.com 472 8% 5.124 89% 115 2% - 0% 71 1% 5.782
C-1000 362 3% 10.583 96% 81 1% 5 0% 33 0% 11.064
Coca-Cola 1.653 5% 29.347 89% 999 3% 69 0% 885 3% 32.953
Fugro 6 1% 385 90% 20 5% 15 4% 2 0% 428
Heineken 5.726 15% 32.332 82% 494 1% 122 0% 751 2% 39.425
KLM 2.316 9% 22.601 86% 617 2% 90 0% 740 3% 26.364
NS 703 12% 4.970 85% 103 2% - 0% 87 1% 5.863
Philips 4.641 12% 26.260 68% 3.404 9% 138 0% 4.007 10% 38.450
PostNL 77 6% 1.207 91% 27 2% - 0% 12 1% 1.323
TomTom 1.308 4% 29.787 91% 630 2% 67 0% 956 3% 32.748
Unibail-Rodamco 3 1% 487 95% 9 2% 12 2% 1 0% 512
Total 18.461 8% 186.695 83% 7.144 3% 681 0% 11.706 5% 224.687
62 Content Analysis
4-5-2 Volume of Firm-Related Social Media Messages
As described in chapter 1, this thesis analyses the availability of user-generated social media content on two
dimensions. These dimensions are customer relation type and industry type. The availability of user-generated
content has empirically been measured during a period of time. In this thesis, the variable called average daily
mentions serves as a measure to describe the amount – or volume – of generated ?rm-related user-generated
social media content.
The eighteen ?rms have been monitored for a period of two weeks. Each time that a social media post that
contained one of the ?rms’ names and that was publicly accessible has been downloaded. Some ?rms were
mentioned more extensive than others. As such, it is possible to gain insight in the number of social media
messages that are daily generated on the web per ?rm. Table 4-8 shows for each ?rm how many posts have
been collected. The ?nal column shows how many times – on average – the ?rm has been mentioned on a daily
basis. The average daily mentions for each ?rm i have been calculated based on formula 4-1.
Average_Daily_Mentions
i
=
Total_Collected_Posts
i
Measured_Days
i
=
Total_Collected_Posts
i
MAX_Date
i
? MIN_Date
i
(4-1)
Table 4-8: Average Daily Mentions per Firm
The second column of the table indicates the total amount of messages that have been collected in relation with
the ?rm. In the third column, this number is divided by the number of days at which messages have been found,
hence representing the average daily mentions of the ?rms.
Firm Collected posts Average daily mentions
1 ABN AMRO 7.067 544
2 Aegon 1.449 111
3 Akzo Nobel 922 77
4 Albert Heijn 11.581 965
5 Arcadis 455 38
6 ArcelorMittal 5.532 461
7 Blokker 2.769 231
8 Bol.com 5.782 482
9 C-1000 11.064 922
10 Coca-Cola 32.953 2.996
11 Fugro 428 36
12 Heineken 39.425 3.285
13 KLM 26.364 2.197
14 NS 5.863 489
15 Philips 38.450 2.958
16 PostNL 1.323 102
17 TomTom 32.748 2.519
18 Unibail-Rodamco 512 39
? 224.687
Volume per Firm
As can be concluded from the ?nal column in table 4-8, the average daily mentions di?ers from ?rm to ?rm.
From this table, we can conclude that the available user-generated content di?ers from ?rm to ?rm, and that
the applicability of social media data for business intelligence purposes will not be possible for all ?rms, since
not for each ?rm UGC is generated. Figure 4-13 illustrates the average daily mentions of di?erent ?rms in
our sample. The ?gure as been ordered from highly mentioned ?rms to less mentioned ?rms. In the following
paragraph, the volume of social media messages is investigated from a customer relation type perspective.
Volume per Customer Relation Type
Our sample consists of a mix of ?rms that pursue a B2C or a B2B relation. One of the objectives of this thesis
is to investigate whether and to what extent B2C ?rms are more often subject of discussion on social media
than B2B ?rms. With the collected data we can analyse this topic.
4-5 Descriptive Statistical Analysis 63
Heineken
Coca-Cola
Philips
3.000
3.500
TomTom
KLM
2.000
2.500
t
i
o
n
s
[
x
1
/
d
a
y
]
Albert Heijn
1 000
1.500
2.000
A
v
e
r
a
g
e
D
a
i
l
y
M
e
n
t
i
o
n
s
[
Albert Heijn
C-1000
ABN AMRO
NS Bol.com
Arcelor
Mittal
Blokker
Aegon P tNL Unibail
500
1.000
Blokker
Aegon PostNL
AkzoNobel
Unibail
Rodamco Arcadis Fugro
0
Firm
Figure 4-13: Average Daily Mentions of Firms
Bar chart illustrating the variation in the volume of ?rm-related social media posts. Firms have been ordered
descending.
Heineken
Coca-Cola
Philips
3.000
3.500
TomTom
KLM
2.500
x
1
/d
a
y
]
1.500
2.000
A
v
e
r
a
g
e
D
a
ily
M
e
n
t
io
n
s
[
x
1
/d
a
y
Bol.com NS
Albert Heijn
C-1000
ABN AMRO Arcelor
Mittal
500
1.000
A
v
e
r
Bol.com NS
PostNL
Blokker
Aegon Unibail
Rodamco
AkzoNobel
Arcelor
Mittal
Arcadis Fugro
0
500
B2B B2C
Figure 4-14: Average Daily Mentions of Firms
An overview of the average daily produced ?rm-related social media messages. Firms have been clustered based on
their customer relation type.
Figure 4-14 shows the daily volume of ?rm-related social media content, in which the ?rms are clustered on
their relation type and consequently ordered descending. This ?gure suggests that B2C ?rms – coloured in red
– are more likely to ?nd social media content that is related to their ?rm than ?rms performing B2B relations
(coloured in blue). Figure 4-14 shows one remarkable value; the average daily mentions of ArcelorMittal. When
analysing the content of the messages related to this ?rm, the explanation is discovered. ArcelorMittal has
constructed the belvedere for the Olympic Games, called the ArcelorMittal Orbit. During the measurement
period, the tower has been opened for the public, leading to discussions on social media.
In table 4-9, the average daily mentions of B2C ?rms have been consolidated, as are the B2B ?rms. Thus,
the ?nal column of table 4-9 presents an average of an average. Hence, the values are normalised and thereby
eliminating the fact that the number of respondents di?ers between the two groups. The ?rst hypothesis
formulated at the beginning of this chapter was:
H
1
: The volume of ?rm-related social media messages is higher for B2C ?rms than for B2B ?rms.
64 Content Analysis
Table 4-9: Average Daily Mentions
The table consolidates the messages of all ?rms operating the same customer relation type, i.e. B2C or B2B. The
?nal column illustrates the average daily mentions of an individual ?rm operating either a B2C or B2B relation.
Customer relation Collected posts Average daily mentions per ?rm
B2C 216.838 1.369
B2B 7.849 130
? 224.687
With the ?gures presented in table 4-9, a bar chart is created in order to draw conclusions with respect to the
?rst hypothesis. Figure 4-15 depicts the average daily volume of ?rm-related social media messages, consolidated
per customer relation type.
? = 1.369
? = 1.223
N = 13
600
800
1.000
1.200
1.400
1.600
g
e
D
a
i
l
y
M
e
n
t
i
o
n
s
[
x
1
/
d
a
y
]
? = 130
? = 186
N = 5
0
200
400
600
B2C B2B
A
v
e
r
a
g
Figure 4-15: Average Daily Mentions of Firms
Bar chart illustrating the average daily volume of ?rm-related social media messages. Firms have been consolidated
per customer relation type.
Figure 4-15 strongly suggests that B2C ?rms are far more often subject of discussion on social media sites than
B2B ?rms. Thus, the results of our content analysis strongly suggest that the ?rst hypothesis is to be accepted,
implying that the volume of ?rm-related social media messages di?ers for performing B2B or B2C relations,
with B2C ?rms being highly more mentioned on social media than B2B ?rms.
Volume per Industry Type
The second dimension on which the volume of ?rm-related social media content is investigated relates to
industries. Our sample consists of eighteen ?rms active in seven di?erent industries, see table 4-3 for an
overview. As a ?rst step to identify possible di?erences in the volume of daily messages between industries,
the ?rms have been clustered on industry type in ?gure 4-16, and have consequently been sorted in descending
order.
Figure 4-16 suggests that there exists a di?erence in the amount of user-generated content between di?erent
industries. Therefore, the di?erent volumes are consolidated per industry, and analysed in the following
paragraphs. Furthermore, ?gure 4-16 reveals that while an industry average may be lower than the average
of an other industry, an individual ?rm may still be mentioned higher than a ?rm of an other industry. For
example, ABN AMRO is mentioned more often than PostNL, while the industry ?nancial institutions is on
average less mentioned than the transport & storage industry. These insights suggest that there are company
speci?c aspects that also in?uence the amount of messages that are created in relation to the ?rms, i.e. the
industry type is not the only aspect in?uencing the volume of ?rm-related messages.
We examine the availability of user-generated content in the di?erent industries by comparing the average daily
mentions of the di?erent industries with each other. Table 4-10 presents the number of mentions of ?rms,
4-5 Descriptive Statistical Analysis 65
Heineken
Coca-Cola
Philips
3.000
3.500
TomTom
KLM
2.500
x
1
/d
a
y
]
1.500
2.000
A
v
e
r
a
g
e
D
a
ily
M
e
n
t
io
n
s
[
x
1
/d
a
y
Albert Heijn
C-1000
ABN AMRO
NS Bol.com
Arcelor
Mittal
500
1.000
A
v
e
r
NS Bol.com
Arcelor
Mittal
Blokker
Aegon PostNL
AkzoNobel
Unibail
Rodamco Arcadis Fugro
0
500
Information&
Communication
Industry Transport &Storage Wholesale &Retail Financial Institutions Mining&Quarrying
Consultancy, Research
&Other Specialised
Business Services Communication Business Services
Figure 4-16: Average Daily Mentions of Firms
Bar chart illustrating the average daily volume of ?rm-related messages. Firm have been clustered over the industries
and consequently ordered descending.
consolidated across di?erent industries. The second column of the table shows the total amount of social media
posts that have been collected in the corresponding industry. The third column shows the average daily mentions
of a ?rm in the corresponding industry. The values in the third column thus represent an average of an average,
thereby eliminating the fact that the number of ?rms – respondents – di?ers per industry type. Figure 4-17
presents these values. This ?gure suggests that the existence of ?rm-related UGC di?ers among industry type,
implying that di?ers per industry whether or not there exists user-generated content on social media.
Table 4-10: Average Daily Mentions, Consolidated per Industry
Industry Total collected posts Average daily mentions per ?rm
Mining and quarrying 44.904 269
Industry 72.378 3.080
Wholesale and retail 25.414 706
Transport and storage 33.550 929
Information and communication 38.530 1.500
Financial institutions 9.028 231
Consultancy, research and other
specialised business services
883 37
? 224.687
Figure 4-16, 4-17 and table 4-10 provided insight in the variations in the volume of ?rm-related social media
messages across di?erent industries. With this insight, we can examine the second hypothesis of this chapter,
which was formulated as:
H
2
: The volume of ?rm-related social media messages di?ers between industries.
The results of our analysis illustrate variations in the volume of ?rm-related social media messages, which
suggest – based on our sample – that there exists a variation in the daily volume of social media messages that
are created. When ordered descending, the industry ?rms are mentioned mostly, followed by information and
communication, transport and storage, wholesale and retail, mining and quarrying, ?nancial institutions and
consultancy, research and other specialised business services being the least mentioned on social media. Thus,
the results of this analysis indicate that it matters in which industry a ?rm is active whether or not the ?rm
will be subject on social media. It is therefore that – taken into account our sample – we accept the second
hypothesis, implying that the volume of ?rm-related social media messages di?ers between industries. However,
the categories on the two dimensions that are researched in this thesis are not fully independent. For example,
66 Content Analysis
? = 3.080
? = 179
N = 3
3.000
3.500
2.500
3.000
y
M
e
n
t
io
n
s
[
x
1
/d
a
y
]
? = 1.500
? = 1.441
N = 3
1.500
2.000
A
v
e
r
a
g
e
D
a
i
l
y
M
e
n
t
? = 929
? = 1.115
N = 3
? = 706
? = 412
N = 3
500
1.000
1.500
? = 269
? = 272
N = 2
? = 231
? = 273
N = 3
N = 37
? = 2
N = 2
0
500
Industry Information and Transport and Wholesale and Mining and Financial Consultancy,
Industry Information and
Communication
Transport and
Storage
Wholesale and
Retail
Mining and
Quarrying
Financial
Institutions
Consultancy,
Research and
Other Specialised
Business Services
Figure 4-17: Average Daily Mentions of Firms
Bar chart illustrating the volume of daily mentions for ?rms in di?erent industries. The social media messages
related to ?rms in the same industry have been consolidated.
the category consultancy, research and other specialised business services solely consists of B2B ?rms. This
issue is discussed later in this thesis.
4-5-3 Subjects of Social Media Posts
Next to an assessment of the amount of social media posts that are created on the web, this thesis examines
the subjects of the social media posts in order to link the messages to key-performance indicators. The social
media messages of the sample ?rms have been classi?ed into one of the categories that have been established
in section 4-3-1. These categories are based on ten categories of commonly applied key-performance indicators.
Consequently, the collected social media posts of the ?rms in the sample have been classi?ed into one of these
categories. The results of this activity are documented in appendix B from ?rm to ?rm. In this section, the
subjects of social media posts are analysed. First, the subjects of social media messages are discussed from ?rm
to ?rm. Next, the social media posts are analysed from the two dimensions which are the perspectives of this
thesis. Consequently, we analyse whether or not the customer relation type in?uences the types of subjects that
are contained in social media messages. Finally the same analysis is executed, only this time from an industry
perspective.
Subjects per Firm
Figure 4-18 shows the results of the classi?cation process, in which all ?rms are displayed. The percentages
in the ?gure indicate how many of the classi?ed posts are assigned to the corresponding KPI category. The
colours of the bars represent the main categories of subjects of social media posts. When purely looking at
the colours, it becomes clear that some ?rms’ social media posts contain much ?nancial result (‘orange’) posts,
while others contain a high portion of customer relations (‘red’) posts. Furthermore, we see the existence of
community (‘blue’) posts in each ?rm. For reasons of readability, under-represented categories of subjects have
been grouped under a category called other (‘grey’). Figure 4-18 shows two remarkable values of the ‘other’
category. For KLM, these messages mainly have been classi?ed as being spam. The three letters are used by
other people on social media as well, for instance because these are the initials of the person. Also C-1000 shows
a remarkably high percentage of ‘other’ posts. A closer look at C-1000’s messages reveals that many of these
posts can not be classi?ed into one of the ten categories and are hence classi?ed as ‘unde?ned’ posts. Mainly,
these posts contain expressions of people who use the C-1000 stores as a point of reference to meet each other.
A full overview is presented in appendix B.
Figure 4-18 illustrates that the subjects of social media posts related to ?rms di?ers from ?rm to ?rm. In
4-5 Descriptive Statistical Analysis 67
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
P
e
r
c
e
n
t
a
g
e
o
f
S
o
c
i
a
l
M
e
d
i
a
P
o
s
t
s
0%
10%
Short-termfinancial results Customer relations Community Other
Figure 4-18: Social Media Posts Subject Classi?cation
Stacked bar chart illustrating the percentages of social media posts assigned to di?erent categories based on the
posts’ subjects. Small percentages of categories have been merged under ‘other’.
the following paragraphs, we investigate whether or not it is likely that the factors customer relation type and
industry type a?ect the type of subjects of ?rm-related social media posts.
Subjects per Customer Relation Type
In this paragraph we assess whether the subjects of social media messages di?er for di?erent customer relation
types. Illustratively, we examine amongst other whether or not the percentage of product and service quality
related messages di?ers for ?rms performing a B2B or a B2C relation. As a ?rst step to analyse di?erences
in subjects across B2B and B2C ?rms, the individual ?rms have been grouped into their respective customer
relation type, and the percentages of the subjects have been plotted in ?gure 4-19. Again, for readability issues,
under-represented categories of social media subjects have been grouped under an other category.
The third hypothesis that is examined in this chapter was formulated as:
H
3
: The subjects of ?rm-related social media messages di?er between ?rms performing B2B and
B2C relations.
As can be concluded from ?gure 4-19, social media posts related to B2B ?rms contain a high percentage of
posts related to ?nancial results (‘orange’), while this percentage is under-represented for B2C ?rms. Such
information is not of any additional value for a ?rm, since these posts contain information that is yet available
at the ?rm. On the contrary, social media messages related to B2C ?rms contain a high portion of customer
relations (‘red’) related posts in comparison with B2B ?rms. For both type of ?rms it holds that a high portion
of the social media messages reveal the communities’ perceptions of the ?rm (‘blue’ bars). However, B2B ?rms’
community related social media posts are created by professionals, while in the community messages related
to B2C ?rms, these messages are created by consumers. For a detailed overview of the percentages of subjects
related to each ?rm, please see table B-1 in appendix B. These insights suggest that the subjects of social media
messages related to ?rms that pursue di?erent customer relation types vary, and that the third hypothesis is
to be accepted. Thus, ?rms performing di?erent customer relation types will ?nd di?erent subjects in their
?rm-related social media messages.
68 Content Analysis
40%
50%
60%
70%
80%
90%
100%
P
e
r
c
e
n
t
a
g
e
o
f
S
o
c
i
a
l
M
e
d
i
a
P
o
s
t
s
0%
10%
20%
30%
0%
B2B B2C
P
e
r
c
e
n
Short-termfinancial results Customer relations Community Other
Figure 4-19: Social Media Posts Subject Classi?cation, Consolidated per Customer Relation Type
Stacked bar chart illustrating the portion of social media messages related to di?erent subjects. Firms have been
consolidated per customer relation type. Categories that are under-represented are merged under ‘other’ posts.
Subjects per Industry Type
In the previous paragraph the availability of user-generated social media content related to di?erent social
media post categories has been investigated across the customer type dimension. This section performs the
same analysis, only this time the industry dimension serves as the distinguishing factor of the ?rm types. In the
appendix, ?gure D-1 (page 113) lists – in detail – for each industry the average amount of social media posts
related to the di?erent categories of social media posts.
The fourth hypothesis that is examined in this chapter was formulated as:
H
4
: The subjects of ?rm-related social media messages di?er between industries.
Figure 4-20 depicts the portions of social media messages of the di?erent subjects across the seven industries.
Under-represented subjects of social media messages have been merged in the other category. When looking at
the colours of the bars, di?erences are seen – again – in customer relations (‘red’) type of posts and ?nancial
results (‘orange’) posts. E.g. the posts related to wholesale and retail ?rms contain a higher portion of customer
relation posts than ?nancial results posts. The contrary is seen in mining and quarrying, and consultancy ?rms.
Thus, our results suggest that the subjects of social media messages di?er per industry type, implying that the
fourth hypothesis is to be accepted.
4-6 Interpretation of the Results
In the ?fth step of the content analysis, the results are interpreted into meaningful conclusions. In the beginning
of this chapter, the hypotheses that are to be examined by the content analysis have been formulated. These
hypotheses relate to two aspects: (i) volume of social media posts, and (ii) subjects of social media posts. The
content analysis has been designed and executed in a manner to examine these hypotheses, of which the results
are discussed in this section.
4-6-1 Volume of Social Media Posts related to Firms
The collecting process of social media messages related to the ?rms in our sample resulted in di?erent amounts
of messages for di?erent ?rms. E.g. for Heineken, 39.425 messages have been collected while only 428 posts have
4-6 Interpretation of the Results 69
20%
30%
40%
50%
60%
70%
80%
90%
100%
P
e
r
c
e
n
t
a
g
e
o
f
S
o
c
i
a
l
M
e
d
i
a
P
o
s
t
s
0%
10%
20%
Consultancy, research and
other specialised business
services
Mining and quarrying Financial institutions Transport and storage Information and communication Wholesale and retail Industry
Short-termfinancial results Customer relations Community Other
Figure 4-20: Social Media Posts Subject Classi?cation
The ?gure consolidates social media messages of ?rms in the same industries. The ?gure illustrates that the subjects
of social media messages related to ?rms di?ers per industry type.
been found that are related to Fugro. We constructed a variable labelled average daily mentions by dividing
the total amount of collected posts for a ?rm by the amount of days that the ?rm has been monitored. The
variable average daily mentions is deemed as the variable that re?ects the volume of social media posts. In our
sample, the mean average daily mentions amounts 1.025 messages per ?rm per day.
The average daily mentions di?ers from ?rm to ?rm. In order to draw generic conclusions, i.e. not ?rm speci?c,
the ?rms were ?rstly grouped into either business-to-business or busines-to-consumer ?rms. As the analysis
showed, ?rms conducting a B2C relation will ?nd more social media posts that are related to them than ?rms
active in the B2B sector. In our sample, we ?nd an average of 130 daily mentions of a B2B ?rm, while a B2C
?rm is daily – on average – 1.369 times mentioned. Secondly, the ?rms have been grouped into seven industries.
Again, the average daily mentions have been analysed and di?erences between the volumes are illustrated.
4-6-2 Subjects of Social Media Posts related to Firms
The second aspect of this thesis’ content analysis analyses the subjects of the social media posts. As discussed,
the subjects of the social media messages are to be linked to the ?rm’s key-performance indicators. The collected
social media messages have been classi?ed into one of the 28 categories of social media messages that have been
established. Figure 4-18 (page 67) illustrates that subjects of social media messages di?er from ?rm to ?rm.
In order to draw generic conclusions of the subjects of the social media messages related to di?erent ?rms, the
?rms have ?rstly been grouped based on their customer relation type. We can conclude that the subjects of
social media messages related to B2B ?rms contain a higher percentage of short term ?nancial results, news
and professionals related messages than messages related to B2C ?rms. Next, the analysis indicates that the
social media messages related to B2C ?rms contain a higher percentage of posts related to customer relations,
product and service quality and product and service innovation than messages related to B2B ?rms. Secondly,
?rms have been grouped into seven di?erent industries. In the same way as with the analysis of the volume
of social messages, we ?nd that the subjects of social media messages di?er among the ?rms participating in
the di?erent industries of our sample. The majority of social media messages related to ?rms (41%) express
how the external stakeholders of a ?rm perceive the company. In this thesis, such posts have been classi?ed as
community posts. 18% of the social media messages in our dataset contained the name of a ?rm, but did not
70 Content Analysis
contain any valuable information for the ?rm and have consequently been assigned as unde?ned posts. About
11% of the social media messages relate to ?nancial results, which consist of ?nancial performance discussions
(5%) and stock related discussions (6%). Table 4-11 lists the interpretation of the results of the content analysis
in a summarily manner.
Table 4-11: Conclusion of Content Analysis
Volume of Social Media Messages Subjects of Social Media Messages
Customer
Relation
B2C ?rms are more often subject of
discussion on social media than B2B
?rms.
B2C ?rms related social media messages are more
often subjected to customer relations, product and
service quality and product and service innovation
than B2B related ?rms. On the other hand, B2B
?rms’ related messages are more often subjected to
?nancial results, news and professionals discussing
the ?rm. However, the information contained in the
social media posts of B2B ?rms is often yet available
to the ?rm, hence not o?ering added value to the
?rm’s richness of management information.
Industry Our analysis shows a variation in
the volume of social media messages
across di?erent industries.
Our analysis indicates that there is a di?erence in
the subjects of social media posts related to ?rms in
di?erent industries.
4-7 Sub Conclusion: Social Media Posts that relate to KPI Categories
and the Performance Prism Perspectives
Firstly, our analysis indicates that there exists a di?erence in the volume of ?rm-related social media messages
that are daily generated. These di?erences are indicated when comparing ?rms with ?rms, but also when we
compare between B2B and B2C ?rms and when a comparison between di?erent industries is made. With respect
to the volume of ?rm-related social media content, we can state that especially B2C ?rms are able to collect
social media data for business intelligence purposes because it are these ?rms that are subject of interest on
social media sites.
Secondly, in our analysis social media messages have been assigned to di?erent categories of KPIs. The assigning
of messages to KPI categories was based on the subject of the messages. As the analysis indicated, it di?ers
from ?rm to ?rm which kind of KPIs are candidates to be measured using social media data. When taking a
customer relation type perspective, the analysis indicates that KPIs related to the community, i.e. the metrics
that re?ect the attitude of external stakeholders towards the ?rms, are the ones that are particularly suited
to be measured using social media data. Social media messages related to community metrics provide a ?rm
with insight that cannot be generated with internal systems, the information contained in these messages are
created by individuals discussing the ?rm and/or the ?rm’s products / services. Additionally, we see that a
substantial part of the social media messages related to B2C ?rms are related to customer relations metrics.
These messages contain questions and/or complaints of customers and are purposed to get in contact with the
?rm. Several ?rms embrace these messages by establishing a web care team that actively responds to customers
writing messages purposed to contact the ?rm. As regards to B2B ?rms, we see that a high percentage of
the social media posts relate to short-term ?nancial results. Regrettably, the information in these messages
are also available to the ?rm without the existence of social media messages. Most likely, the ?rm is aware of
the information in these messages before it is available on social media. Next, we see that the percentage of
professionals posts (a sub category of community) is higher for B2B ?rms than for B2C ?rms. These messages
contain valuable information for the ?rm, such as market analyses and the position of the ?rm in that situation,
or forecasts for macro economic developments and the e?ects on the ?rm and/or the ?rm’s ecosystem. However,
we have to bear in mind that the volume of social media messages related to B2B ?rms is much lower than for
B2C ?rms.
In chapter 2, a framework has been presented that illustrates the relation between the ?ve performance prism
perspectives and ten categories of key-performance indicators (page 23). As indicated, this thesis examines
possibilities to measure operational performance – i.e. the key-performance indicators of a ?rm – by means of
4-7 Sub Conclusion: Social Media Posts that relate to KPI Categories and the Performance Prism Perspectives 71
4%
Stakeholder
Satisfaction
Strategies
Processes
Capabilities
Stakeholder
Contribution
1. Short-termfinancial results
2. Customer relations 3. Employee relations
4. Operational performance
5. Product and service quality
6. Alliances
7. Supplier relations
8. Environmental performance
9. Product and service innovation
10. Community
KPI Category
Performance PrismPerspective
Social Media Data
9%
11%
0%
0%
0%
1%
41%
0%
Represented in social media data
1%
Not represented in social media data
Mediumrepresented in social media data
Figure 4-21: Social Media Data related to Key-Performance Indicators
Figure illustrates the link between social media data, key-performance indicators and performance prism perspectives.
The colours indicate for which KPI categories - and hence for which performance prism perspectives - social media
data can be found that relates to these categories.
social media data. With the knowledge we gained with a content analysis of a sample of social media messages
related to ?rms, we can conclude which categories of KPIs are candidates to be measured using social media data.
Figure 4-21 schematically shows for which type of KPIs there exists social media data that is related to these
KPIs. KPI categories that were not represented in the sample – i.e. smaller than 1% of the total messages that
have been classi?ed – have been coloured red, implying that the respective KPI category is under-represented in
social media data and hence not able to be measured by means of social intelligence. These under-represented
KPI categories are related to alliance metrics, supplier relations metrics, environmental performance metrics
and operational performance metrics. Next, categories of KPIs that were somehow represented in the analysed
messages – i.e. between 1% and 5% of all classi?ed messages – have been coloured orange. As can be concluded,
these KPI categories relate to employee relations, product and service quality and product and service innovation.
Finally, categories of KPIs that were – in comparison with the other categories – highly represented in the sample
data have been coloured in green. Formally, KPI categories of which more than 5% of the sample data could
be assigned to the respective category have been coloured green.
Consequently, since section 2-4 yet established links between the ten KPI categories and the ?ve performance
prism perspectives, we can draw conclusions on the applicability of social business intelligence for the di?erent
performance prism perspectives. Corresponding to the colours of the KPI categories, the ?ve performance
prism perspectives have been coloured red, yellow or green. The colours represent the applicability of social
intelligence for the di?erent perspectives. As can be concluded, performance metrics in the domains stakeholder
contribution and stakeholder satisfaction are especially suited to be measured using social media data.
The results of the content analysis showed that the applicability of social business intelligence di?ers from ?rm
to ?rm. Especially ?rms B2C ?rms are likely to ?nd ?rm-related messages. Whereas ?gure 4-21 shows the
overall percentages of social media messages related to the di?erent KPI categories, these ?gures vary from ?rm
to ?rm and are thus higher for B2C ?rms. For a detailed overview, see appendix D).
Chapter 5
Blueprint of a Social Business Intelligence
Procedure
Chapter 4 illustrated that ?rm-related social media messages contain information that can be linked to a ?rm’s
key-performance indicators. However, chapter 4 also showed that not all categories of KPIs can be linked to
the content created on social media, simply because the user-generated social media content does not relate
to all categories of key-performance indicators. For those KPIs that are related to the subjects of the social
media messages, a procedure is required that prescribes how a ?rm should acquire and process these social
media messages for business intelligence purposes. A blueprint for a procedure in which social media data are
collected and processed in a way that corresponds with companies’ general business intelligence processes is
developed in this chapter. We refer to such a procedure as a social business intelligence procedure. Given the
insights gained in chapter 4, we state that the following chapter is only relevant for certain ?rms; ?rms that
are mentioned on social media. Firms that are unable to ?nd related social media data should not invest in the
development of social business intelligence procedures.
Section 5-1 starts this chapter by formulating the requirements of a social business intelligence procedure.
Next, section 5-2 provides the blueprint of the procedure, and discusses the necessary steps that form the
procedure. In section 5-3 the procedure is veri?ed. In section 5-4, the real-time aspect of social business
intelligence is discussed. In section 5-5 traditional business intelligence procedures are compared with social
business intelligence. Finally, section 5-6 concludes the ?ndings of this chapter.
5-1 Requirements Formulation
Based on the business intelligence concepts that are discussed in chapter 2, the possibilities of social media
monitoring tools that are discussed in chapter 3, and the experience that we gained in chapter 4 by performing
a content analysis on the social media messages related to di?erent ?rms, nineteen requirements for a business
intelligence procedure have been formulated. Section 5-1-1 describes these requirements. In section 5-1-2 the
formulated requirements are veri?ed on consistency with general business intelligence procedures.
5-1-1 Description of Requirements
Table 5-1 lists the requirements for a social business intelligence procedure. These requirements are discussed
in the following sections. A social business intelligence procedure should:
1. Have access to social media platforms
Obviously, to acquire intelligence from social media messages, a ?rm should have access to the platforms
where these messages are produced.
2. Identify the social media platforms at which the ?rm is discussed
The fundamental purpose of social intelligence is to acquire insight in the the perception of the ?rm’s
5-1 Requirements Formulation 73
Table 5-1: Requirements
A social business intelligence procedure should . . .
1 Have access to social media platforms
2 Identify the social media platforms at which the ?rm is discussed
3 Identify the volume of social media messages related to the ?rm
4 Remove the spam from social media messages that initially seemed to relate to the ?rm
5 Anonymise personal data
6 Identify who the people are that discuss the ?rm on social media
7 Identify what the subjects of the social media messages related to the ?rm are
8 Determine whether the information contained in the social media messages related to the ?rm
o?ers additional value
9 (Automatically) Classify the social media messages related to a ?rm into categories
10 Relate the (categories of) subjects of the social media messages to the ?rm’s key-performance
indicators
11 Determine the ?rm’s social reputation
12 Determine the social reputation of the ?rm’s product(s)
13 Determine relations between social media metrics and the ?rm’s (social) key-performance indicators
14 Update the status of the social media metrics and the values of the KPIs constantly
15 Present the slope of the relations between social media metrics and KPIs on a time chart
16 Interpret the gained intelligence and position it into the ?rm’s developments
17 Assign the gained intelligence to the right persons in a ?rm
18 Allow a ?rm to engage on social media platforms
19 Regularly update the search terms to anticipate on changes
external stakeholders – including (potential) customers, competitors and parter ?rms – towards the ?rm
and/or the ?rm’s products/services. As illustrated in chapter 4, the majority of social media messages
that are related to ?rms, and publicly accessible, are written on Twitter. However, the distribution of
the platforms where the ?rm is discussed may vary from ?rm to ?rm. Therefore, before starting the
monitoring of social media messages, a ?rm should investigate where – i.e. on which platforms – the ?rm
is subject of discussion.
An overview of the platform distribution provides a ?rm insight into which social media platforms the
?rm should focus, engage or advertise. Though a ?rm may be subject of discussion on multiple platforms,
it does not imply that the ?rm is required to monitor these platforms individually. The social media
monitoring tools o?er the possibility to monitor and engage on multiple social media platforms through
one dashboard.
3. Identify the volume of social media messages related to the ?rm
The fact that a ?rm is subject of discussion on social media is of less value whenever there are little
messages available for a ?rm to analyse. Furthermore, the existence of more social media messages related
to a ?rm o?er opportunities to identify correlations between the amount of these messages and the ?rm’s
KPIs. Such analyses are of less value when there are little social media messages available. As chapter 4
revealed, the volume of social media messages di?ers from ?rm to ?rm. Especially business-to-consumer
?rms are discussed on social media, implying that these ?rms have the opportunity to acquire social
intelligence. A constant monitoring of the amount of social media messages related to the ?rms allows
for the detection of sudden deviations, illustrating that there is “something going on”, which may require
attention from the ?rm’s management.
4. Remove the spam from social media messages that initially seemed to relate to the ?rm
As experienced in chapter 4, many social media messages contain the name of the ?rm in the post, though
they do not relate to the ?rm. Especially ?rms carrying a commonly used name or an abbreviation (such
as KLM), are likely to receive many spam messages in their social media messages. Though it may help
to use speci?c user names for the ?rm’s web care team (e.g. @KLM_WebCare), the drawback of such a
name is that the ?rm will not detect all ?rm related messages since users will nevertheless use the generic
name in their posts. Spam related messages are to be removed from the dataset since they do not contain
any value for the ?rm.
5. Anonymise personal data
74 Blueprint of a Social Business Intelligence Procedure
As illustrated in section 3-4, the European Commission has drafted new Regulation on processing personal
data. As a consequence, ?rms are not naturally allowed to process data that allows one to retrace a natural
person from that data. In order to be in compliance with the expected new Regulation, ?rms intending
to collect and process social media data should anonymise the personal data.
6. Identify who the people are that discuss the ?rm on social media
Though A. M. Kaplan and Haenlein (2010) argue that the usage of social media is diversifying in terms
of the users’ age, it is wise to determine who the people are that discuss the ?rm on social media. It is
out of scope of this thesis to describe how di?erent customer groups (e.g. di?erent generations, men /
women, di?erent cultures) should be treated, but whenever a ?rm decides to engage into the social media
conversations, it should be aware of the people that make up their social media environment. Furthermore
a ?rm can decide that it does not consider the people that produce the social media messages as critical
customers, and therefore does not undertake any action.
7. Identify what the subjects of the social media messages related to the ?rm are
Whereas the second requirement of a social business intelligence procedure ensures that a ?rm has insight
in the amount of messages that are produced and containing the ?rm’s name, it is also valuable for a
?rm to have insight in what it is that social media users discuss in relation with the company. The
identi?cation of the subjects of social media posts forms the basis for the translation of social media posts
to key-performance indicators. Furthermore, the identi?cation of subjects – combined with the volume
of messages – provides a ?rm with insight in the topics that are “trending”, i.e. popular topics at the
moment. Trending topics related to ?rms can serve as a measure describing what people consider as
important, and which may be action points for the ?rm.
8. Determine whether the information contained in the social media messages related to the
?rm o?ers additional value
The content analysis of a set of social media messages related to ?rms revealed that there are also messages
that do contain the ?rm’s name, but neither do contain information that is of any value for the ?rm. We
have classi?ed posts of no value as unde?ned posts.
Furthermore chapter 4 illustrated that there are also messages that contain information that must be
available to the ?rm without analysing the social media. Especially for B2B ?rms, many social media posts
contain information about the ?nancial performance or share prices of the ?rm. Generally, a publication
or press article has been the source of these messages. These messages do not contain information that is
not available in the ?rm yet, and are therefore considered of less value for the ?rm.
9. (Automatically) Classify the social media messages related to a ?rm into categories
The unstructured character of social media posts makes it that these messages have to be preprocessed
before an analysis can commence. Classifying the messages into categories, e.g. into categories of subjects
(as we have done in chapter 4), categories of languages, categories of men and women, categories of
many or less followers, etc. allows a ?rm to structurally analyse the messages and derive that particular
information that the ?rm is interested in.
The unstructured nature and the large amount of messages that are generated in relation with some ?rms
makes it that social media data can be termed as “big data”. It is therefore desired that the classi?cation
process of the social media messages into categories runs automatically. Automatic classi?ers are existing
solutions to this problems, and are also available for text. These classi?ers require so-called “training sets”
in order to establish criteria at which a piece of text is either classi?ed in e.g. category A or in B. As we
have experienced, the subjects of social media messages di?er from ?rm to ?rm. Two social media posts
containing the word “Senseo” and “TomTom One XL” are both related to a product, but do not contain
the same words. Therefore, training sets should be established for speci?c ?rms. Our classi?cation can
be used to train classi?ers for the ?rms that participated in the sample of this thesis.
10. Relate the (categories of) subjects of the social media messages to the ?rm’s key-performance
indicators
As we have showed in chapter 4, it is possible to classify social media posts into categories that are related
to KPIs. The subjects of the social media posts serve as the basis to assign a certain social media message
to a certain key-performance indicator. For example, when a ?rm manages by a KPI representing the
customer satisfaction towards a certain product, it can use the social media messages related to that
product as a measure to determine the satisfaction level. As we have seen in the previous chapter, TV
commercials are also subject of discussion on social media. A ?rm may determine the success of such a
campaign by counting the messages that relate to the commercial.
5-1 Requirements Formulation 75
11. Determine the ?rm’s social reputation
A social business intelligence procedure should determine the ?rm’s reputation on social media. The
volume of messages related to the ?rm is not of any value whenever there is no insight in the nature
of these messages since it matters whether or not these messages are positive or negative. Social media
monitoring tools o?er the possibility to determine the sentiment of a social media post. Generally, posts
are classi?ed as either positive, neutral or negative. As section 3-3-1 illustrated, the sentiment analyses
may not always be as reliable. However, as we expect, sentiment analysis tools will be improved and able
to determine the sentiment of the most more accurate. The ?rm’s social reputation – e.g. measured by the
percentage of positive posts related to the ?rm – is an interesting indicator that may reveal correlations
with other KPIs, such as sales.
12. Determine the social reputation of the ?rm’s product(s) / service(s)
Whereas it is necessary to determine the ?rm’s social reputation, a ?rm may be interested in the reputation
of a particular product or service that it provides. Again, sentiment analysis is required for theses posts.
The social reputation of products – e.g. measured by the percentage of positive posts related to that
product or service – may reveal correlations with the sales or amount of returns of that product.
13. Determine relations between social media metrics and the ?rm’s (social) key-performance
indicators
One of the fundamental purposes of business intelligence is to identify which activities of a ?rm deliver
value. In order to determine which social media metrics actually relate to the ?rm’s, the social business
intelligence procedure should contain a step in which the relations between social media metrics and the
?rm’s KPIs are determined. An example of such a relation may be the amount of positive messages about
product x and the sales in a certain period of product x.
14. Update the status of the social media metrics and the values of the KPIs constantly
In order to develop real-time business intelligence, the system should automatically monitor the social
media metrics. “This will only be satis?ed whenever the right KPIs are de?ned before the metrics are
monitored” (Azvine et al., 2005). As such, the ?rm gets insight in the values of the social media metrics
and the values of the KPIs.
15. Present the slope of the relations between social media metrics and KPIs on a time chart
Whereas the previous requirement ensures that the information that is derived from social media is
presented in real-time, this requirement ensures that the slope of the values are presented in a way so that
deviations over time are easily recognised. Sudden events may trigger social media metrics to ?uctuate,
these events are able to be noti?ed when the values are presented in a time chart.
16. Interpret the gained intelligence and position it into the ?rm’s developments
Whereas the derived intelligence may reveal relations of social media metrics and KPIs and provide insight
in the external stakeholders’ perceptions of the ?rm, one should always position this intelligence in the
light of developments of the ?rm.
17. Assign the gained intelligence to the right persons in a ?rm
When it turns out that certain KPIs are in?uenced by social media metrics, and these KPIs are not
performing su?ciently, the acquired intelligence should be communicated to the responsible departments
in the ?rm. The departments can provide clarifying factors for the under performing KPIs, and can take
the acquired intelligence (e.g. related to the feature of a certain products) into their decision-making
process.
18. Allow a ?rm to engage on social media platforms
A social business intelligence procedure should allow ?rms to engage with the users on social media. As
the content analysis in chapter 4 illustrated, many ?rms engage in the social media discussions. Though
we cannot verify this statement, it is expected that there will be generated more user-generated content
whenever a ?rm actively participates on social media. We will elaborate about this statement in the
further research section (section 6-6).
19. Regularly update the search terms to anticipate on changes
An up-and-running social business intelligence procedure has been started by search terms that are related
to the ?rm. Since a ?rm is always in development, it will launch new products, services and employees
will come and go. Therefore, the search terms should be updated whenever there are events that in?uence
the required search terms. For example, whereas Microsoft’s search terms include “Windows 7”, it should
add “Windows 8” to these search terms by the time it launches – or pre-launches – this new product.
76 Blueprint of a Social Business Intelligence Procedure
5-1-2 Requirements Check on Business Intelligence Concepts
Chapter 2 described the business intelligence concept as it is applied within ?rms. Especially section 2-3
elaborated about the activities that make up the business intelligence process. Van Beek (2006) argues that a
BI process consists of three main tasks, being (i) registering, (ii) processing and (iii) reacting. Additionally, the
processing task consists of 15 sub tasks required to process the registered data. In total, 17 (1+15+1) tasks can
be distinguished that are required for a business intelligence process. We verify the requirements for a social
business intelligence procedure – that have been established in section 5-1-1 by controlling whether each of the
BI steps are represented by at least one of the requirements that we have established.
As can be seen concluded from table 5-2 (page 77), each activity is represented by at least one requirement.
This allows us to conclude that the requirements of the social business intelligence procedure are consistent
with existing BI procedures.
5-1 Requirements Formulation 77
T
a
b
l
e
5
-
2
:
B
u
s
i
n
e
s
s
I
n
t
e
l
l
i
g
e
n
c
e
A
c
t
i
v
i
t
i
e
s
a
n
d
S
o
c
i
a
l
B
u
s
i
n
e
s
s
I
n
t
e
l
l
i
g
e
n
c
e
R
e
q
u
i
r
e
m
e
n
t
s
S
o
c
i
a
l
B
u
s
i
n
e
s
s
I
n
t
e
l
l
i
g
e
n
c
e
R
e
q
u
i
r
e
m
e
n
t
s
R e g i s t e r
C o l l e c t
F i l t e r
C o m b i n e
A g g r e g a t e
V i s u a l i s e
I n t e r p r e t
I n t e r n a l i s e
R e v i s e
V e r i f y
E n r i c h
S h a r e
R e m e m b e r
D e c i d e
D i s t r i b u t e
A n t i c i p a t e
R e a c t
1
A
c
c
e
s
s
t
o
s
o
c
i
a
l
m
e
d
i
a
p
l
a
t
f
o
r
m
s
2
I
d
e
n
t
i
f
y
t
h
e
s
o
c
i
a
l
m
e
d
i
a
p
l
a
t
f
o
r
m
s
a
t
w
h
i
c
h
t
h
e
?
r
m
i
s
d
i
s
c
u
s
s
e
d
3
I
d
e
n
t
i
f
y
t
h
e
v
o
l
u
m
e
o
f
s
o
c
i
a
l
m
e
d
i
a
m
e
s
s
a
g
e
s
r
e
l
a
t
e
d
t
o
t
h
e
?
r
m
4
R
e
m
o
v
e
t
h
e
s
p
a
m
f
r
o
m
s
o
c
i
a
l
m
e
d
i
a
m
e
s
s
a
g
e
s
t
h
a
t
i
n
i
t
i
a
l
l
y
s
e
e
m
e
d
t
o
r
e
l
a
t
e
t
o
t
h
e
?
r
m
5
A
n
o
n
y
m
i
s
e
p
e
r
s
o
n
a
l
d
a
t
a
6
I
d
e
n
t
i
f
y
w
h
o
t
h
e
p
e
o
p
l
e
a
r
e
t
h
a
t
d
i
s
c
u
s
s
t
h
e
?
r
m
o
n
s
o
c
i
a
l
m
e
d
i
a
7
I
d
e
n
t
i
f
y
w
h
a
t
t
h
e
s
u
b
j
e
c
t
s
o
f
t
h
e
s
o
c
i
a
l
m
e
d
i
a
m
e
s
s
a
g
e
s
r
e
l
a
t
e
d
t
o
t
h
e
?
r
m
a
r
e
8
D
e
t
e
r
m
i
n
e
w
h
e
t
h
e
r
t
h
e
i
n
f
o
r
m
a
t
i
o
n
c
o
n
t
a
i
n
e
d
i
n
t
h
e
s
o
c
i
a
l
m
e
d
i
a
m
e
s
s
a
g
e
s
r
e
l
a
t
e
d
t
o
t
h
e
?
r
m
o
?
e
r
s
a
d
d
i
t
i
o
n
a
l
v
a
l
u
e
9
(
A
u
t
o
m
a
t
i
c
a
l
l
y
)
C
l
a
s
s
i
f
y
t
h
e
s
o
c
i
a
l
m
e
d
i
a
m
e
s
s
a
g
e
s
r
e
l
a
t
e
d
t
o
a
?
r
m
i
n
t
o
c
a
t
e
g
o
r
i
e
s
1
0
R
e
l
a
t
e
t
h
e
(
c
a
t
e
g
o
r
i
e
s
o
f
)
s
u
b
j
e
c
t
s
o
f
t
h
e
s
o
c
i
a
l
m
e
d
i
a
m
e
s
s
a
g
e
s
t
o
t
h
e
?
r
m
’
s
k
e
y
-
p
e
r
f
o
r
m
a
n
c
e
i
n
d
i
c
a
t
o
r
s
1
1
D
e
t
e
r
m
i
n
e
t
h
e
?
r
m
’
s
s
o
c
i
a
l
r
e
p
u
t
a
t
i
o
n
1
2
D
e
t
e
r
m
i
n
e
t
h
e
s
o
c
i
a
l
r
e
p
u
t
a
t
i
o
n
o
f
t
h
e
?
r
m
’
s
p
r
o
d
u
c
t
(
s
)
1
3
D
e
t
e
r
m
i
n
e
r
e
l
a
t
i
o
n
s
b
e
t
w
e
e
n
s
o
c
i
a
l
m
e
d
i
a
m
e
t
r
i
c
s
a
n
d
t
h
e
?
r
m
’
s
(
s
o
c
i
a
l
)
k
e
y
-
p
e
r
f
o
r
m
a
n
c
e
i
n
d
i
c
a
t
o
r
s
1
4
U
p
d
a
t
e
t
h
e
s
t
a
t
u
s
o
f
t
h
e
s
o
c
i
a
l
m
e
d
i
a
m
e
t
r
i
c
s
a
n
d
t
h
e
v
a
l
u
e
s
o
f
t
h
e
K
P
I
s
c
o
n
s
t
a
n
t
l
y
1
5
P
r
e
s
e
n
t
t
h
e
s
l
o
p
e
o
f
t
h
e
r
e
l
a
t
i
o
n
s
b
e
t
w
e
e
n
s
o
c
i
a
l
m
e
d
i
a
m
e
t
r
i
c
s
a
n
d
K
P
I
s
o
n
a
t
i
m
e
c
h
a
r
t
1
6
I
n
t
e
r
p
r
e
t
t
h
e
g
a
i
n
e
d
i
n
t
e
l
l
i
g
e
n
c
e
a
n
d
p
o
s
i
t
i
o
n
i
t
i
n
t
o
t
h
e
?
r
m
’
s
d
e
v
e
l
o
p
m
e
n
t
s
1
7
A
s
s
i
g
n
t
h
e
g
a
i
n
e
d
i
n
t
e
l
l
i
g
e
n
c
e
t
o
t
h
e
r
i
g
h
t
p
e
r
s
o
n
s
i
n
a
?
r
m
1
8
A
l
l
o
w
a
?
r
m
t
o
e
n
g
a
g
e
o
n
s
o
c
i
a
l
m
e
d
i
a
p
l
a
t
f
o
r
m
s
1
9
R
e
g
u
l
a
r
l
y
u
p
d
a
t
e
t
h
e
s
e
a
r
c
h
t
e
r
m
s
t
o
a
n
t
i
c
i
p
a
t
e
o
n
c
h
a
n
g
e
s
78 Blueprint of a Social Business Intelligence Procedure
5-2 Social Business Intelligence Procedure
A blueprint for a social business intelligence (“SBI”) procedure has been developed, in which all requirements
of section 5-1 are taken care of. An aggregate overview of the procedure is presented in ?gure 5-1.
Reaction based on
social intelligence
Strategic
mapping of
KPIs
Reacting
Search terms
Action plan(s) to
respond to gained
intelligence
g
Collecting
Mapping
insights to
business
units
Data pre
Unstructured
data
Information for
business units
Data pre-
processing
Categorising
Analysing
Structured, combined
(and anonymised) data
Categorised data
Figure 5-1: Blueprint: Social Business Intelligence Procedure
The SBI procedure consists of seven main activities that are related to each other, as ?gure 5-1 illustrates. Each
of the main activities are further exempli?ed in the following sections.
5-2-1 Strategic mapping of KPIs
Firms deduct key-performance indicators from their strategy. This process is further elaborated in section 2-2.
The KPIs that a ?rm eventually established are to be measured. As the content analysis of this thesis revealed,
some KPIs are not appropriate to be measured by social media because there does not exist any content that
related to these KPIs. Other KPIs are best measured by internal systems, and some KPIs are properly measured
by social media. From the list of KPIs that a ?rm uses, a selection can be made of indicators that are to be
measured by social media. Illustratively, ?gure 5-2 highlights the KPIs that are to be measured by social media.
These KPIs form the starting point of the social business intelligence procedure, since it are these KPIs for
which social media data is to be collected and analysed.
Strategy
Driver z Driver y Driver x
KPI 4 KPI 3 KPI 2 KPI 1 KPI 6 KPI 5
KPI 1 KPI 4
I. Strategic mapping of KPIs
KPI 3
S
o
c
ia
l M
e
d
ia
K
P
Is
II. Collecting
Social media
categories available
in the data set,
newaction points
Keywords
VII. Reacting
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Figure 5-2: Blueprint: Social Business Intelligence Procedure (Strategic Mapping of KPIs)
The KPIs selected to be measured by social data determine the categories that are to be analysed – i.e. the
subjects of social media messages – and hence the keywords that are to be used in the collecting process. On
the other hand, the available social media data determines whether or not it is possible to measure the KPI by
social media data. After all, a KPI for which no related social media data exist, can not be measured by social
data. Thus, there exists an interaction between on the one hand what a ?rm wants to measure by social media
5-2 Social Business Intelligence Procedure 79
data, and on the other hand what a ?rm is possible to measure using social media data. As we have seen in
chapter 4, not every KPI is subject of discussion on social media.
5-2-2 Collecting
After the ?rst step, in which the KPIs that are to be measured by social media data have been selected, the
data is to be collected. The step is schematically represented in ?gure 5-3
II. Collecting
Facebook
Twitter
Blogs
YouTube
News sites Etc.
Listening to social media channels
@Company_name
#service_x
#event_z
#product_y
@Competitor_na
me
Etc.
Select keywords
I. Strategic mapping of KPIs
II. Data Pre-Processing
Unstructured data
Search
terms
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Figure 5-3: Blueprint: Social Business Intelligence Procedure (Collecting)
Keywords related to the ?rm, the ?rm’s products/services and the selected KPIs are used to “listen” to multiple
social media channels at which the ?rm could be mentioned. The content analysis of this thesis revealed that it
di?ers per ?rm on which social media platform the ?rm is discussed. It is therefore that the ?rst step involving
social media platforms consists of the determination of the platforms at which the ?rm is discussed. As we have
experienced in chapter 4, search queries related to ?rms will result in unstructured data from multiple social
media platforms. These unstructured data are to be pre-processed, which is the next step in the SBI procedure.
5-2-3 Data Pre-Processing
The third step in the social business intelligence procedure consists of pre-processing the collected data. In
contrast to ‘regular’ BI data, social media data is unstructured, sourced from multiple platforms, containing
spam and personal data, and is therefore required to be pre-processed. Figure 5-4 illustrates this process.
III. Data Pre-Processing
Select attributes to
analyse
Combine different
data formats
IV. Categorising
II. Collecting
Unstructured data
Remove
duplicates
Remove spam
Data ready to be categorised
Anonymise
personal data
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Parse data in
table / database
Figure 5-4: Blueprint: Social Business Intelligence Procedure (Data Pre-Processing)
The collected data consists of social media messages that are sourced from multiple sources in di?erent formats,
such as CSV, JSON, XML, etc. Each data source may employ its own structure of social media messages,
80 Blueprint of a Social Business Intelligence Procedure
and not each platform may contain the same richness in attributes as the other. For instance the Twitter API
o?ers developers the opportunity to extract so called geotags – geographic coordinates of the origination of
the Tweet – while other social media platforms do not o?er this attribute to the messages. Each social media
post should be parsed – structured – into one and the same data format. Next, as we have experienced in
the scraping process of chapter 4, multiple search queries will lead to multiple messages yet available in the
database. Therefore, only social media messages that do not exist in the table should be added. The ?nal
step in the data pre-processing step consists of the removal of spam. After the data pre-processing has been
completed, the data is structured, clean and ready to be categorised.
5-2-4 Categorising
The third step in the SBI procedure consists of categorising the social media posts. The purpose of this step is
to divide the messages into clustered categories at which the ?rm is interested. The aspects at which the social
media posts are categorised may vary. Figure 5-5 schematically shows the third step of the SBI procedure.
IV. Categorising
Categorise the
data on specific
aspects
People
Competitive data
Feedback about competitors’
products
Feedback on people’s attitude to
competitors’ organisation
Feedback about latest advertising
campaign
Feedback about product features
Feedback about howpeople
perceive the brand / company
People’s opinions
Data about requests for customer
service
Data about the performance of
customer service
Feedback on the pricing of your
products / services
Example classifications
Less followers
Many followers
Positive speakers
Encouragers
Promotors
Negative speakers
Complainers
Saboteurs
Subjects e.g.
Example classifications
Public image
Customer relations
Recruitment
Product and service quality
Product and service innovation
Professionals’ opinions
Etc.
Trending Topics
III. Data Pre-Processing
Data ready to be categorised
V. Analysing
Categorised data
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Figure 5-5: Blueprint: Social Business Intelligence Procedure (Categorising)
One can decide to analyse the people that create the messages, and group these people in e.g. people with
many/less followers or friends, or into people that write/negative positive about the ?rm. We have labelled
the four categories of people. Encouragers are the people with less followers though speak positive about the
?rm or its products. Complainers are the people with less followers and write negative about the ?rm. People
with many followers who speak positive about the ?rm have been labelled as promoters, while people with
many followers writing negative have been termed saboteurs. An analysis of the people provides the ?rm with
intelligence about the power of the people that write about the ?rm, and may form the starting point of a social
media engagement strategy.
Another aspect at which social media messages may be classi?ed is based on their subjects. Our content analysis
of chapter 4 also categorised social media messages based on their subjects. The subjects that were represented
in our dataset related to public image, customer relations, recruitment, product and service quality, product and
5-2 Social Business Intelligence Procedure 81
service innovation, professionals’ opinions, etc. By classifying posts into categories based on subjects, it becomes
possible to link the volume of messages related to a certain subject to the companies’ corresponding KPIs. For
instance, public image posts – which may be additionally classi?ed as positive, neutral or negative – are related
to a customer satisfaction KPI. There are plenty of other categories that one can think of to categorise social
media messages, but to link the ?rm’s KPIs to social media data, one should classify the messages based on
their subjects.
Whereas the data on social media is generally publicly accessible, it is possible for a ?rm to perform the same
analysis based on search queries related to competitors and competitors’ products. As such, a competitive
analysis provides the company intelligence about their position with respect to the market average.
Furthermore, word counts can be used to determine so called trending topics; topics that are over-represented in
the social media messages related to the ?rm. Trending topics, or a top 10 of the words that are most frequently
used in the social media messages, provide a ?rm insight in the topics that are discussed on social media in
relation with their ?rm.
5-2-5 Analysing
Once the social media data has been structured and cleaned, the analysis of these data can commence in step 5
of the SBI procedure. It is in this step of the procedure where a translation is made from data to information.
This step is visualised in ?gure 5-6. Depending on the matter of interest, a ?rm can analyse a variety of data
and relations. It would be wise to at least plot the conversation volume – or amount of social media messages
related to the ?rm – against the di?erent social media channels to determine where the conversations related to
the ?rm take place. Next, whenever a category has been established in step 4 in which all messages related to a
certain product or product feature have been grouped, it is possible to determine the attitude of the public to
the product by applying sentiment analysis on these data. Such analysis provides the ?rm with insight in the
the products or product features that are to be improved. Furthermore, a comparable analysis on competitors’
social media data will show the ?rm’s position pertaining to the competitors and competitors’ products.
The most valuable intelligence will be gained when the ?rm combines the social media metrics – such as amount
of mentions, sentiment, messages originating from a certain region, etc. – with the companies’ KPIs, such as
sales volume, market share, customer satisfaction and the amount of customers. The slopes that will be gained
when these metrics are together plotted on a time chart may reveal relations. The right part of ?gure 5-6
illustrates such graphs. A correlation analysis may con?rm these relations. The intelligence that is gained in
the analysis phase may reveal that certain social media metrics are under performing, and that these social
media metrics in?uence key-performance indicators of the company. Consequently, a ?rm may undertake actions
to improve these metrics.
5-2-6 Mapping insights to Business Units
Key-performance indicators are related to di?erent departments in a ?rm, and the managers of these departments
may clarify the under-performance of the metrics and they may suggest actions to improve the KPIs. As ?gure
5-7 illustrates, the intelligence provided by step 5 should be communicated to the responsible business units.
Especially when under-performing KPIs are discovered.
For example, insights related to products should be communicated to the ?rm’s research & development
department, customer satisfaction intelligence to the ?rm’s customer relations management department, etc. It
are the employees of the responsible departments who posses the knowledge and experience to reason why a
KPI is under-performing, and – in collaboration with social media experts – are the ones who may develop an
action procedure to improve the indicator.
5-2-7 Reacting
The ?nal step in the SBI procedure comprises the execution of action plans required to improve under-performing
KPIs by means of social media. Illustratively, ?gure 5-8 shows two type of actions that may result from the
social business intelligence procedure. A ?rm can for instance decide to review its products (features) based
on complaints and suggestions that the SBI procedure provided. Or, a ?rm may decide to intervene in social
media discussions, for instance because customer satisfaction turned out to be low, and – at the same time –
the customer service of the ?rm turned out to be insu?cient.
82 Blueprint of a Social Business Intelligence Procedure
V. Analysing
C
o
m
p
e
t
it
o
r
x
C
o
m
p
e
t
it
o
r
y
F
ir
m
Competition
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
T
w
it
t
e
r
B
lo
g
s
F
a
c
e
b
o
o
k
Channel
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
Map KPIs on the graphs, e.g.:
jan apr jul oct
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
Sales volume
S
a
le
s
v
o
lu
m
e
20,000 1,000
Conversation volume
jan apr jul oct
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
Customer satisfaction
C
u
s
t
o
m
e
r
s
a
t
is
f
a
c
t
io
n
20,000 1
Conversation volume
jan apr jul oct
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
Market share
M
a
r
k
e
t
s
h
a
r
e
20,000 100%
Conversation volume
abc
Product feature
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
Sentiment
Determine
correlations to
identify
dependencies
Map social media reputation
Flag
underperforming
social media
metrics
jan apr jul oct
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
Customer database
C
u
s
t
o
m
e
r
d
a
t
a
b
a
s
e
20,000
10,000
Conversation volume
VI. Mapping insights to
business units
IV. Categorising
Categorised data
Information for business units
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data
Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Figure 5-6: Blueprint: Social Business Intelligence Procedure (Analysing)
VII. React
VII. Reacting
VI. Mapping insights to Business Units
Customer
Relations
Management
Research &
Development
Marketing & Sales
Customer
satisfaction
information
Feedback related
to products
Sales related
information
Management
board
Competitive
intelligence
Action plan(s) to
respond to
gained intelligence
Etc.
V. Analysis
V. Analysing
Information for
business units
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data
Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Figure 5-7: Blueprint: Social Business Intelligence Procedure (Mapping Insights to Business Units)
5-3 Veri?cation of Procedure 83
VII. Reacting
Review product /
services
Intervene in social
media
conversations
Other activities
Reaction
based on
social intelligence
VI. Mapping insights to
Business Units
Action plan(s) to respond
to gained intelligence
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data
Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Figure 5-8: Blueprint: Social Business Intelligence Procedure (Reacting)
5-3 Veri?cation of Procedure
The veri?cation of the developed social business intelligence procedure is veri?ed in this section. We will test
whether the requirements established in section 5-1 are ful?lled. For each individual requirement an activity is
searched for that ful?ls the requirement. If all requirements are ful?lled by at least one activity, we can conclude
that the social business intelligence procedure is veri?ed in accordance with the requirements. Table 5-3 (page
84) lists the seven main components of the SBI procedure in the columns, and the eighteen requirements in the
rows of the table. For each requirement, the activity that serves this requirement has been checked. As can be
concluded, each of the eighteen requirements are at least ful?lled by one of the main components. Therefore,
we can conclude that the procedure is in accordance with the requirements, which are in turn in accordance
with the activities required for general business intelligence.
84 Blueprint of a Social Business Intelligence Procedure
Table 5-3: Veri?cation Matrix
S
B
I
C
o
m
p
o
n
e
n
t
s
S
t
r
a
t
e
g
i
c
m
a
p
p
i
n
g
o
f
K
P
I
s
C
o
l
l
e
c
t
i
n
g
D
a
t
a
p
r
e
-
p
r
o
c
e
s
s
i
n
g
C
a
t
e
g
o
r
i
s
i
n
g
A
n
a
l
y
s
i
n
g
M
a
p
p
i
n
g
i
n
s
i
g
h
t
s
t
o
B
u
s
i
n
e
s
s
U
n
i
t
s
R
e
a
c
t
i
n
g
Requirements for a social business intelligence procedure I II III IV V VI VII
1 Access to social media platforms
2 Identify the social media platforms at which the ?rm is discussed
3 Identify the volume of social media messages related to the ?rm
4 Remove the spam from social media messages that initially seemed
to relate to the ?rm
5 Anonymise personal data
6 Identify who the people are that discuss the ?rm on social media
7 Identify what the subjects of the social media messages related to
the ?rm are
8 Determine whether the information contained in the social media
messages related to the ?rm o?ers additional value
9 (Automatically) Classify the social media messages related to a
?rm into categories
10 Relate the (categories of) subjects of the social media messages to
the ?rm’s key-performance indicators
11 Determine the ?rm’s social reputation
12 Determine the social reputation of the ?rm’s product(s)
13 Determine relations between social media metrics and the ?rm’s
(social) key-performance indicators
14 Update the status of the social media metrics and the values of
the KPIs constantly
15 Present the slope of the relations between social media metrics
and KPIs on a time chart
16 Interpret the gained intelligence and position it into the ?rm’s
developments
17 Assign the gained intelligence to the right persons in a ?rm
18 Allow a ?rm to engage on social media platforms
19 Regularly update the search terms to anticipate on changes
5-4 Real-Time Social Business Intelligence 85
5-4 Real-Time Social Business Intelligence
In the introduction of this thesis, the concept of real-time business intelligence has been introduced. One of
the aspects that makes social media data valuable is the fact that it is created real-time and that these data
are directly available. The real-time aspect of social media data is one of the main reasons why this thesis has
been executed in the ?rst place. In this section we pay attention to the speed of the social business intelligence
procedure as it is presented in section 5-2.
In our analysis, a scan for new social media messages has been executed on a daily basis. However, it would be
valuable for a ?rm to be informed directly whenever social media activity related to the ?rm deviates from its
steady state. For example, an increase in the average amount of hourly ?rm-related messages may indicate an
event of which the ?rm should be aware from a risk management perspective. Deviations in the volume of social
media messages are relatively easy to detect, since detection systems simply count the amount of messages that
has been generated in the past period and compare this amount with the average amount. A scan to detect
variations in the volume of ?rm-related messages should be executed periodically, the results are then almost
immediately available. Figure 5-9 shows an illustrative example of a comparison between the average volume
and the actual volume of today. Such a graph would announce a ?rm that it is suddenly more frequent subject
of discussion on social media than in normally. The commercial tool that has been used in this thesis to analyse
social media automatically refreshes the ?rm-related messages, comparable to the streams o?ered by Twitter.
0
10
20
30
40
50
60
70
80
0
-
1
1
-
2
2
-
3
3
-
4
4
-
5
5
-
6
6
-
7
7
-
8
8
-
9
9
-
1
0
1
0
-
1
1
1
1
-
1
2
1
2
-
1
3
1
3
-
1
4
1
4
-
1
5
1
5
-
1
6
1
6
-
1
7
1
8
-
1
9
1
9
-
2
0
2
0
-
2
1
2
1
-
2
2
2
2
-
2
3
2
3
-
0
V
o
l
u
m
e
[
m
e
s
s
a
g
e
/
h
o
u
r
]
Average Today
Figure 5-9: Illustrative example: Variations in volume of ?rm-related social media messages
However, insight in deviations in the volume of ?rm-related social media messages is not su?cient to speak of
social business intelligence. As the procedure in the previous section illustrated, a process of pre-processing,
categorising, analysing and mapping insights to business units is required after a ?rm has determined which
key-performance indicators are to be measured by social media data is required. All these steps are to be
automated by one critical mechanism; text classi?cation. A tool that is able to automatically classify social
media posts into pre-determined categories – e.g. subjects – is a prerequisite for real-time social business
intelligence. As we have seen in the content analysis of this thesis, manually classifying social media posts is
a time-consuming process and therefore not advised. The further development of automatic text-classi?cation
tools and the incorporation of such tools in social media monitoring packages is advised. The programming of
each automatic text classi?er requires a training set of texts that have manually been classi?ed. Our dataset
can be used for such purposes. Furthermore, the automated classi?cation is to be linked with internal data.
So far, business intelligence tools like SAP, Oracle, QlikView, etc. manage and process internal data, whereas
dedicated tools like uberVU, NetBase, Radian6, etc. are used to process external – social – data. Thus, data
from di?erent systems is required to measure the in?uence of social media metrics on organisational performance.
Such systems are to be developed.
In section 3-3-1 we concluded that the current state of social business intelligence can be termed ‘early adoption’.
With the experiences gained in chapter 4 and the required features mentioned in the previous paragraph, the
next (research and development) steps required for social business intelligence systems are determined. These
steps are further elaborated in section 6-6 ‘further research’.
5-5 Social Business Intelligence versus Business Intelligence
The newly developed social business intelligence di?ers from ‘traditional’ business intelligence methods. In this
section, the most important di?erences concerned with social business intelligence are discussed. Table 5-4 lists
86 Blueprint of a Social Business Intelligence Procedure
the main di?erences between the two concepts, which are consequently elaborated.
Table 5-4: ‘Traditional’ and Social Business Intelligence compared
‘Traditional’ Business Intelligence Social Business Intelligence
Data is structured Data is unstructured
Only the values of the data automatically ?uctuate
in the course of time
The nature of the data may alter in the course of
time
Data are mainly numerical Data are mainly textual
Relations between data and KPIs are obvious Relations between data and KPIs are fuzzy
Few sources of data Multiple sources of data
Data sources are internal Data sources are external
One data format Multiple data formats
Origins of the data are known Origins of the data are unknown
Data represents known subjects (products in stock,
sales per month, etc.)
It is beforehand not clear what is contained in the
data (complaints, suggestions, etc.)
Spam
Data may contain personal data for which no explicit
consent of data processing is provided
The far most important aspect of social media data is the fact that it is unstructured. When collecting social
media data, one big mishmash of textual data is found containing di?erent subjects in di?erent languages. On
the contrary, ‘normal’ BI systems source data from structured sources in which pre-determined variables are
stored. As a consequence, the collecting (also referred to as extracting) process will deliver structured data in
any case. In other words, in normal BI systems it is known beforehand what will be measured, whereas the
subjects of social media messages di?er from ?rm to ?rm.
Additionally, the nature of the social media data may alter in the course of time. Due to the fact that anyone
who has access to social media is able to formulate new subjects, it is not unlikely that the subjects contained
in social media messages may vary. This aspect of social business intelligence is not found in normal BI
processes. In normal BI the to be measured variables are pre-determined, the values of these variables will
alter in the course of time. However, the nature of the data will not change. E.g. a metric measuring the
amount of products that are currently in stock will not suddenly start measuring the number of employees
or any other variable. However, subjects contained in social media messages may change. Therefore, social
business intelligence systems are required to cope with ?uctuations in the nature of the data.
Apart from (simple) word counting mechanisms, whether purposed to show the volume of ?rm-related messages
or to show trending topics, social media data is textual in nature. Fundamental data analysis methods and
algorithms underlying traditional business intelligence are developed to process numerical data. Therefore, a
translation step is necessary from raw textual data to numerical data before such analyses can be executed on
social media data. In our social business intelligence procedure this step is provided by categorising the textual
data. The categorising step ensures that equivalent social media posts are grouped – i.e. structured – so that
the number of messages in each group serve as the basis for numerical analyses. E.g. if all messages containing
the subject ‘product x’ have been grouped in one category, the number of the messages in this group are ready
to serve as the input for further (numerical) analyses.
In this thesis we have assigned social media posts to key-performance indicators based on the subjects contained
in these messages. Whereas in traditional business intelligence relations between data and KPIs are obvious
and linear, these relations are less evident in social business intelligence. For instance, the volume of messages
related to a certain product may in?uence the sales of that product. However, this relation is not self-evident
since a causal relation between the two variables is not guaranteed. Other factors – such as lotteries – may
in?uence the chatter volume, but this does not necessarily underwrite the intention of people to buy the product.
The fact remains that these relations may still exist, and it is therefore that monitoring and analysing social
media data in relation with the ?rm’s KPIs may reveal valuable intelligence. As indicated in chapter 4, KPIs
related to customer relations and the perceptions of stakeholders can be measured using social media data.
Whereas in traditional BI data is stored in relatively few (structured) data sources, there are many social media
platforms from which data is to be sourced. These sources di?er from traditional BI sourcing systems since
they are in the external environment of a ?rm, implying that the data formats and other institutional aspects
are determined by the social media platforms. Furthermore, the platforms can change the format in which the
5-6 Sub Conclusion 87
data is delivered. Sourcing data from multiple social media platforms means combining di?erent formats. This
step requires more e?ort than in traditional BI systems. In addition, the attributes that are passed to a ?rm
when it crawls social media data di?ers per platform.
The creators of the data – social media users – are unknown to the ?rm in social business intelligence. As
such, it can be hard to determine the trustworthiness of the data. A user can post whatever he or she wants
on the web, without ensuring that the message actually re?ects his or her opinion or intention. However, there
are ample examples of self-regulating platforms on the web, of which Wikipedia is probably the most famous.
Contributions of users to Wikipedia that are incorrect are automatically corrected by other users with good
intentions. Moreover, since messages are grouped in our procedure, it is rather easy to target the popular topics
(which require attention) and determine the trustworthiness. Next, we expect that natural language processing
tools will be improved, allowing the detection of cynicism and other di?culties concerned with social media
data.
In traditional BI systems it is beforehand crystal clear what will be measured, e.g. the time to assemble a
product from ?ve components, which clearly a?ects the operating expenditures of a company through the cost
of workers. In social business intelligence, the contents in the data are not clear beforehand. As we have seen in
chapter 4, subjects of social media messages di?er from ?rm to ?rm. Thus, not each KPI that a ?rm is willing
to be measured by social media data can actually be measured by these data. It is therefore that the contents
of the social media data determine what can be measured. Each company can be willing to measure KPIs
by social media data, however without existence of any data, this will not be possible. In traditional business
intelligence, a ?rm is much less dependent on external stakeholders for the possibilities of BI.
Reaction based on
social intelligence
Strategic
mapping of
KPIs
Reacting
Search terms
Action plan(s) to
respond to gained
intelligence
g
Collecting
Mapping
insights to
business
units
Data pre
Unstructured
data
Information for
business units
Data pre-
processing
Categorising
Analysing
Structured, combined
(and anonymised) data
Categorised data
Figure 5-10: Speci?c Steps in Social Business Intelligence
When recalling the cycle visualising the social business intelligence procedure developed in section 5-2, the
speci?c steps required in social business intelligence can be highlighted. Figure 5-10 shows the social business
intelligence cycle, in which the speci?c social BI steps are highlighted in orange. In the orange steps, a di?erent
method is required as compared to ‘traditional’ BI. Since our procedure is based on existing business intelligence
methods, the procedure also shows overlap. The steps that are relatively equal to standard BI are coloured
in blue. As can be concluded from ?gure 5-10, di?erent activities are mainly required in the collecting and
processing steps of business intelligence. It is in these steps where the data is converted from unstructured to
structured data that is ready for analysis.
5-6 Sub Conclusion
A social business intelligence procedure (“SBI”) should ?t within the general ‘way of executing’ business
intelligence, since it is not possible that social media metrics measure all key-performance indicators as well
as ?rm’s internal BI systems will do. Therefore, SBI is considered as an additional component to business
88 Blueprint of a Social Business Intelligence Procedure
intelligence, rather than a replacing procedure. However, as the content analysis of chapter 4 illustrated, there
are certain categories of KPIs that are in?uenced by – or at least related to – a substantial amount of social
media messages that are related to the ?rm. For these type of KPIs, which di?er from ?rm to ?rm, a procedure
has been developed that prescribes the necessary steps to acquire, process and ?nally gain intelligence from
?rm related social media messages. The procedure is based on general BI concepts, existing technologic social
media analysis solutions and the experience gained in the execution of a content analysis into the social media
messages related to eighteen di?erent ?rms.
A SBI procedure consists of seven main components, being (i) strategic mapping of KPIs, (ii) collecting, (iii)
data pre-processing, (iv) categorising, (v) analysing, (vi) mapping insights to the business units, and (vii)
reacting. The seven steps can be interpreted as a cycle, i.e. the output of the last step in?uences the ?rst step.
(i) Strategic mapping of KPIs The very ?rst step of social business intelligence sets the scene for the objects
that are to be collected and analysed. Namely, in the ?rst step the key-performance indicators that are to
be measured by social media data are selected. As we have seen in chapter 4, not each type of KPI is to be
measured by social media data since there does simply not exist any related social media data to these types
of KPIs. Firms should mainly focus on KPIs related to customer relations, public image and – to a less extent
– on product and service innovation when selecting KPIs that are to be measured using social media data.
Whenever a ?rm has selected the social KPIs, it can start collecting the appropriate data.
(ii) Collecting The second step of the SBI procedure related to data collection. In contradiction to regular BI
systems, the data is to be sourced from external parties in social business intelligence. People create ?rm-related
messages on di?erent platforms, of which the vast majority of publicly accessible messages are created on Twitter.
The search terms that are used to ?lter out the content at which the ?rm is interested should be based on the
social KPIs selected in the previous step.
(iii) Data pre-processing The social media data has been collected from multiple platforms which adhere to
their own data format. The di?erent format are to be combined into one uniform database, so that – in a
later step – data analysis can be applied on the complete dataset. Furthermore, the ?rm should select those
attributes that are necessary for the analysis, not each platform o?ers the same richness of attributes to a social
media post. In addition, the data should be anonymised to be in compliance with new Regulations regarding
data privacy. Finally, spam – i.e. social media posts that do not relate to the ?rm – should be removed from
the collected data.
(iv) Categorising The data pre-processing step resulted in a structured database in which the social media
messages from multiple platforms are combined. In the categorising step, the messages are clustered on di?erent
issues of interest, depending on the ?rm’s subject of interest. E.g., messages related to certain products can be
categorised, or one can cluster the messages that are created by people with many followers, etc. Again, the
criteria at which the messages are categorised are determined by the selection of the social KPIs in the ?rst
step.
(v) Analysing So far, the collected data has not provided any insights. It is in this step of the social business
intelligence procedure where data is transformed into information. The categories that were established in the
previous step are analysed in this step. For instance, sentiment analysis can be applied on the categories related
to the ?rm’s products in order to acquire intelligence related to customer experiences of the products. However,
the most valuable intelligence is gained when social media data is related to internal data. For instance, the
volume of social media messages related to a certain product may be correlated with the sales volume of that
product. It is in this phase of the SBI procedure where such relations are explored.
(vi) Mapping insights to business units In the ?rst step of the procedure, KPIs have been selected. These
KPIs typically relate to a certain function of the ?rm, and hence have an ‘owner’. The intelligence gained in the
previous step relates to KPIs, and should feed back to the owner of the KPI. Generally, it are the people in the
?rm that are responsible for the KPI who are the ones that can reason how the KPI is in?uenced. Therefore,
these people are the ones that can draft an action plan in case the KPI needs improvement.
5-6 Sub Conclusion 89
(vii) Reacting The ?nal step of the social intelligence procedures consists of the execution of the action plans
that are developed in collaboration with people from the business lines that are responsible for the respective
KPIs. Actions on the gained intelligence may involve revisions of internal processes or strategies, or external
interventions such as social media engagement.
The developed social business intelligence procedure is based on general business intelligence processes, and the
requirements of the social BI procedure have been derived from general BI processes. For the veri?cation of the
SBI procedure, all requirements have been checked on ful?lment by systematically tracking which requirement
is ful?lled by which activity.
Chapter 6
Conclusions & Discussion
First of all, this chapter presents the conclusions of the research in section 6-1. In section 6-2 the contributions
of this work to existing and future research are discussed. Section 6-3 proceeds by discussing the implications of
the ?ndings in this thesis for practice. Section 6-4 consequently re?ects on the thesis and the research process.
In section 6-5 the research is critically reviewed, and limitations are discussed. Finally, section 6-6 provides
suggestions for future research related to the subject of this thesis; social business intelligence.
6-1 Conclusions
Firms are increasingly using social media, while at the same time business intelligence systems are increasingly
applied for performance measurement of business activities. Though these two concepts o?er room for synthesis,
it also raises questions related to the applicability and opportunities o?ered by combining social media and
business intelligence. So far, it is not clear which ?rms are able to ?nd ?rm-related social media data and if
they are able, how these data should be incorporated in the business intelligence processes of ?rms. As one of
the ?rst researches into the opportunities of leveraging social media data for BI purposes, this this was purposed
to draw generic conclusions on the applicability of social business intelligence by distinguishing ?rms on generic
aspects. Firms were distinguished on customer relation type – either B2C or B2B – and on industry type.
Therefore, the main research question of this thesis has been formulated as:
How can ?rms use social media data for business intelligence, taking into account the ?rm’s speci?c
industry and relationship with end-users?
The main research question has been divided into three sub questions, which are answered in the following
sections. The ?rst sub question was de?ned as:
1. What is the current state of social media in relation with business intelligence?
Social media is a natural consequence of Web 2.0, and can be de?ned as Web 2.0 based applications allowing
users to create and share user-generated content with pre-selected users and/or communities. The applications
through which users are active are known as social media platforms. In 2012, there are many social media
platforms available, which di?er in scope and functionality. Each platform adheres to its own policy regarding
data crawling and data format.
Social media is a topic on the agenda of many ?rms in 2012. Though many ?rms acknowledge the opportunities
of social media, there also exists a degree of reluctance from managers towards social media. Research indicates
that executives who avoid social media do not understand what social media is, how to engage with it and learn
from it. On the other hand, ?rms that do embrace the world of social media particularly perform activities in
the ?eld of marketing, customer relations management, reputation management and co-creation / prosuming
activities through the various social media platforms.
Business intelligence – irrespective of the variables to be measured – can be perceived as a cycle consisting
of three main steps; (i) register, (ii) process and (iii) react. Before a BI cycle can commence, it has to be
6-1 Conclusions 91
determined ‘what to measure’. The variables that are to be registered are generally aligned with a ?rm’s
strategy and corresponding business model, termed key-performance indicators (“KPIs”). The three BI steps
are required when a ?rm intends to apply business intelligence on the ?rm’s social media activities. However,
social media data di?ers from “regular” business information. Unlike internal business data, social media data
is created by non-professionals and stored into a variety of databases that are owned by external parties who
employ their own database structure and access limitations. Therefore, a di?erent BI approach is required for
social media data.
Firms employ di?erent key-performance indicators. Especially lower level KPIs are ?rm speci?c, while top level
KPIs are generic and employed by many ?rms. Based on Adam & Neely’s (2001) generic performance prism
perspectives, ten categories of key-performance indicators have been established. The ten categories are de?ned
as (short-term) ?nancial results, customer relations, employee relations, operational performance, product and
service quality, alliances, supplier relations, environmental performance, product and service innovation and
community. It are these categories of KPIs for which related social media data has been searched for.
In social business intelligence, a ?rm analyses the activities on social media related to the ?rm and determines
the e?ect of these activities on the ?rm’s performance. Existing social media monitoring tools – which are
becoming increasingly available on the market – mainly reveal the performance of the ?rm on social media as
a separate component of the ?rm. The intelligence that such monitoring tools provide relate to the volume of
posts, engagement of users, sentiment, geography, topics and themes in the social media messages, in?uencer
ranking, channel distribution, etc. However, the purpose of business intelligence is to reveal the underlying
parameters that determine the ?rm’s performance, that is, not limited to solely social media performance.
In order to understand the in?uence of social media activities on the ?rm’s performance, a link between the
company’s key-performance indicators and social media parameters is required. In social business intelligence,
such links are required.
The second sub question was formulated as follows:
2. In which contexts are ?rms able to acquire social media data for business intelligence?
In this research, the context of a ?rm has been described based on two generic dimensions. Firstly, ?rms
were distinguished from each other based on the industry in which they operate. Secondly, ?rms’ contexts
were described by distinguishing di?erent customer relations types; i.e. B2B or B2C relations. The volume of
messages that contain the name of a ?rm di?ers from ?rm to ?rm. E.g. in our sample 39.425 messages related
to Heineken have been collected, while during the same period only 428 messages related to Fugro have been
found. Our analysis indicates that there exists variation in the volume of ?rm-related social media posts across
di?erent industries. Firms classi?ed as industrials, information & communication were more frequently subject
of discussion on social media than consulting or mining & quarrying ?rms. Our analysis also illustrates that
there exists variation in the volume of social media posts across B2B and B2C ?rms. B2C ?rms are far more
often subject of discussion on social media than B2B ?rms.
Apart from an assessment of the volume of social media content related to ?rms, we analysed the subjects of the
messages in order to gain an understanding of the type of information contained in the social media messages.
Our analysis shows that the subjects of social media messages di?er from ?rm to ?rm. The majority of social
media messages related to ?rms (41%) express how the external stakeholders of a ?rm perceive the company. In
this thesis, such posts have been classi?ed as community posts. 18% of the social media messages in our dataset
contained the name of a ?rm, but did not contain any valuable information for the ?rm and have consequently
been assigned as unde?ned posts. About 11% of the social media messages relate to ?nancial results, which
consist of ?nancial performance discussions (5%) and stock related discussions (6%).
The content analysis of this research suggests that the subjects of social media messages related to B2B
?rms contain a higher percentage of short term ?nancial results, news and professionals related messages than
messages related to B2C ?rms. Unfortunately for B2B ?rms, such type of information is yet available internally.
Acquiring social media data to gain additional management information is therefore of less value for B2B ?rms.
Next, the analysis indicates that the social media messages related to B2C ?rms contain a higher percentage
of posts related to customer relations, product and service quality and product and service innovation than
messages related to B2B ?rms. It are these types of information that deliver additional value to the ?rm, since
this information is not available at ?rms internally.
In addition, the content analysis of this research suggests that the subjects of social media posts di?er between
industries, but that the majority of the subjects in each industry relates to community, i.e. social media posts
92 Conclusions & Discussion
revealing how the community perceives the company. The results indicate that ?rms active in the information
& communication, ?nancial institutions and transport & storage industries are more subjected to social media
messages related to customer relations, while ?rms active in the mining and quarrying and consulting industries
will ?nd messages related to ?nancial performance.
As indicated in chapter 1, this thesis describes a ?rm’s context based on two dimensions; customer relation type
and industry type. By distinguishing ?rms based on customer relation type, we can state that B2C ?rms are
able to acquire (i) a high volume of social media messages related to their ?rm and (ii) social media messages
that contain information that is not yet available in the internal information systems of the ?rm, hence enriching
the business intelligence. Next, when distinguishing ?rms based on their industry we conclude that the volume
of ?rm related messages di?ers between ?rms. Additionally, the analyses suggest that the subjects of the social
media messages di?er from industry to industry. However, in our sample, there exists interaction between
the customer relation type and the industries. The di?erences in volume and subjects are more visible when
distinguishing between B2B and B2C ?rms rather than distinguishing between industries. Table 6-1 summarises
these conclusions.
Table 6-1: Conclusion of Content Analysis
Volume of Social Media Messages Subjects of Social Media Messages
Customer
Relation
B2C ?rms are more often subject of
discussion on social media than B2B
?rms.
B2C ?rms related social media messages are more
often subjected to customer relations, product and
service quality and product and service innovation
than B2B related ?rms. On the other hand, B2B
?rms’ related messages are more often subjected to
?nancial results, news and professionals discussing
the ?rm. However, the information contained in the
social media posts of B2B ?rms is often yet available
to the ?rm, hence not o?ering added value to the
?rm’s richness of management information.
Industry Our analysis shows a variation in
the volume of social media messages
across di?erent industries.
Our analysis indicates that there is a di?erence in
the subjects of social media posts related to ?rms in
di?erent industries.
The third sub question of this thesis has been formulated as:
3. Which processes are required to incorporate social media data into general business intelligence
frameworks?
‘Social’ data di?ers from internally generated and collected data on the following aspects. Firstly, social data
has not been veri?ed before it is published. Anyone can create a social media message, it will not be veri?ed
before it is available to the world. Social media messages may contain jokes or cynicism, making it hard for
?rms to interpret what the writer of the message actually means. Secondly, social media data is unstructured.
The data are sourced from multiple sources in di?erent formats and languages. Each source may employ its
own structure of social media messages, and not each platform may contain the same richness in attributes as
the other. Thirdly, the unstructured nature of the data and the huge amount of data that is generated makes
it that social media data can be labelled as ‘big data’, implying that the problems of big data may also be
applicable on social media data. Fourth, as with many data on the web social media messages contain spam,
which is to be removed before commencing an analysis of the messages. Fifth, while internal data may represent
evident relations (e.g. between the number of employees and the ?rm’s revenues), relations between social media
metrics and key-performance indicators are less evident. When a ?rm intends to add a social component to its
existing business intelligence, these aspects should be considered.
Structuring social media data is an important activity required to make an analysis on such data. Firstly, the
sourced data has to adopt one and the same data structure. Whereas each social media platform will deliver data
to its own favour – e.g. by CSV, JSON, XML, or other formats – the data should be parsed into one common
data format. Secondly, dividing the messages into categories makes the dataset ready for interpretation. Many
types of categories are possible. A classi?cation based on the people reveals which users are actively engaged
with the ?rm, which users have much power (in terms of followers and friends), which users speak positive
6-2 Contributions to Research 93
about the ?rm, etc. A classi?cation based on the subjects of the social media messages reveals which topics are
considered important by the social media users, and, more important a classi?cation based on subjects allows
a ?rm to link social media messages to the ?rms’ key-performance indicators. For instance, public image posts
– which may be additionally classi?ed as positive, neutral or negative – are related to a customer satisfaction
KPI. Though there are plenty of other categories to classify social media messages, to link the ?rm’s KPIs to
social media data the messages should be categorised based on their subjects.
Reaction based on
social intelligence
Strategic
mapping of
KPIs
Reacting
Search terms
Action plan(s) to
respond to gained
intelligence
g
Collecting
Mapping
insights to
business
units
Data pre
Unstructured
data
Information for
business units
Data pre-
processing
Categorising
Analysing
Structured, combined
(and anonymised) data
Categorised data
Figure 6-1: Blueprint: Social Business Intelligence Procedure
A social business intelligence (“SBI”) procedure prescribes how a ?rm should collect and process social media
data to gain intelligence, at which the ?rm can consequently base their decision-making. A SBI procedure
consists of seven main components, being (i) strategic mapping of KPIs, (ii) collecting, (iii) data pre-processing,
(iv) categorising, (v) analysing, (vi) mapping insights to the business units, and (vii) reacting. The seven steps
can be interpreted as a cycle, i.e. the output of the last step in?uences the ?rst step. Figure 6-1 schematically
shows the social business intelligence procedure.
At the start of this thesis, the objective has been formulated as:
The objective of this research is to develop a procedure to utilise social media data for business
intelligence, for which the applicability is investigated for ?rms in di?erent industries and for
di?erent relations with end-users.
Taking into account the answers on the research questions, we state that the objective of this thesis has
been achieved. A social business procedure has been developed, veri?ed on consistency with general business
intelligence processes and tailored to the challenges arising from processing social media data. In addition, the
applicability of this procedure is investigated. The results of our study indicate that especially ?rms performing
B2C relations are able to execute social business intelligence, because (i) these ?rms are subject of discussion
on social media, hence ?rm-related social media exists for these ?rms and (ii) the information contained in
B2C related messages o?er additional information for the ?rm. Furthermore, the results of this study also
indicate that there exists a di?erence in the volume of ?rm-related content between di?erent industries, in
which industrials and information & communication ?rms are more frequent subject of discussion on social
media than consulting and mining & quarrying ?rms.
6-2 Contributions to Research
Social media is a hot topic in the academic world. Existing research in the ?eld of social media is often aimed at
marketing e?orts or other activities in which the ?rm expresses or should express itself to the outside world. This
94 Conclusions & Discussion
thesis focused on the incoming information from a ?rm point of view, i.e. the extraction of information from
social media to support decision-making. The fact that this thesis investigated the applicability of social media
data for organisational decision-making, makes it that this thesis touches the world of business intelligence. In
our opinion, this aspect distinguishes this thesis from other research.
Future research aimed at deriving information – in whatever form – from social media for business purposes,
should be aware that the type of ?rm a?ects the applicability of such activities, and that one should not draw
generic conclusions applicable to all ?rms. As this research indicates, the existence of and subjects contained
in ?rm-related social media messages di?ers between ?rms. Furthermore, the method of this research is partly
based on traditional content analysis, in which a new type of data has been analysed. In the following section,
our experiences of applying a content analysing on social media data are shared.
6-2-1 Methodological Innovation
The content analysis methodology is not new. Krippendor? (2004) refers to propaganda analysis during World
War I as one of the earlies structured approaches of analysing texts. However, the fact that this thesis performed
a content analysis on social media messages makes it that part of the research in this thesis is innovative. The
structured approach that was executed was based on earlier work from Bos and Tarnai (1999), whom created
a framework to execute a content analysis. Though their framework did not speak of social media platforms
or data, we found that their research framework – with adjustments – is also applicable on social media data.
The adjustments of – or additions to – the framework relate to the data collection and preparation steps of
the content analysis procedure. Whereas textual data is considered structured in the framework of Bos and
Tarnai (1999), a content analysis on social media data requires a data structuring process before commencing
the categorisation process. The strength of the framework lies in the fact that it is generic, hence applicable
in many domains. With the experience of the execution of the content analysis on social media data in this
thesis, we recommend using the framework of Bos and Tarnai (1999) for future social media content analyses.
Figure 6-2 shows the additional steps required when applying a content analysis on social media. These steps
are highlighted in blue, and are established based on our experiences in the execution of the content analysis
on social media data.
Research outline, research questions,
formulation of hypotheses, material to
investigate
Operationalising the categories,
determining the sample, determining the
unit of analysis
Establishment of categories
Theoretical level
Determining reliability and validating the
categories
Appropriate statistical analyses
Data collection and evaluation
Immanent interpretation of the results,
discussion of the results on the basis of
the problem
Interpretation of the results
Select platforms, select keywords,
determine measurement period, determine
attributes to scrape
Social media domain
Create uniformdatabase, create uniform
data structure, merge data from multiple
platforms, remove spam
Pretest
Data preparation
Figure 6-2: Adjusted (Social Media) Content Analysis Procedure
The additional steps required when applying a content analysis take place after the second step. After the
research questions, the formulation of hypotheses, the determination of the categories that are to be analysed
and the sample establishment, the social media domain comes in. In this step, the researcher needs to determine
from which platforms the data is to be sourced. Each platform adheres to its own data format, and not each
platform provides access to all posts. Furthermore, each platform has its own focus, implying that di?erent
people are active on di?erent platform. Next, the keywords are to be determined. Comparable to search engines,
social media scrapers scan for keywords in the many posts created on the web. A researcher may decide to
6-3 Implications for Practice 95
base its keywords on user names, hashtags (subject of message, assigned by the creator of the message), the
researcher may decide to scrape all posts in a certain area, or during a time. Other elements to select the
messages that are to be analysed are also possible. Next, the attributes to scrape are to be determined. Each
social media post exists of various attributes, e.g. user name, time, location, content, hyperlink, etc. It di?ers
per platform which attributes are shared.
Next, the data is to be prepared. In case that the research exists of data sourced from multiple platforms,
the data is to be merged and structured into a uniform database. In addition, the data is likely to contain
spam. These messages are to be removed before commencing the analysis. The framework then proceeds in the
original steps of Bos and Tarnai (1999).
6-3 Implications for Practice
The ?ndings in this research have implications for (consulting) ?rms willing to use social media data for business
intelligence. First, the ?ndings of this research indicate that B2B ?rms are less likely to ?nd social media data.
Furthermore, if a B2B ?rm will ?nd messages related to the ?rm, these messages are likely to contain information
that is yet available to the ?rm. Hence, the applicability of social business intelligence for B2B ?rms is limited.
On the contrary, B2C ?rms are often subject of discussion, and the messages related to B2C ?rms contain
information that is not yet available to the ?rm internally. Taken into account the results of this research, the
opportunities and promises of social media found in many reports and white papers are mainly applicable to
B2C ?rms. Therefore, ?rms and ?rms o?ering consultancy on the domain of social business intelligence, should
be aware that the opportunities of social business intelligence are limited to B2C ?rms.
In addition, a stepwise procedure for social business intelligence has been developed. Such a procedure was
necessary to be developed since the new data source – social media platforms – di?ers from the systems at
which normally data is collected and stored. This procedure is applicable on ?rms for which social media data
is available.
6-4 Re?ection
In this section, a re?ection on the research process is presented. First, the developed Twitter scraper is discussed.
Next, additional research steps that were executed whenever there was more time available are presented.
Finally, a detailed stepwise approach of our data collection process is presented.
6-4-1 Twitter Scraper
During the early stages of this Master thesis project, a software tool has been developed that scrapes messages
created on Twitter. The tool has been written in PHP language and is designed to work with MySQL databases,
which we managed using phpMyAdmin. At the same time that the tool was up and running, access to one of
the commercial social media monitoring tools (uberVU) was granted to the author of this thesis, for which we
are grateful. Clients of uberVU pay a monthly fee of at least $1.000 to access the software. The tool allowed us
to scan a variety of social media platforms, whereas our own tool solely scraped Twitter. Therefore, the decision
was made to use the commercial o?-the-shelf software rather than our own to collect the data. In addition,
uberVU o?ers additional features like sentiment analysis, in?uencer ranking, location of the message, etc. that
are not available in our tool. However, our own scraper – though it solely scrapes tweets – may serve research
projects that are aimed at tweets.
6-4-2 If I had More Time
This research has been executed during a period of six months, i.e. from July 2012 to December 2012. The
limited time available for this research has implications for both the depth and the breadth of the research, hence
on the conclusions and the applicability of the conclusions. In case that the research time would be longer, we
would have surveyed more ?rms so that the analysed categories (B2B/B2C and industries) would have consisted
of more respondents, allowing for statistical analyses. Whereas the conclusions of this research are exploratory,
the statistical testing of di?erences between groups of ?rms would allow for the generalisation of the statements.
Furthermore, adding more respondents to the sample leads to more industries being represented in the sample,
96 Conclusions & Discussion
so that the conclusions of the research are also applicable on other industries. Next, we would have analysed
more messages in the content analysis. The social media messages in this research have been collected during
a period of two weeks. In case the messages would have been collected during a longer period of time, e.g. six
months, the sample would have existed of more messages. As such, it would be possible to gain insight in the
‘steady volume’ of daily messages, and, more interestingly, deviations in the steady volume. Deviations might
be due to the announcement of ?nancial ?gures, marketing events, etc. However, to perform a content analysis
on many social media messages, an automatic classi?er is required. Manually classifying the posts, as we did
in this research, would then take too much time. To train such an automatic classi?er, the manually classi?ed
posts in this research are suited. Furthermore, if we had more time, we would have investigated whether or not
other factors, such as the number of employees, revenues, market capitalisation, etc. also a?ect the number of
social media messages that are related to a ?rm. With respect to the social business intelligence procedure that
has been developed, we would have validated the framework by pilot projects and the involvement of business
intelligence experts.
6-4-3 Stepwise Description of Data Collection Process
In this section we share our method of the data collection process. These steps are also incorporated in the
adapted content analysis framework in section 6-2-1.
• Determine keywords / search terms
After establishing the sample, the search terms required to ?lter out the related social media messages
have been determined. In this thesis we searched for the ?rm names. However, it is also possible to search
more speci?c, e.g. on the name of a product or a speci?c event.
• Use search terms in social media monitoring tool
The search terms were consequently used in the social media monitoring tool.
• Export the search results into a database
This functionality is not o?ered by each social media monitoring tool, but vital for the data collection
process. Tools that do not o?er the exportation of search results in whatever format are not suited for
further analyses on the data because most analysis software requires the data to be stored on a local
machine.
• Structure the database
Depending on the output of the export process, the database is to be structured. The social media
monitoring tools used in this thesis exported the messages into comma-separated values, which could
easily be loaded into MS Excel. The richness of attributes o?ered by the social media monitoring tools
determines the complexity of the structuring of the database.
• Daily search for new results
In order to collect a large dataset, daily runs for new messages were executed. The social media monitoring
tools used in this thesis o?ered to possibility to export up to 10,000 messages per search run. Pre-testing the
collection process illustrated that this constraint was su?cient to get a complete picture of the ?rm-related
messages created on the social media platforms by daily searching for new results. Whenever a tool has
been used that o?ered a lower exporting capacity, e.g. 1,000 messages, the frequency of searching for new
results would have been higher.
• Verify that the new search results do not yet exist in the database
A daily run for ?rm-related messages resulted in the collection of duplicates, i.e. messages that were
already collected yesterday. These messages have been identi?ed based on their unique URL that was
contained as an attribute to each message using LOOKUP functions in MS Excel. More speci?cally, the
value of each URL was LOOKED UP in the existing spreadsheet. The textual format of these lookup
values required some computer power, but our 4GB RAM / i5 machine turned out to have su?cient power
for these calculations. Whenever a message did found a match, this meant that the message did yet exist.
The messages that did yet exist were not added to the database.
• Start analyses
The data collection process resulted in a structured database in a format so that MS Excel could handle
the data for analyses.
Whereas the above steps are described in detail, the underlying ideas will be applicable on each social media
data collection process.
6-5 Limitations 97
6-5 Limitations
This research and consequently its outcomes have limitations that should be taken into consideration when
adopting the conclusions of this thesis. The limitation are discussed in this section.
The ?rst aspect of the research limitations, or aspects to consider when interpreting the conclusions related to
the population from which the social media messages are drawn. This aspect has been assigned by Krippendor?
(2004) as an important issue to consider when performing a content analysis. Not everyone uses social media,
and even less people actually create content on the platforms. Therefore, ?rms that analyse social media data
should be aware that these data do not represent the full (potential) client base. It is very likely that social
media users have other preferences than non-social media users. A ?rm should always place the conclusions
from social business intelligence in the light of their complete client base before it makes a decision to undertake
actions, because the actions may only serve those needs of the ones that engaged on social media. Nevertheless,
anno 2012 social media is relatively young. The user groups – e.g. age groups or countries – that use social
media may increase in the coming years. We expect that social media will be further embedded in the lives of
people that grow up in the social media era.
Secondly, each social media platform has its own privacy policy. This implies that a user either has the possibility
to determine whether or not it shares its messages to the public, or that the platform determines the publicly
available messages. As a consequence, only messages that were publicly available have been analysed in our
dataset. It is likely that people who have not publicised their social media messages also discuss ?rms or ?rms’
products / services, these messages are not available in our dataset. Still, the dataset is representative for ?rms
conducting social business intelligence, since they will not get access to private messages either.
Third, not all social media platforms have been part of our analysis. Platforms such as Sina Weibo – the Chinese
counterpart of Twitter –, Qzone, Renren, Habbo are not part of our analysis. However, the platforms that did
exist in our sample are the ones used in the Western world. Therefore we state that the conclusions of our
report are valid for Western world ?rms.
Fourth, the volume of messages, likes, shares, retweets, etc. can easily be in?uenced by a ?rm, though this does
not necessarily mean that the user is actually engaged with the ?rm. For instance, a ?rm may decide to ra?e an
iPad or organise other lotteries. People can participate in the lottery by e.g. sharing a promotion message or by
‘liking’ the ?rm’s page. Whereas such activities certainly lead to an increase in the number of likes, shares, etc.,
the underlying reason why people pay attention to the ?rm is for the price, and not necessarily the engagement
in the company or its products. We refer to content that is generated according to such mechanisms as biased
chatter, and doubt if such activities actually lead to an improvement of the ?rm’s KPIs, e.g. the number of
sales.
Fifth, one of the conclusions of this research is that the volume and the subjects of social media messages di?er
among industries. Due to time restrictions, we have not been able to analyse the social media messages of more
than eighteen ?rms. As a result, each industry consisted of two or three ?rms, which we deem a small sample.
It is therefore that the conclusions of this research are to be perceived as exploratory rather than con?rming
hypotheses.
Sixth, our research grouped ?rms based on the industry type in which they are active. With respect to the
volume of ?rm-related messages, intra group di?erence have been spotted. These notable ?ndings indicate that
the industry aspect is not the only determining factor in?uencing the volume of ?rm-related messages. Other
factors, such as (world-wide) brand awareness or the size of the company are likely to in?uence the amount
of ?rm-related messages that are daily generated. These company speci?c aspects have deliberately not been
taken into account since the purpose of this thesis is to draw generic conclusions on the applicability of social
business intelligence.
Next, this thesis analysed the applicability of social business intelligence on two dimensions; industry type and
customer relation type. We did not correct for interaction e?ects between these two groups. It is e.g. likely
that there exist more B2B ?rms in the consulting industry. However, the results of this study are exploratory
and – from the insights we gained – future research containing larger samples should correct for such interaction
e?ects.
Next, this thesis mainly focused on the business perspective of social business intelligence. Less attention has
been paid to the technical perspective. For example, the question “What kind of database is best to store
the unstructured data that is captured in text form?” is unanswered. Though the developed social business
procedure prescribes which components are required to collect and analyse social media data in relation with
the ?rm’s performance, the technical requirements related to these components are underexposed.
98 Conclusions & Discussion
Finally, the social business intelligence procedure that has been developed in this thesis is compliant with general
business intelligence concepts that are adhered to in ?rms. However, the procedure has – due to time restrictions
– not been validated, that is, tested on a real case. Nevertheless, the individual components of the procedure
are tested. The general BI steps are yet used by ?rms, and the collection and classi?cation processes of social
media messages have been performed in this thesis.
6-6 Future Research
During the execution of this research, ideas for future research related to this thesis have been devised. These
ideas are presented in this section. The suggested researches build further on the conclusions of this thesis.
6-6-1 Classi?er
We have manually classi?ed social media posts in categories. With these manually classi?ed posts, it is possible
to create an automatic classi?cation process. In automatic classifying, a classi?er will be “trained” so that
it recognises which words and phrases relate to a certain category. The classi?ed messages in this research
can serve as the training set for an automatic classi?er. As we have experienced, and which is also argued
by Gianfortoni, Adamson, and Rosé (2011), classi?cation of social media posts, e.g. by gender, age, political
a?liation and sentiment analysis is di?cult, and even more problems arise when models trained in one domain
are applied in another domain. Therefore, a social media classi?er should not be used generally on each domain.
We even argue that each ?rm requires its own classi?er, only because the product names of ?rms di?er.
In the development of the social business intelligence procedure we argued that the categorising process is one
of the important steps in structurally analysing social media messages. This process is even more challenged
by the increase of user-generated social media content showing big data characteristics. From a social business
intelligence view, it is desired that research in the ?eld of automatic text classifying – tailored to ?rms – proceeds.
6-6-2 Social Media Posts Categories
A part of this research required the establishment of social media posts categories. Whereas the starting point of
the establishment of the categories in this thesis was based on former research, the social media messages in the
dataset have driven the establishment of additional categories. Future research in the domain of social media,
and more speci?cally the classi?cation of social media messages can use the categories that were established in
this thesis.
6-6-3 The Real Source
Many messages are forwarded – retweetet and shared – from users to others. Thereby, messages do not stay
within one social media platform. It is interesting to investigate which platforms contains the most initial
creations of information. As such, ?rms can manage their reputation by actively following those platforms that
create the most initial messages, before the message goes viral and may harm the ?rm’s reputation.
6-6-4 Case Study: Relations of Social Media Metrics and Key-Performance Indicators
The social business intelligence procedure that has been developed in this thesis contains a component in which
social media messages are assigned to key-performance indicators based on the subject of the messages. A
research in which the relations between the social media messages and the actual values of various KPIs of a
?rm are investigated would reveal the strength of these relations. Thereby, the KPIs that have been assigned
in this thesis as being able to be measured using social media data could be used for such an analysis.
Appendix A
Performance Prism Perspectives and
Key-Performance Indicators Categories
Table A-1 on the next page assigns each KPI category de?ned by Ittner et al. (2003) to a performance prism
perspective de?ned by Neely et al. (2001). Based on their subjects, social media posts will be assigned to KPI
categories in this thesis. With the assignment of KPI categories to performance prism perspectives, we can
derive conclusions of the existence of social media data related to di?erent performance prism perspectives.
100 Performance Prism Perspectives and Key-Performance Indicators Categories
Table A-1: Assigning Key-Performance Indicator Categories to Performance Prism Perspectives
KPI Category Performance Prism
Perspective
Elucidation
Short-term ?nancial results Stakeholder satisfaction Shareholders are the actors that are interested in the
return on their investment.
Customer relations Stakeholder satisfaction
/ contribution
On the one hand, KPIs related to customer relations
can represent the customer satisfaction. On the
other hand, customers may also contribute to the
?rm, e.g. by payments and/or co-creation activities.
Employee relations Stakeholder satisfaction
/ contribution
Employees are the actors interested in getting
awarded for their contributing value to the ?rm.
Therefore, KPIs related to customer relations can
involve both perspectives.
Operational performance Processes KPIs related to operational performance re?ect the
performance of business processes, generally in time
of volume, speed, reliability, etc.
Product & service quality Capabilities KPIs related to product and service quality represent
how capable a ?rm is in performing its activities.
Alliances Stakeholder satisfaction
/ contribution
Metrics related to the ?rm’s alliances are on the
one hand purposed to satisfy the participating
parties and on the other purposed to measure the
contributing value of the alliance to the ?rm.
Supplier relations Stakeholder satisfaction
/ contribution
KPIs related to supplier relations are purposed to
measure the contributing value of the suppliers’
products/services to the ?rm, or the metrics can
specify the satisfaction of the customers (e.g. in
terms of the price paid for the product).
Environmental performance Stakeholder satisfaction A ?rm’s activities may a?ect the environment.
Groups representing the environment may not
be satis?ed whenever the ?rm’s activities a?ect
the environment. KPIs re?ecting environmental
performance hence relate to stakeholder satisfaction.
Product & service innovation Processes One of the key business process relates to the
development of new products and services.
Community Stakeholder satisfaction
/ contribution
Metrics related to the ?rm’s community relate to the
public image of the company. As such, these metrics
involve stakeholders.
Appendix B
Classi?cation of Social Media Posts
In section 4-3-1, categories for social media posts have been established. Consequently, the collected social media
posts of the ?rms in the sample have been classi?ed into one of these categories. The results are discussed in
this chapter. The results will be discussed from ?rm to ?rm. However, we will refer to ?gure B-1 – which is
depicted below – when discussion the individual ?rms. This ?gure contains a heat-map of the KPI categories
per ?rm, indicating which category is presented the most (green) and which the least (red) in the sample. The
percentages that are contained in the heat-map are visualised in a stacked bar chart in ?gure B-2.
KPI Category
A
B
N
A
M
R
O
A
e
g
o
n
A
k
z
o
N
o
b
e
l
A
l
b
e
r
t
H
e
i
j
n
A
r
c
a
d
i
s
A
r
c
e
l
o
r
M
i
t
t
a
l
B
l
o
k
k
e
r
B
o
l
.
c
o
m
C
-
1
0
0
0
C
o
c
a
-
C
o
l
a
F
u
g
r
o
H
e
i
n
e
k
e
n
K
L
M
N
S
P
h
i
l
i
p
s
P
o
s
t
N
L
T
o
m
T
o
m
U
n
i
b
a
i
l
-
R
o
d
a
m
c
o
A
v
e
r
a
g
e
1. Short-term financial results 1% 5% 21% 0% 37% 16% 0% 0% 0% 0% 36% 3% 1% 0% 1% 2% 3% 75% 11%
1.1 Financial performance discussions 0% 0% 1% 0% 32% 8% 0% 0% 0% 0% 0% 0% 1% 0% 1% 0% 0% 47% 5%
1.2 Stock related discussions 1% 5% 20% 0% 5% 8% 0% 0% 0% 0% 36% 3% 0% 0% 0% 2% 3% 28% 6%
2. Customer relations 17% 4% 0% 20% 2% 0% 7% 17% 1% 4% 0% 2% 14% 33% 0% 39% 6% 0% 9%
2.1 Explaining firm 4% 0% 0% 5% 0% 0% 0% 3% 0% 1% 0% 1% 4% 4% 0% 2% 2% 0% 1%
2.2 Understanding firm 2% 0% 0% 4% 0% 0% 0% 4% 0% 1% 0% 0% 7% 4% 0% 1% 2% 0% 1%
2.3 Thanking firm 1% 0% 0% 1% 1% 0% 0% 1% 0% 1% 0% 0% 1% 1% 0% 0% 1% 0% 0%
2.4 Informing firm 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 11% 0% 0% 0% 0% 1%
2.5 Questioning customer 3% 1% 0% 3% 0% 0% 5% 4% 0% 1% 0% 1% 1% 4% 0% 5% 1% 0% 2%
2.6 Complaining customer 4% 2% 0% 6% 0% 0% 3% 4% 1% 0% 0% 1% 1% 9% 0% 30% 0% 0% 3%
2.7 Thanking customer 2% 1% 0% 2% 0% 0% 0% 1% 0% 0% 0% 0% 1% 1% 0% 1% 0% 0% 0%
3. Employee relations 1% 4% 10% 5% 5% 2% 8% 0% 8% 1% 10% 0% 0% 3% 2% 21% 0% 1% 4%
3.1 Recruitment 1% 2% 9% 1% 4% 2% 1% 0% 0% 1% 8% 0% 0% 3% 2% 7% 0% 1% 2%
3.2 Employee posts 1% 2% 1% 4% 1% 0% 7% 0% 8% 0% 1% 0% 0% 0% 0% 14% 0% 0% 2%
4. Operational performance 1% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 1% 0% 0% 0% 0% 0%
5. Product and service quality 1% 0% 0% 1% 0% 0% 1% 0% 0% 2% 0% 1% 0% 1% 7% 1% 7% 0% 1%
6. Alliances 0% 2% 0% 0% 0% 0% 0% 1% 0% 0% 0% 0% 0% 0% 0% 3% 0% 0% 0%
7. Supplier relations 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 1% 0% 0% 0% 0% 0% 0% 0%
8. Environmental performance 0% 0% 0% 2% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
9 Product and service innovation 2% 4% 1% 1% 0% 0% 0% 0% 0% 1% 0% 1% 0% 1% 1% 0% 8% 0% 1%
10 Community 71% 43% 55% 28% 44% 60% 27% 72% 25% 57% 49% 44% 7% 30% 61% 23% 26% 15% 41%
10.1 Promotion 56% 2% 1% 1% 8% 2% 0% 20% 2% 2% 0% 3% 1% 2% 1% 0% 2% 1% 6%
10.2 News 3% 4% 21% 2% 5% 6% 0% 0% 0% 1% 1% 4% 0% 1% 0% 1% 0% 1% 3%
10.3 Public image 10% 36% 11% 26% 18% 32% 26% 50% 23% 50% 3% 33% 4% 27% 6% 20% 22% 3% 22%
10.4 Professionals 3% 1% 3% 0% 12% 20% 0% 2% 0% 2% 45% 3% 2% 0% 0% 2% 0% 11% 6%
10.5 Distributors 0% 0% 18% 0% 0% 0% 1% 0% 0% 1% 0% 2% 0% 0% 54% 0% 2% 0% 4%
Undefined 4% 16% 10% 43% 11% 18% 37% 4% 50% 23% 4% 28% 2% 2% 1% 11% 49% 9% 18%
Spam 0% 22% 2% 0% 1% 3% 19% 6% 15% 13% 1% 18% 75% 28% 27% 0% 2% 0% 13%
Classified Posts 7.067 1.449 922 2.848 455 1.097 1.788 2.574 1.651 1.078 428 2.050 2.498 1.441 1.623 1.013 1.100 512
Figure B-1: Social Media Posts Classi?cation
ABN AMRO
ABN AMRO Group N.V. is a Dutch bank with 6.8 million clients and around 25.000 employees. The ?rm
organises multiple marketing events each year, of which the pictures were uploaded to the ?rm’s Picasa pro?le
during our sample period. This declares the high percentage of social media posts made on this platform, as
can be seen in table C-1. If we would neglect the Picasa posts, which are somehow irregular posts and moreover
made by the ?rm on its own, we would ?nd that 90% of the social media posts have been sourced from Twitter,
which is in line with the other ?rms.
102 Classi?cation of Social Media Posts
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Unibail-Rodamco Arcadis Fugro Akzo Nobel ArcelorMittal Aegon PostNL NS Albert Heijn ABN AMRO Bol.com KLM Blokker TomTom Coca-Cola Heineken C-1000 Philips
P
e
r
c
e
n
t
a
g
e
o
f
S
o
c
ia
l
M
e
d
ia
P
o
s
t
s
Short-termfinancial results Customer relations Employee relations Operational performance Product and service quality Alliances Supplier relations Environmental performance Product and service innovation Community Undefined Spam Short-termfinancial results Customer relations Employee relations Operational performance Product and service quality Alliances Supplier relations Environmental performance Product and service innovation Community Undefined Spam
Figure B-2: Social Media Posts Classi?cation
We have classi?ed all 7.067 collected social media posts in the ABN AMRO data set. The results are shown in
?gure B-1. The columns show the social media post categories as a percentage of the total classi?ed posts of
that ?rm. For ABN AMRO, we can conclude that the majority of the social media posts (71%) are related to
the community category, which is due to the large share of promotion type of posts. As explained, this ?gure
is high because of the Picasa posts made by the ?rm in the sample period. Secondly, we see that many posts
in the ABN AMRO data set are related to customer relations. ABN AMRO operates a web care team that
actively monitors the social media sites for customers that complain or ask questions. Here, we see that the
web care team of ABN AMRO partly replaces the traditional telephone help desk. The table also shows that
a substantial part of the posts are not able to be classi?ed into one of the categories. These posts particularly
relate to people using ABN AMRO o?ces as a point of recognition to meet or illustrate where they are. For
example, posts contain appointments like “let’s meet in front of the ABN AMRO o?ce before we go into town”.
Whereas these messages de?nitely contain the name of the ?rm, they do not contain any relevant information
for the ?rm that can be marked as “social intelligence”.
Aegon
Aegon N.V. provides insurances, pensions and asset management products to over 47 million clients in the
world. As can be concluded from table C-1, the majority (81%) of the collected posts have Twitter as a source.
The channel distribution of Aegon is in line with the distribution of other ?rms.
Over the sample period, 1.449 posts have been collected containing the word “Aegon”. All these posts have been
classi?ed in the categories that were established in section 4-3-1. Figure B-1 shows the results of the classi?cation
of Aegon’s social media posts. As can be concluded, the majority (43%) of the social media posts are related
to the KPI category involving community related indicators. More speci?cally, the majority of the community
related posts are classi?ed as public image posts. Public image posts contain the ?rm’s name, indicating that
people are talking about the ?rm, but the posts are not purposed to get in contact with the ?rm. When
analysing the public image posts more detailed, we ?nd that many of these posts relate to Aegon’s sponsoring
activities. The ?rm is for example sponsor of the Dutch soccer club Ajax, it sponsors a tennis centre called
Aegon Arena and it sponsors the Dutch rowing team. The public image social media posts related to sponsoring
activities may serve as a measure to determine the exposure of the sponsoring activities. A substantial part of
Aegon’s social media messages have been classi?ed as spam. Spam messages contain the name of the ?rm but
do not relate the ?rm. The dataset of Aegon revealed that a person named “Aegon The Conqueror” showed up
often in posts. Aegon The Conqueror is a character in a popular TV series called Game of Thrones. These kind
of spam messages, that show up in the dataset because the name of the ?rm is commonly used for the naming
of other entities or people, do not contain any information that may be valuable for the company.
103
Akzo Nobel
Akzo Nobel N.V. is a Dutch ?rm active in paint, lacquer, coatings and other specialised chemical products. The
company operates in 80 countries and employs around 55.000 people. As can be concluded from table C-1, the
922 posts that have been collected are for the majority sourced from Twitter.
Figure B-1 shows the results of the classi?cation process of the 922 social media messages that have been
collected. As can be concluded, around 21% of the social media messages that are related to Akzo Nobel refer
to news articles. Around 18% of the social media posts are made by distributors of Akzo Nobel’s products
who promote their products on retail websites like Amazon. Another substantial part – 20% – of the messages
relates to stock related discussions. A remarkable phenomenon in the classi?cation of Akzo Nobel’s social media
messages is the fact that there are almost no messages related to customer relations, whereas we have seen
customer related messages in the social media posts of other companies.
Albert Heijn
Albert Heijn is a Dutch supermarket, which is a subsidiary of Royal Ahold N.V. Albert Heijn operates around
850 stores in the Netherlands, and is with 34% market share the market leader in the Netherlands. The company
also operates stores in Belgium, Germany and Curacao. Table C-1 illustrates that almost all of the collected
social media messages containing Albert Heijn have been sourced from Twitter.
As ?gure B-1 shows, a substantial part (20%) of the social media messages of Albert Heijn involve customer
relations management. On the one hand customers are responsible for many complaints (6% of the classi?ed
posts), while Albert Heijn’s web care team actively responses to these messages by either showing an
understanding (4% of the classi?ed posts) of the customer’s complaint, or even explaining (5% of the classi?ed
posts) the customer something that they asked. Here, we clearly see that the ?rm’s help desk moves to social
media. Another substantial part (26%) of social media messages involving Albert Heijn relate to the ?rm’s
public image. These messages contain customer’s opinions about e.g. the latest Albert Heijn commercial or
about the products that they bought in the store.
Arcadis
Arcadis N.V. is a Dutch engineering consultancy, o?ering solutions in the ?eld of infrastructure, civilised areas
and environmental projects. The company is active in more than 70 countries and employing around 18.000
people. During the period of monitoring the companies, we collected 455 posts related to Arcadis. This ?gure
is substantially lower than the number of posts that we collected from other ?rms. Of the 455 social media
posts that have been collected and are related to Arcadis, around 93% has been derived from Twitter.
Shown by ?gure B-1 on page 101, 32% of the collected messages involve Arcadis’ ?nancial performance
discussions. These posts particularly involve professionals discussing the ?nancial performance of the company
and talking about future projections of the company’s ?nancial position. Next, 18% of the Arcadis’ posts have
been classi?ed as public image posts. These posts particularly refer to publicised articles or news messages that
have been spread by the company. Presumably, the company has spread these articles / news messages to gain
exposure. The ?rm could use the amount of public image messages that refer to the articles as a measure to
determine the exposure as a result of the publicised articles.
ArcelorMittal
ArcelorMittal is the world’s largest steel producer, active in 27 countries and employing 320.000 people. During
the period of monitoring, we collected 5.532 social media posts related to ArcelorMittal. Of these posts, around
83% have been derived from Twitter. Other sources involved blogs (5%), news sites (2%) and other platforms.
As can be seen in ?gure B-1, social media messages related to ArcelorMittal do not involve customer relations.
Rather, the social media messages relate to the ?rm’s public image (32% of the posts) and professionals (20%
of the posts) talking about the company. The posts classi?ed as public image involve marketing activities of the
?rm. Especially the “ArcelorMittal Orbit”, a steel tower constructed by the ?rm on the 2012 London Olympics,
was subject of discussion. This category of social media messages is the only one that involves non-professional
104 Classi?cation of Social Media Posts
people, because all other social media messages related to ArcelorMittal involve professionals. 20% of the
sample posts have been classi?ed as professionals, in which professionals discuss joint-ventures or the industries’
outlook. Finally, a substantial amount of the posts (16%) relate to the companies ?nancial results.
Blokker
Blokker is a Dutch store selling products related to household. The Blokker stores are subsidiaries of Blokker
Holding B.V., which operates over 2.900 stores in 11 countries, thereby employing 25.000 people. In total, 2.769
social media messages related to Blokker have been collected, of which the majority (91%) has been derived
from Twitter. Please see table C-1 on page 108 for a distribution of the sources of the social media posts.
Figure B-1 (page 101) shows that 27% of the classi?ed social media posts are related to the community category.
The community category comprises messages that reveal the community’s perception of the ?rm. Of these 27%
community related posts, the vast majority consists of social media posts that have been classi?ed as public
image posts. Public image posts are messages that are made by individuals and contain the ?rm’s name, without
explicitly seeking contact with the ?rm. In the case of Blokker, many messages involve statements of people
announcing to their followers that they are planning to visit one of the stores, or people referring to articles that
are sold by the ?rm. As can be concluded, customer also ask questions to the ?rm and also complain about the
?rm. However, no messages of the company have been found in the sample that respond to these messages.
Bol.com
Bol.com is an online web-shop selling a variety of products, such as books, DVDs, games, blu-rays, electronics,
computers, etcetera. Since the foundation of the company in 1999, it has shown solely growth percentages of
18% y-o-y and above in terms of revenues. As of 2012, Bol.com is a subsidiary of Royal Ahold N.V. We have
collected 5.782 social media posts that are related to Bol.com, of which 89% has been derived from Twitter.
Of the classi?ed posts related to Bol.com, around 50% are related to the ?rm’s public image. For Bol.com, these
public image posts particularly involve people who share to their followers their recent purchase of a product
through Bol.com or people illustrating to other people that a certain product can be bought at Bol.com. Next,
a substantial amount of the posts consists of promotion activities. These posts are made by the ?rm or by
?rm’s selling products through Bol.com’s website for marketing purposes. Finally, 17% of the Bol.com classi?ed
social media posts relate to customer relations. We clearly see that people ask questions or complain, and that
Bol.com’s web care team is consequently responsible for the social media posts that have been classi?ed as
either explaining ?rm (3%) or understanding ?rm (4%).
C1000
C1000 is a Dutch supermarket organisation with a market share of 11,5%, employing around 7.000 people and
operating 425 stores in the Netherlands. In the future, many C1000 stores will be turned into Jumbo stores
as Jumbo Supermarkten acquired C1000 in 2012. Allmost all of the 5.782 social media posts that have been
collected in relation with C1000 have been derived from Twitter.
The majority of the social media posts (50%) of C1000 have been classi?ed as unde?ned, implying that these
messages cannot be classi?ed into one of the categories that have been established. When taking a closer look
at the unde?ned messages, we see that many users refer to C1000 as a location to meet each other, or users
share that they are heading for or just returned from C1000. These messages do not contain any management
information, and are therefore classi?ed as unde?ned. The other portion of the classi?ed social media messages
related to C1000 are classi?ed as public image posts. These messages contain statements of customers sharing
their followers what they have bought or seen at a C1000 store. Next, marketing campaigns are discussed by
people.
Coca-Cola
Coca-Cola is one of the many drinks o?ered by The Coca-Cola Company, selling Coca-Cola all over the world
(except for North-Korea and Cuba). As expected, Coca-Cola is one of the company’s in our sample that
105
delivered the most messages because it is one of the most famous brands of the world. During the monitoring
period, 32.953 messages containing the word Coca-Cola have been collected. In line with other companies in
our sample, 89% of the collected Coca-Cola social media messages have been derived from Twitter.
As ?gure B-1 illustrates, half of the classi?ed posts of Coca-Cola can be positioned under the category labelled
public image. These posts contain perceptions of customers to the company, to marketing campaigns of the ?rm
or are made by people talking about the ?rm’s sponsorships. Next, a substantial part of the classi?ed social
media posts of coca-cola are classi?ed as unde?ned. These messages to not contain any valuable management
information, and contain for instance statements of people that they are drinking Coca-Cola right now, or that
they wish that they were drinking one now.
Fugro
Fugro N.V. is a Dutch company that collects and interprets data related to the earth’s surface. The company
provides advice to ?rms active in the oil- and gas industry, the mining industry and the construction industry.
Fugro is active in over 50 countries, operating 275 o?ces and employing around 14.000 employees. During the
monitoring period, merely 428 social media posts related to Fugro have been collected. Of all ?rms in our
sample, there is no ?rm with less search results. Though the sample of Fugro is small, it shows a channel
distribution that is comparable to the other ?rms in our sample; around 90% of the derived post are sourced
from Twitter.
As indicated by ?gure B-1 on page 101, the vast majority (45%) of the social media posts related to Fugro
are classi?ed as posts made by professionals. These posts consists of professionals talking about new vessels
that Fugro either ordered or received, or how macro trends are e?ecting the market in which Fugro operates.
Furthermore, automated messages creating tools post a message each time that a Fugro vessel leaves or arrives
at a harbour. These posts are also classi?ed as professionals. Another substantial part of the social media posts
related to Fugro have been assigned to the category labelled stock related discussions. Unfortunately the two
categories that are responsible for the majority of the social media post categories do not o?er any information
to the company that is not available at the company internally.
Heineken
Heineken N.V. is a Dutch multinational providing beer and other drinks. The company is active in 178 countries,
employing 70.000 people. During our period of monitoring 39.425 social media posts have been collected by the
search terms related to Heineken, making Heineken the ?rm with the most mentions of our sample. As can be
concluded from table C-1, 82% of these social media posts have been sourced from Twitter, while Facebook is
responsible for 15% of the messages.
Of the 2.050 posts that have been classi?ed, 33% have been classi?ed as public image posts (see table B-1). Many
of these posts relate to Heineken commercials seen by people on the television, or other ways that Heineken
pursues to expose itself such as the Holland Heineken House at the 2012 London Olympics. The number of
public image posts can serve as a measure to determine the success of the desired exposure by these kind of
marketing events. Another substantial part of Heineken’s posts are considered as spam, because they refer to
other entities, people naming themselves Heineken on the web or people that are actually named Heineken.
KLM
KLM – Koninklijke Luchtvaart Maatschappij – N.V. is a Dutch airliner that operates 116 airplanes across the
globe. The ?rm has three subsidiaries – KLM Cityhopper, Martinair and Transavia.com – while the parent
company is Air France-KLM. The search terms used to ?lter out the social media posts related to KLM (see
table 4-5 on page 58) resulted in 26.364 messages which have been scraped. 86% of these messages have been
sourced from Twitter, while Facebook is responsible for 9% of the posts. These ?gures are in line with the
channel distribution of other ?rms in the sample.
10% of the collected posts – i.e. 2.498 posts – have been classi?ed into one of the social media categories that
have been established in section 4-3-1. Figure B-1 shows that an astonishing 75% of these posts have been
classi?ed as being spam. A closer look at the spam classi?ed posts shows that the letters K, L and M are
106 Classi?cation of Social Media Posts
used by many people in their username, e.g. @Klm_babe, @DaOne_KLM, @klm_klm_klm, @klm_nico, and
@miyu_klm. Probably, the initials of these people correspond with the name of the ?rm. However, the dataset
of KLM also shows a substantial amount of posts classi?ed as customer relations. KLM operates a webcare team
that actively monitors the messages directed to KLM, at which the employees of the webcare team consequently
respond. Again – as we have seen with ABN AMRO and Albert Heijn – we see that the traditional customer
help-desk is (partly) moving to the social media.
NS
Nederlandse Spoorwegen N.V. (“NS”) is a Dutch railway company operating the main rail network in the
Netherlands. During the monitoring period, 5.863 social media posts related to NS have been collected. 85%
of the NS posts have been sourced from Twitter, while Facebook is responsible for 12% of these posts.
Figure B-1 shows that 33% of the classi?ed posts are related to customer relations. Especially the category
informing ?rm is over-represented in the dataset, this is due to the ?rm that uses social media to inform
customers that certain tracks of the network are subject to delays. These posts do not contain any information
that is not available internally, because the nature of the direction of these messages is outgoing; from ?rm to
customers. Next, 9% of the classi?ed posts are complaining customers, while 4% of the posts are questioning
customers. These posts contain information that may not be available to the ?rm. The ?rm operates a web
care team that answers questions and shows understanding for the experienced problems (4% and 4% of the
posts respectively).
Philips
Koninklijke Philips Electronics N.V. is a Dutch electronics ?rm active in more than 60 countries and employing
122.000 people. The ?rm is organised into three main divisions: Philips Consumer Lifestyle, Philips Healthcare
and Philips Lighting. Philips is the largest manufacturer of lighting in the world. During the monitoring period
32.748 posts have been collected using the Philips search queries of table 4-5, corresponding to almost 3.000
daily posts. Philips is the second largest ?rm in our sample in terms of collected social media messages. 68%
of the messages have been sourced from Twitter, 12% from Facebook, 9% from Blogs and 10% from other
platforms including Friendfeed and YouTube.
As indicated by ?gure B-1, the vast majority (54%) of the social media messages related to Philips are classi?ed
as distributors posts. The distributors posts are made by professionals that are selling Philips products to
consumers. Often, distributors use Amazon.com as a site to sell the products, while they use social media to
announce the public their o?ers. Philips is a common surname. As a result, many posts in the Philips sample
have been classi?ed as spam as they do contain the name Philips, but do not relate to the ?rm. 7% of the
classi?ed posts are labelled as product and service quality posts. These posts comprise product- reviews and
experiences of users, containing valuable information for R&D related activities.
PostNL
PostNL N.V. is a mail and parcel company operating in the Netherlands, Germany, Italy and the United
Kingdom. In total, 1.323 social media messages have been collected using the search terms related to PostNL.
91% of these posts have been collected from Twitter, which is in line with other ?rms in the sample.
More than in any other dataset of the ?rms, 30% of PostNL’s social media messages have been classi?ed as
complaining customers. The messages contain statements of customers who are complaining about the service
of the ?rm, about broken parcels, late deliveries, etc. The ?rm operates a web care team, though it does
only respond to a limited amount of complaining and questioning customers. 20% of the PostNL posts have
been classi?ed as public image posts, posts made by individuals talking about the ?rm. A surprising ?gure in
PostNL’s social media classi?cation overview is the high percentage – 14% – of employee posts. Apparently,
PostNL’s employees – and especially postmen – share that they are working at the ?rm.
107
TomTom
TomTom N.V. is a Dutch producer of automotive navigation systems. TomTom is Europe’s leading manufacturer
of navigation systems. The ?rm employs around 3.500 people. During the monitoring period, 32.748 social media
messages have been collected using the search queries related to TomTom. In line with other ?rms, 91% of
these messages have been derived from Twitter.
Of all classi?ed TomTom messages, 49% has been labelled unde?ned, implying that these messages could not
be assigned to any of the other categories. These messages relate to the ?rm, although people use TomTom in
their message, though these messages do not contain any valuable information for the ?rm. The high percentage
of unde?ned posts is due to the fact that people use the word TomTom as a term for navigation systems in
generally, or to refer to anyone who is navigating. Apparently, TomTom has become a word in the general
vocabulary used by the society, though the relation to the ?rm TomTom is not always present. 7% of the
posts are related to product and service quality, in which users share experiences of the usage of TomTom’s
products. Another 8% of the posts are classi?ed as product and service innovation posts, in which users either
make innovative suggestions for future products, or share their opinion towards new products / services. These
two categories contain valuable information for R&D departments, on the one hand to measure the success of
existing products and on the other hand to develop new products.
Unibail-Rodamco
Unibail-Rodamco is a ?rm specialised in commercial property investments. It is the largest commercial real
estate company in Europe, managing three types of assets; shopping centers, convention centers and o?ce
properties. Unibail-Rodamco employs around 1.500 people. Only 512 social media messages related to
Unibail-Rodamco have been collected, i.e. 39 daily posts on average. 95% of these posts have been sourced
from Twitter.
More than in any other ?rm in our sample, as illustrated by ?gure B-1 75% of the posts related to
Unibail-Rodamco relate to ?nancial results. These posts are either related to ?nancial performance discussions
of the ?rm or to stock related discussions. Furthermore, 11% of the classi?ed posts are messages classi?ed as
professionals; people writing about the ?rm from a professional point of view.
Appendix C
Social Media Platform Distribution
Table C-1: Social Media Channel Distribution
Platform Facebook Twitter Blogs News Other Total
Abs % Abs % Abs % Abs % Abs % Abs
ABN AMRO 124 2% 3.000 42% 70 1% 15 0% 3.858 55% 7.067
Aegon 110 8% 1.173 81% 79 5% 20 1% 67 5% 1.449
Akzo Nobel 30 3% 806 87% 43 5% 25 3% 18 2% 922
Albert Heijn 328 3% 11.116 96% 77 1% 1 0% 59 1% 11.581
Arcadis 8 2% 422 93% 9 2% 10 2% 6 1% 455
ArcelorMittal 439 8% 4.569 83% 296 5% 89 2% 139 3% 5.532
Blokker 155 6% 2.526 91% 71 3% 3 0% 14 1% 2.769
Bol.com 472 8% 5.124 89% 115 2% - 0% 71 1% 5.782
C1000 362 3% 10.583 96% 81 1% 5 0% 33 0% 11.064
Coca-Cola 1.653 5% 29.347 89% 999 3% 69 0% 885 3% 32.953
Fugro 6 1% 385 90% 20 5% 15 4% 2 0% 428
Heineken 5.726 15% 32.332 82% 494 1% 122 0% 751 2% 39.425
KLM 2.316 9% 22.601 86% 617 2% 90 0% 740 3% 26.364
NS 703 12% 4.970 85% 103 2% - 0% 87 1% 5.863
Philips 4.641 12% 26.260 68% 3.404 9% 138 0% 4.007 10% 38.450
PostNL 77 6% 1.207 91% 27 2% - 0% 12 1% 1.323
TomTom 1.308 4% 29.787 91% 630 2% 67 0% 956 3% 32.748
Unibail-Rodamco 3 1% 487 95% 9 2% 12 2% 1 0% 512
Total 18.461 8% 186.695 83% 7.144 3% 681 0% 11.706 5% 224.687
109
2%
42%
55%
ABN AMRO
8%
81%
5%
1%
5%
Aegon
n=7.067 n=1.449
1%
0%
Facebook Twitter Blogs News Other Platforms
81%
Facebook Twitter Blogs News Other Platforms
2%
2%
2%
1%
Arcadis
8%
5%
2%
2%
ArcelorMittal
n=455 n=5.532
93%
Facebook Twitter Blogs News Other Platforms
83%
Facebook Twitter Blogs News Other Platforms
3%
1%
0%
0% 5%
3%
0%
3%
n=11.064 n=32.953
96%
C-1000
Facebook Twitter Blogs News Other Platforms
89%
Coca-Cola
Facebook Twitter Blogs News Other Platforms
110 Social Media Platform Distribution
3%
87%
5%
3%
2%
Akzo Nobel
3%
1%
0%
0%
Albert Heijn
n=922 n=11.581
87%
Facebook Twitter Blogs News Other Platforms
96%
Facebook Twitter Blogs News Other Platforms
6%
3%
0%
0%
Blokker
8%
2%
0%
1%
Bol.com
n=2.769 n=5.782
91%
Facebook Twitter Blogs News Other Platforms
89%
Facebook Twitter Blogs News Other Platforms
1%
5%
4%
0%
15%
1%
0%
2%
n=428 n=39.425
90%
Fugro
Facebook Twitter Blogs News Other Platforms
82%
Heineken
Facebook Twitter Blogs News Other Platforms
111
4%
2%
0%
3%
TomTom
1%
2%
2%
0%
Unibail-Rodamco
n=32.748 n=512
9%
2%
0%
3%
KLM
12%
2%
0%
1%
NS
91%
Facebook Twitter Blogs News Other Platforms
95%
Facebook Twitter Blogs News Other Platforms
n=26.364 n=5.863
86%
Facebook Twitter Blogs News Other Platforms
85%
Facebook Twitter Blogs News Other Platforms
12%
9% 0%
11%
6%
2%
0%
1%
n=38.450 n=1.323
68%
Philips
Facebook Twitter Blogs News Other Platforms
91%
PostNL
Facebook Twitter Blogs News Other Platforms
Appendix D
Descriptive Statistics of Social Media
Post Categories
Social Media Categories across Di?erent Customer Relation Types
Table D-1: Social Media Post Categories: Across Customer Relation Type (Descriptives)
Customer Relation Type B2C B2B Total
µ N ? µ N ? µ N ?
short_term_?nancial_results 1,27% 13 1,54% 37,00% 5 23,27% 11,20% 18 20,01%
?nancial_performance_discussions 0,20% 13 0,44% 17,70% 5 21,04% 5,06% 18 13,01%
stock_related_discussions 1,07% 13 1,56% 19,31% 5 13,10% 6,13% 18 10,62%
customer_relations 12,76% 13 12,34% 0,56% 5 0,93% 9,37% 18 11,80%
explaining_?rm 1,92% 13 1,87% 0,07% 5 0,10% 1,41% 18 1,79%
understanding_?rm 1,95% 13 2,19% 0,07% 5 0,10% 1,43% 18 2,04%
thanking_?rm 0,43% 13 0,41% 0,26% 5 0,59% 0,39% 18 0,45%
informing_?rm 0,83% 13 2,98% 0,00% 5 0,00% 0,60% 18 2,54%
questioning_?rm 2,15% 13 1,79% 0,08% 5 0,09% 1,57% 18 1,78%
complaining_customer 4,73% 13 8,16% 0,02% 5 0,04% 3,42% 18 7,19%
thanking_customer 0,60% 13 0,64% 0,07% 5 0,10% 0,45% 18 0,59%
employee_relations 4,08% 13 5,91% 5,59% 5 4,17% 4,50% 18 5,41%
recruitment 1,34% 13 1,91% 4,85% 5 3,78% 2,31% 18 2,93%
employee_posts 2,74% 13 4,43% 0,66% 5 0,51% 2,16% 18 3,86%
operational_performance 0,22% 13 0,50% 0,00% 5 0,00% 0,16% 18 0,43%
product_and_service_quality 1,77% 13 2,51% 0,05% 5 0,12% 1,29% 18 2,26%
alliances 0,44% 13 0,88% 0,13% 5 0,20% 0,35% 18 0,76%
supplier_relations 0,11% 13 0,28% 0,00% 5 0,00% 0,08% 18 0,24%
environmental_performance 0,15% 13 0,43% 0,00% 5 0,00% 0,11% 18 0,37%
product_and_service_innovation 1,52% 13 2,21% 0,28% 5 0,63% 1,17% 18 1,97%
community 39,81% 13 20,50% 44,61% 5 17,79% 41,14% 18 19,39%
promotion 7,08% 13 15,50% 2,46% 5 3,15% 5,79% 18 13,28%
news 1,22% 13 1,46% 6,82% 5 8,21% 2,78% 18 4,90%
public_image 25,50% 13 14,61% 13,61% 5 12,12% 22,20% 18 14,67%
professionals 1,27% 13 1,21% 18,14% 5 15,98% 5,96% 18 11,02%
distributors 4,73% 13 15,36% 3,58% 5 8,00% 4,41% 18 13,48%
unde?ned 20,74% 13 18,88% 10,19% 5 4,90% 17,81% 18 16,77%
spam 17,13% 13 20,00% 1,58% 5 1,24% 12,81% 18 18,28%
113
S
o
c
i
a
l
M
e
d
i
a
C
a
t
e
g
o
r
i
e
s
a
c
r
o
s
s
D
i
?
e
r
e
n
t
I
n
d
u
s
t
r
i
e
s
I n d u s t r y
S
o
c
i
a
l
M
e
d
i
a
P
o
s
t
C
a
t
e
g
o
r
y
?
N
?
?
N
?
?
N
?
?
N
?
?
N
?
?
N
?
?
N
?
?
N
?
s
h
o
r
t
_
t
e
r
m
_
f
i
n
a
n
c
i
a
l
_
r
e
s
u
l
t
s
1
8
,
5
9
%
2
3
,
4
7
%
1
,
3
3
%
3
1
,
5
1
%
0
,
0
0
%
3
0
,
0
0
%
1
,
1
4
%
3
0
,
9
5
%
1
,
5
4
%
2
2
,
1
8
%
2
7
,
1
5
%
3
4
1
,
8
2
%
3
6
,
2
3
%
2
0
,
6
8
%
1
1
,
2
0
%
1
8
2
0
,
0
1
%
f
i
n
a
n
c
i
a
l
_
p
e
r
f
o
r
m
a
n
c
e
_
d
i
s
c
u
s
s
i
o
n
s
4
,
5
8
%
2
5
,
2
5
%
0
,
2
5
%
3
0
,
4
3
%
0
,
0
0
%
3
0
,
0
0
%
0
,
5
2
%
3
0
,
8
4
%
0
,
1
8
%
2
0
,
2
6
%
1
5
,
8
2
%
3
2
7
,
4
0
%
1
5
,
9
3
%
2
2
2
,
5
3
%
5
,
0
6
%
1
8
1
3
,
0
1
%
s
t
o
c
k
_
r
e
l
a
t
e
d
_
d
i
s
c
u
s
s
i
o
n
s
1
4
,
0
1
%
2
8
,
7
2
%
1
,
0
8
%
3
1
,
6
5
%
0
,
0
0
%
3
0
,
0
0
%
0
,
6
3
%
3
1
,
0
8
%
1
,
3
6
%
2
1
,
9
3
%
1
1
,
3
3
%
3
1
4
,
4
8
%
2
0
,
2
9
%
2
2
1
,
8
6
%
6
,
1
3
%
1
8
1
0
,
6
2
%
c
u
s
t
o
m
e
r
_
r
e
l
a
t
i
o
n
s
0
,
3
1
%
2
0
,
1
8
%
2
,
2
7
%
3
1
,
8
3
%
9
,
6
0
%
3
9
,
7
4
%
2
8
,
7
6
%
3
1
2
,
9
9
%
1
1
,
2
5
%
2
7
,
9
4
%
7
,
1
6
%
3
9
,
1
6
%
1
,
1
0
%
2
1
,
5
5
%
9
,
3
7
%
1
8
1
1
,
8
0
%
e
x
p
l
a
i
n
i
n
g
_
f
i
r
m
0
,
0
5
%
2
0
,
0
8
%
0
,
4
1
%
3
0
,
3
6
%
1
,
5
6
%
3
2
,
7
0
%
3
,
4
2
%
3
1
,
4
4
%
2
,
1
2
%
2
0
,
6
8
%
1
,
5
4
%
3
2
,
4
2
%
0
,
1
1
%
2
0
,
1
6
%
1
,
4
1
%
1
8
1
,
7
9
%
u
n
d
e
r
s
t
a
n
d
i
n
g
_
f
i
r
m
0
,
0
5
%
2
0
,
0
8
%
0
,
2
9
%
3
0
,
4
7
%
1
,
3
2
%
3
2
,
2
9
%
3
,
9
0
%
3
3
,
1
7
%
2
,
9
7
%
2
1
,
2
4
%
0
,
9
6
%
3
1
,
2
7
%
0
,
1
1
%
2
0
,
1
6
%
1
,
4
3
%
1
8
2
,
0
4
%
t
h
a
n
k
i
n
g
_
f
i
r
m
0
,
0
0
%
2
0
,
0
0
%
0
,
3
1
%
3
0
,
5
4
%
0
,
2
9
%
3
0
,
5
1
%
0
,
5
9
%
3
0
,
5
3
%
0
,
5
6
%
2
0
,
0
3
%
0
,
3
1
%
3
0
,
3
7
%
0
,
6
6
%
2
0
,
9
3
%
0
,
3
9
%
1
8
0
,
4
5
%
i
n
f
o
r
m
i
n
g
_
f
i
r
m
0
,
0
0
%
2
0
,
0
0
%
0
,
0
0
%
3
0
,
0
0
%
0
,
0
0
%
3
0
,
0
0
%
3
,
5
9
%
3
6
,
2
1
%
0
,
0
0
%
2
0
,
0
0
%
0
,
0
0
%
3
0
,
0
0
%
0
,
0
0
%
2
0
,
0
0
%
0
,
6
0
%
1
8
2
,
5
4
%
q
u
e
s
t
i
o
n
i
n
g
_
f
i
r
m
0
,
1
0
%
2
0
,
0
1
%
0
,
7
4
%
3
0
,
5
8
%
2
,
7
3
%
3
2
,
1
4
%
3
,
0
9
%
3
2
,
3
0
%
2
,
5
5
%
2
2
,
3
2
%
1
,
0
5
%
3
1
,
3
6
%
0
,
1
1
%
2
0
,
1
6
%
1
,
5
7
%
1
8
1
,
7
8
%
c
o
m
p
l
a
i
n
i
n
g
_
c
u
s
t
o
m
e
r
0
,
0
5
%
2
0
,
0
6
%
0
,
3
9
%
3
0
,
2
1
%
3
,
1
2
%
3
2
,
5
4
%
1
3
,
5
5
%
3
1
5
,
1
7
%
2
,
3
1
%
2
2
,
6
2
%
1
,
8
9
%
3
1
,
9
0
%
0
,
0
0
%
2
0
,
0
0
%
3
,
4
2
%
1
8
7
,
1
9
%
t
h
a
n
k
i
n
g
_
c
u
s
t
o
m
e
r
0
,
0
5
%
2
0
,
0
8
%
0
,
1
3
%
3
0
,
1
4
%
0
,
5
8
%
3
0
,
9
0
%
0
,
6
3
%
3
0
,
1
2
%
0
,
7
4
%
2
1
,
0
4
%
0
,
7
7
%
3
0
,
9
0
%
0
,
1
1
%
2
0
,
1
6
%
0
,
4
5
%
1
8
0
,
5
9
%
e
m
p
l
o
y
e
e
_
r
e
l
a
t
i
o
n
s
6
,
1
8
%
2
5
,
5
2
%
0
,
8
6
%
3
0
,
7
8
%
6
,
9
8
%
3
2
,
0
5
%
8
,
0
6
%
3
1
1
,
5
7
%
0
,
1
7
%
2
0
,
2
5
%
2
,
0
7
%
3
1
,
4
4
%
7
,
2
1
%
2
3
,
6
8
%
4
,
5
0
%
1
8
5
,
4
1
%
r
e
c
r
u
i
t
m
e
n
t
5
,
5
8
%
2
5
,
3
1
%
0
,
8
1
%
3
0
,
8
5
%
0
,
7
4
%
3
0
,
4
9
%
3
,
2
7
%
3
3
,
5
6
%
0
,
1
7
%
2
0
,
2
5
%
1
,
2
3
%
3
0
,
6
0
%
5
,
9
6
%
2
3
,
4
6
%
2
,
3
1
%
1
8
2
,
9
3
%
e
m
p
l
o
y
e
e
_
p
o
s
t
s
0
,
6
1
%
2
0
,
2
1
%
0
,
0
5
%
3
0
,
0
8
%
6
,
2
3
%
3
2
,
3
7
%
4
,
7
8
%
3
8
,
1
7
%
0
,
0
0
%
2
0
,
0
0
%
0
,
8
0
%
3
0
,
9
6
%
1
,
0
3
%
2
0
,
5
3
%
2
,
1
6
%
1
8
3
,
8
6
%
o
p
e
r
a
t
i
o
n
a
l
_
p
e
r
f
o
r
m
a
n
c
e
0
,
0
0
%
2
0
,
0
0
%
0
,
0
0
%
3
0
,
0
0
%
0
,
0
0
%
3
0
,
0
0
%
0
,
4
7
%
3
0
,
7
3
%
0
,
0
2
%
2
0
,
0
3
%
0
,
4
5
%
3
0
,
7
8
%
0
,
0
0
%
2
0
,
0
0
%
0
,
1
6
%
1
8
0
,
4
3
%
p
r
o
d
u
c
t
_
a
n
d
_
s
e
r
v
i
c
e
_
q
u
a
l
i
t
y
0
,
1
4
%
2
0
,
1
9
%
3
,
6
7
%
3
3
,
7
3
%
0
,
9
5
%
3
0
,
5
9
%
0
,
5
3
%
3
0
,
4
1
%
3
,
3
1
%
2
4
,
5
7
%
0
,
3
1
%
3
0
,
5
3
%
0
,
0
0
%
2
0
,
0
0
%
1
,
2
9
%
1
8
2
,
2
6
%
a
l
l
i
a
n
c
e
s
0
,
2
2
%
2
0
,
3
1
%
0
,
0
3
%
3
0
,
0
5
%
0
,
0
0
%
3
0
,
0
0
%
0
,
8
6
%
3
1
,
4
8
%
0
,
3
8
%
2
0
,
2
8
%
0
,
7
6
%
3
1
,
2
5
%
0
,
1
2
%
2
0
,
1
7
%
0
,
3
5
%
1
8
0
,
7
6
%
s
u
p
p
l
i
e
r
_
r
e
l
a
t
i
o
n
s
0
,
0
0
%
2
0
,
0
0
%
0
,
3
4
%
3
0
,
5
9
%
0
,
0
5
%
3
0
,
0
8
%
0
,
0
2
%
3
0
,
0
4
%
0
,
0
0
%
2
0
,
0
0
%
0
,
0
6
%
3
0
,
1
0
%
0
,
0
0
%
2
0
,
0
0
%
0
,
0
8
%
1
8
0
,
2
4
%
e
n
v
i
r
o
n
m
e
n
t
a
l
_
p
e
r
f
o
r
m
a
n
c
e
0
,
0
0
%
2
0
,
0
0
%
0
,
0
5
%
3
0
,
0
5
%
0
,
5
5
%
3
0
,
9
0
%
0
,
0
0
%
3
0
,
0
0
%
0
,
0
0
%
2
0
,
0
0
%
0
,
0
5
%
3
0
,
0
9
%
0
,
0
0
%
2
0
,
0
0
%
0
,
1
1
%
1
8
0
,
3
7
%
p
r
o
d
u
c
t
_
a
n
d
_
s
e
r
v
i
c
e
_
i
n
n
o
v
a
t
i
o
n
0
,
7
1
%
2
1
,
0
0
%
1
,
2
1
%
3
0
,
4
0
%
0
,
3
5
%
3
0
,
5
5
%
0
,
3
7
%
3
0
,
3
7
%
3
,
8
8
%
2
5
,
4
3
%
2
,
0
7
%
3
2
,
2
2
%
0
,
0
0
%
2
0
,
0
0
%
1
,
1
7
%
1
8
1
,
9
7
%
c
o
m
m
u
n
i
t
y
5
7
,
4
5
%
2
4
,
0
9
%
5
4
,
6
4
%
3
9
,
2
9
%
2
6
,
8
6
%
3
1
,
4
7
%
2
0
,
1
5
%
3
1
1
,
8
0
%
4
9
,
2
8
%
2
3
2
,
8
3
%
4
2
,
9
0
%
3
2
8
,
3
8
%
4
6
,
7
4
%
2
3
,
6
2
%
4
1
,
1
4
%
1
8
1
9
,
3
9
%
p
r
o
m
o
t
i
o
n
1
,
7
9
%
2
0
,
6
9
%
1
,
8
2
%
3
0
,
9
6
%
0
,
8
7
%
3
0
,
8
4
%
1
,
3
3
%
3
0
,
9
1
%
1
1
,
3
9
%
2
1
2
,
6
3
%
1
9
,
2
6
%
3
3
1
,
4
8
%
4
,
0
7
%
2
5
,
4
3
%
5
,
7
9
%
1
8
1
3
,
2
8
%
n
e
w
s
1
3
,
4
2
%
2
1
0
,
4
7
%
1
,
7
0
%
3
2
,
0
6
%
0
,
6
5
%
3
1
,
0
2
%
0
,
7
5
%
3
0
,
3
9
%
0
,
0
0
%
2
0
,
0
0
%
2
,
4
0
%
3
1
,
6
5
%
3
,
3
3
%
2
3
,
0
6
%
2
,
7
8
%
1
8
4
,
9
0
%
p
u
b
l
i
c
_
i
m
a
g
e
2
1
,
6
9
%
2
1
4
,
5
7
%
2
9
,
6
5
%
3
2
2
,
5
6
%
2
4
,
7
9
%
3
1
,
4
8
%
1
6
,
8
2
%
3
1
1
,
7
7
%
3
5
,
8
7
%
2
2
0
,
1
5
%
1
6
,
3
2
%
3
1
7
,
5
8
%
1
0
,
8
7
%
2
1
0
,
7
4
%
2
2
,
2
0
%
1
8
1
4
,
6
7
%
p
r
o
f
e
s
s
i
o
n
a
l
s
1
1
,
6
0
%
2
1
1
,
9
6
%
2
,
0
7
%
3
1
,
4
5
%
0
,
1
1
%
3
0
,
1
1
%
1
,
2
5
%
3
0
,
8
5
%
1
,
0
3
%
2
1
,
4
6
%
4
,
9
2
%
3
4
,
9
9
%
2
8
,
4
7
%
2
2
2
,
8
5
%
5
,
9
6
%
1
8
1
1
,
0
2
%
d
i
s
t
r
i
b
u
t
o
r
s
8
,
9
5
%
2
1
2
,
6
5
%
1
9
,
3
9
%
3
3
1
,
5
2
%
0
,
4
5
%
3
0
,
7
7
%
0
,
0
0
%
3
0
,
0
0
%
1
,
0
0
%
2
1
,
4
1
%
0
,
0
0
%
3
0
,
0
0
%
0
,
0
0
%
2
0
,
0
0
%
4
,
4
1
%
1
8
1
3
,
4
8
%
u
n
d
e
f
i
n
e
d
1
3
,
5
8
%
2
5
,
5
5
%
1
7
,
0
9
%
3
1
4
,
6
5
%
4
3
,
5
2
%
3
6
,
5
9
%
5
,
0
6
%
3
4
,
8
6
%
2
6
,
2
5
%
2
3
1
,
8
5
%
9
,
5
9
%
3
6
,
3
9
%
7
,
5
9
%
2
5
,
1
2
%
1
7
,
8
1
%
1
8
1
6
,
7
7
%
s
p
a
m
2
,
8
3
%
2
0
,
6
3
%
1
8
,
5
1
%
3
5
,
8
4
%
1
1
,
1
5
%
3
9
,
8
4
%
3
4
,
5
9
%
3
3
7
,
7
4
%
3
,
9
1
%
2
2
,
7
1
%
7
,
4
3
%
3
1
2
,
4
5
%
1
,
0
2
%
2
0
,
1
2
%
1
2
,
8
1
%
1
8
1
8
,
2
8
%
W h o l e s a l e a n d R e t a i l
T r a n s p o r t a n d S t o r a g e
I n f o r m a t i o n a n d
C o m m u n i c a t i o n
F i n a n c i a l I n s t i t u t i o n s
C o n s u l t a n c y , R e s e a r c h
a n d O t h e r S p e c i a l i s e d
B u s i n e s s S e r v i c e s
T o t a l
M i n i n g a n d Q u a r r y i n g
I n d u s t r y
F
i
g
u
r
e
D
-
1
:
S
o
c
i
a
l
M
e
d
i
a
P
o
s
t
C
a
t
e
g
o
r
i
e
s
:
A
c
r
o
s
s
I
n
d
u
s
t
r
i
e
s
(
D
e
s
c
r
i
p
t
i
v
e
s
)
114 Descriptive Statistics of Social Media Post Categories
Boxplots of Social Media Categories across Firms
Case Processing Summary
Cases
Valid Missing Total
N Percent N Percent N Percent
alliances
supplier_relations
environmental_performan
ce
product_and_service_inno
vation
community
promotion
news
public_image
professionals
distributors
undefined
spam
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
short_term_financial_results
short_term_financial_results
0,8
0,6
0,4
0,2
0,0
Unibail-Rodamco
Page 2
customer_relations
0,4
0,3
0,2
0,1
0,0
explaining_firm
Page 5
employee_relations
0,25
0,20
0,15
0,10
0,05
0,00
PostNL
recruitment
Page 13
operational_performance
1,25E-2
1,0E-2
7,5E-3
5,0E-3
2,5E-3
0,0E0
ABN AMRO
NS
PostNL
Bol.com
product_and_service_quality
Page 16
product_and_service_quality
0,08
0,06
0,04
0,02
0,00
Philips
TomTom
alliances
Page 17
alliances
0,030
0,025
0,020
0,015
0,010
0,005
0,000
PostNL
Aegon
supplier_relations
Page 18
115
supplier_relations
1,2E-2
1,0E-2
8,0E-3
6,0E-3
4,0E-3
2,0E-3
0,0E0
Heineken
ABN AMRO
Albert Heijn
NS
environmental_performance
Page 19
environmental_performance
0,020
0,015
0,010
0,005
0,000
Albert Heijn
ABN AMRO
product_and_service_innovation
Page 20
product_and_service_innovation
0,08
0,06
0,04
0,02
0,00
TomTom
Aegon
community
Page 21
community
0,8
0,6
0,4
0,2
0,0
promotion
Page 22
undefined
0,6
0,5
0,4
0,3
0,2
0,1
0,0
spam
Page 28
spam
0,8
0,6
0,4
0,2
0,0
KLM
Page 29
Appendix E
Corporate Engagement
The next page lists the user names of ?rms that have been found in our dataset. These user names have been
used to assign social media messages in the category ‘?rm-to-customer’.
117
I
D
A
B
N
A
M
R
O
#
P
o
s
t
s
A
e
g
o
n
#
P
o
s
t
s
A
k
z
o
N
o
b
e
l
#
P
o
s
t
s
A
l
b
e
r
t
H
e
i
j
n
#
P
o
s
t
s
A
r
c
a
d
i
s
#
P
o
s
t
s
A
r
c
e
l
o
r
M
i
t
t
a
l
#
P
o
s
t
s
B
l
o
k
k
e
r
#
P
o
s
t
s
B
o
l
.
c
o
m
#
P
o
s
t
s
C
-
1
0
0
0
#
P
o
s
t
s
1
a
b
n
a
m
r
o
_
t
e
g
e
l
e
n
2
A
e
g
o
n
R
e
l
i
g
a
r
e
L
i
f
e
I
n
s
u
r
a
n
c
e
1
A
k
z
o
N
o
b
e
l
_
A
&
A
C
2
A
l
b
e
r
t
H
e
i
j
n
N
u
l
a
n
d
3
A
R
C
A
D
I
S
G
S
1
A
r
c
e
l
o
r
M
i
t
t
a
l
U
S
A
1
B
l
o
k
k
e
r
2
b
o
l
.
c
o
m
b
a
b
y
4
9
C
1
0
0
0
K
o
e
t
s
i
e
r
1
2
A
B
N
A
M
R
O
C
o
m
m
e
r
c
i
a
l
1
A
E
G
O
N
S
e
g
u
r
o
s
1
A
k
z
o
N
o
b
e
l
G
r
a
d
u
a
t
e
s
2
A
l
b
e
r
t
H
e
i
j
n
S
t
i
e
n
s
3
A
R
C
A
D
I
S
L
F
N
1
a
r
c
e
l
o
r
m
i
t
t
a
l
_
d
0
B
l
o
k
k
e
r
2
B
o
l
.
c
o
m
F
i
l
m
T
V
4
7
C
1
0
0
0
N
o
o
r
d
m
a
n
1
3
A
B
N
A
M
R
O
E
c
h
t
2
A
E
G
O
N
U
K
1
a
k
z
o
n
o
b
e
l
a
e
r
o
1
A
l
b
e
r
t
H
e
i
j
n
S
t
r
i
j
p
1
A
R
C
A
D
I
S
S
h
e
l
t
e
r
4
A
r
c
e
l
o
r
M
i
t
t
a
l
D
o
f
a
s
c
o
0
b
l
o
k
k
e
r
h
o
l
t
e
n
1
b
o
l
.
c
o
m
g
a
m
e
s
9
C
1
0
0
0
P
u
t
s
1
4
a
b
n
a
m
r
o
e
c
h
t
2
a
e
g
o
n
_
n
l
3
A
k
z
o
N
o
b
e
l
A
e
r
o
0
a
l
b
e
r
t
h
e
i
j
n
1
a
r
c
a
d
i
s
_
l
f
n
0
A
r
c
e
l
o
r
M
i
t
t
a
l
2
B
o
l
.
c
o
m
M
u
z
i
e
k
4
4
C
1
0
0
0
_
F
r
a
n
e
k
e
r
1
5
A
B
N
A
M
R
O
M
E
P
P
E
L
5
4
7
A
E
G
O
N
N
e
d
e
r
l
a
n
d
9
G
r
a
d
s
@
A
k
z
o
N
o
b
e
l
1
A
l
b
e
r
t
H
e
i
j
n
X
L
T
i
l
b
u
r
g
2
6
5
A
R
C
A
D
I
S
B
e
l
g
i
q
u
e
8
A
r
c
e
l
o
r
M
i
t
t
a
l
O
r
b
i
t
1
B
o
l
.
c
o
m
P
a
r
t
n
e
r
1
1
1
c
1
0
0
0
b
e
e
k
m
a
n
1
6
a
b
n
a
m
r
o
n
i
j
m
e
g
e
n
3
.
8
5
1
A
E
G
O
N
R
e
l
i
g
a
r
e
1
5
A
k
z
o
N
o
b
e
l
1
A
l
b
e
r
t
B
e
z
o
r
g
s
e
r
v
i
c
e
1
A
R
C
A
D
I
S
2
7
B
o
l
.
c
o
m
s
e
r
v
i
c
e
2
7
5
c
1
0
0
0
d
a
l
f
s
e
n
2
7
A
B
N
A
M
R
O
1
A
E
G
O
N
1
A
k
z
o
N
o
b
e
l
A
e
r
o
s
p
a
c
e
1
A
l
b
e
r
t
H
e
i
j
n
1
A
R
C
A
D
I
S
C
o
r
p
C
o
m
m
1
B
o
l
.
c
o
m
a
c
t
i
e
s
2
6
C
1
0
0
0
B
e
e
k
m
a
n
1
8
A
B
N
A
M
R
O
N
V
1
A
E
G
O
N
N
e
d
e
r
l
a
n
d
4
4
A
l
b
e
r
t
H
e
i
j
n
1
6
1
8
1
b
o
l
.
c
o
m
5
4
C
1
0
0
0
K
e
e
s
E
l
z
i
n
g
a
6
9
A
B
N
A
M
R
O
M
a
r
k
e
t
s
7
a
e
g
o
n
s
e
g
u
r
o
s
1
A
H
R
a
p
p
o
r
t
s
t
r
a
a
t
1
b
o
l
_
c
o
m
1
2
c
1
0
0
0
d
a
y
s
2
1
0
a
b
n
a
m
r
o
t
u
r
b
o
1
A
E
G
O
N
C
l
a
s
s
i
c
4
b
o
l
_
c
o
m
_
a
c
t
i
e
s
6
c
1
0
0
0
m
u
l
d
e
r
1
1
1
A
B
N
A
M
R
O
Y
E
R
S
E
K
E
O
P
E
N
1
b
o
l
_
c
o
m
_
a
p
i
1
c
1
0
0
0
n
i
j
e
n
e
r
i
n
g
1
1
2
A
B
N
A
M
R
O
E
c
o
n
B
u
r
e
a
u
2
1
b
o
l
_
c
o
m
_
b
a
b
y
2
c
1
0
0
0
o
f
f
e
r
e
i
n
2
1
3
b
o
l
_
c
o
m
_
b
a
n
e
n
2
c
1
0
0
0
p
a
n
n
i
n
g
e
n
1
1
4
b
o
l
_
c
o
m
_
g
a
m
e
s
2
c
1
0
0
0
p
u
t
s
2
1
5
b
o
l
_
c
o
m
_
k
o
k
e
n
4
8
c
1
0
0
0
w
a
h
l
e
1
1
6
b
o
l
_
c
o
m
_
m
u
z
i
e
k
1
C
1
0
0
0
W
i
e
g
m
a
n
s
6
1
7
b
o
l
_
c
o
m
_
p
a
r
t
n
e
r
2
1
8
w
e
r
k
e
n
b
i
j
b
o
l
.
c
o
m
8
O
w
n
p
o
s
t
s
4
.
4
3
7
8
0
8
2
7
7
4
2
4
5
6
9
9
3
0
%
T
o
t
a
l
s
a
m
p
l
e
6
3
%
6
%
1
%
2
%
9
%
0
%
0
%
9
%
0
%
I
D
C
o
c
a
-
C
o
l
a
#
P
o
s
t
s
F
u
g
r
o
#
P
o
s
t
s
H
e
i
n
e
k
e
n
#
P
o
s
t
s
K
L
M
#
P
o
s
t
s
N
S
#
P
o
s
t
s
P
h
i
l
i
p
s
#
P
o
s
t
s
P
o
s
t
N
L
#
P
o
s
t
s
T
o
m
T
o
m
#
P
o
s
t
s
U
n
i
b
a
i
l
R
o
d
a
m
c
o
#
P
o
s
t
s
1
C
o
c
a
-
C
o
l
a
G
B
1
F
u
g
r
o
S
e
a
c
o
r
e
1
H
e
i
n
e
k
e
n
M
u
s
i
c
H
a
l
l
1
9
K
L
M
A
i
r
l
i
n
e
s
2
N
S
o
n
l
i
n
e
6
P
h
i
l
i
p
s
A
r
e
n
a
2
O
R
P
o
s
t
N
L
1
3
t
o
m
t
o
m
b
e
n
e
l
u
x
3
U
n
i
b
a
i
l
R
o
d
a
m
c
o
3
2
C
o
c
a
-
C
o
l
a
2
4
x
7
1
H
e
i
n
e
k
e
n
N
L
4
K
L
M
B
e
l
g
i
u
m
1
N
S
S
t
a
t
i
o
n
Z
w
o
l
l
e
2
P
h
i
l
i
p
s
A
r
g
e
n
t
i
n
a
1
P
o
s
t
N
L
1
T
o
m
T
o
m
7
0
3
C
o
c
a
-
C
o
l
a
A
r
g
e
n
t
i
n
a
3
H
e
i
n
e
k
e
n
t
w
e
e
t
s
8
K
L
M
C
a
n
a
d
a
3
N
S
S
t
a
t
i
o
n
s
6
1
P
h
i
l
i
p
s
A
V
E
N
T
U
K
3
2
P
o
s
t
N
L
P
r
i
d
e
1
T
o
m
T
o
m
B
r
a
s
i
l
4
4
C
o
c
a
-
C
o
l
a
B
o
i
D
a
D
o
n
4
h
e
i
n
e
k
e
n
_
h
o
u
s
e
0
K
L
M
C
o
u
p
o
n
C
o
d
e
s
1
N
S
U
n
o
f
f
i
c
i
a
l
7
7
P
h
i
l
i
p
s
B
r
a
s
i
l
1
P
o
s
t
N
L
V
a
n
d
a
a
g
2
T
o
m
T
o
m
D
e
u
t
s
c
h
l
a
n
d
2
5
C
o
c
a
-
C
o
l
a
B
r
a
s
i
l
3
h
e
i
n
e
k
e
n
_
t
w
e
e
t
s
0
K
L
M
D
e
u
t
s
c
h
l
a
n
d
9
N
S
V
e
r
t
r
a
g
i
n
g
3
P
h
i
l
i
p
s
C
o
r
p
C
o
m
m
s
1
P
o
s
t
N
L
W
e
b
c
a
r
e
2
7
T
o
m
t
o
m
E
n
g
l
i
s
h
1
6
C
o
c
a
-
C
o
l
a
C
a
n
a
d
a
4
7
H
e
i
n
e
k
e
n
E
e
t
c
a
f
e
2
K
L
M
E
C
U
A
D
O
R
1
7
4
4
N
S
v
e
r
t
r
a
g
i
n
g
1
P
h
i
l
i
p
s
E
n
t
L
i
g
h
t
i
n
g
8
P
o
s
t
N
L
R
e
c
r
u
i
t
m
e
n
t
2
T
o
m
T
o
m
F
r
a
n
c
e
7
7
C
o
c
a
C
o
l
a
B
e
l
g
i
u
m
2
H
e
i
n
e
k
e
n
H
o
u
s
e
6
A
i
r
F
r
a
n
c
e
-
K
L
M
2
N
S
1
3
2
P
h
i
l
i
p
s
L
i
g
h
t
i
n
g
3
T
o
m
T
o
m
I
t
a
l
i
a
3
8
C
o
c
a
-
C
o
l
a
1
2
R
o
y
a
l
D
u
t
c
h
A
i
r
l
i
n
e
s
1
N
S
H
i
s
p
e
e
d
2
2
P
h
i
l
i
p
s
1
T
o
m
T
o
m
O
f
f
i
c
i
a
l
1
5
9
C
o
c
a
-
C
o
l
a
C
l
o
t
h
i
n
g
2
K
L
M
f
a
r
e
s
2
6
n
s
v
e
r
t
r
a
g
i
n
g
n
l
5
P
h
i
l
i
p
s
F
a
s
t
4
T
o
m
T
o
m
P
o
r
t
u
g
a
l
3
1
0
C
o
c
a
-
C
o
l
a
C
o
l
o
m
b
i
a
4
K
L
M
I
r
e
l
a
n
d
2
N
S
_
W
i
j
c
h
e
n
2
P
h
i
l
i
p
s
G
r
o
o
m
i
n
g
2
T
o
m
T
o
m
S
o
u
t
h
A
f
r
i
c
a
6
1
1
C
o
c
a
-
C
o
l
a
C
R
1
K
L
M
I
T
A
L
I
A
5
n
s
_
z
w
o
l
l
e
1
P
h
i
l
i
p
s
H
e
a
l
t
h
c
a
r
e
2
7
T
o
m
T
o
m
S
p
a
i
n
7
1
2
C
o
c
a
-
C
o
l
a
E
c
u
a
d
o
r
1
K
L
M
K
a
z
a
k
h
s
t
a
n
6
P
h
i
l
i
p
s
H
o
m
e
L
i
g
h
t
i
n
g
3
T
o
m
T
o
m
S
w
i
t
z
e
r
l
a
n
d
4
1
3
C
o
c
a
-
C
o
l
a
E
l
S
a
l
v
a
d
o
r
1
K
L
M
M
e
x
i
c
o
9
P
h
i
l
i
p
s
I
n
L
i
g
h
t
i
n
g
3
1
4
C
o
c
a
-
C
o
l
a
E
s
p
a
ñ
a
2
K
L
M
R
U
S
S
I
A
1
P
h
i
l
i
p
s
I
n
d
i
a
2
1
5
C
o
c
a
C
o
l
a
F
M
C
O
L
O
M
B
I
A
3
K
L
M
S
p
a
i
n
3
4
P
h
i
l
i
p
s
I
n
d
o
n
e
s
i
a
1
0
1
1
6
C
o
c
a
-
C
o
l
a
F
M
E
C
U
A
D
O
R
2
K
L
M
U
K
1
6
P
h
i
l
i
p
s
J
a
p
a
n
9
1
7
C
o
c
a
-
C
o
l
a
F
r
e
e
s
t
y
l
e
2
K
L
M
z
a
k
e
l
i
j
k
1
P
h
i
l
i
p
s
J
o
b
s
N
A
9
1
8
C
o
c
a
-
C
o
l
a
G
B
2
k
l
m
_
e
s
1
P
h
i
l
i
p
s
J
o
b
s
N
A
9
1
9
C
o
c
a
-
C
o
l
a
G
u
a
t
e
m
a
l
a
1
K
L
M
_
J
P
1
0
P
h
i
l
i
p
s
L
i
g
h
t
i
n
g
1
2
2
0
C
o
c
a
-
C
o
l
a
H
B
C
I
t
a
l
i
a
1
k
l
m
_
p
t
1
P
h
i
l
i
p
s
S
p
a
i
n
4
2
1
C
o
c
a
-
C
o
l
a
H
o
n
d
u
r
a
s
1
k
l
m
_
u
k
1
P
h
i
l
i
p
s
S
u
p
p
o
r
t
U
K
1
2
2
C
o
c
a
-
C
o
l
a
I
n
d
o
n
e
s
i
a
3
k
l
m
_
u
s
1
P
h
i
l
i
p
s
T
h
a
i
l
a
n
d
1
0
2
3
C
o
c
a
-
C
o
l
a
K
o
r
e
a
1
0
5
k
l
m
f
a
r
e
s
3
P
h
i
l
i
p
s
T
u
r
k
i
y
e
2
6
2
4
C
o
c
a
-
C
o
l
a
N
i
c
a
r
a
g
u
a
1
k
l
m
i
t
a
l
i
a
1
P
h
i
l
i
p
s
U
K
G
r
a
d
u
a
t
e
s
4
6
2
5
C
o
c
a
-
C
o
l
a
P
a
n
a
m
á
1
k
l
m
m
a
l
a
y
s
i
a
1
P
h
i
l
i
p
s
_
N
e
d
e
r
l
a
n
d
6
2
6
C
o
c
a
-
C
o
l
a
V
e
n
e
z
u
e
l
a
1
1
k
l
m
n
o
r
w
a
y
1
P
h
i
l
i
p
s
C
a
r
e
N
L
4
2
6
C
o
c
a
-
C
o
l
a
V
e
n
e
z
u
e
l
a
1
1
k
l
m
n
o
r
w
a
y
1
P
h
i
l
i
p
s
C
a
r
e
N
L
4
2
7
C
o
c
a
-
C
o
l
a
W
e
b
3
K
L
M
R
o
y
a
l
D
u
t
c
h
A
i
r
l
i
n
e
1
p
h
i
l
i
p
s
j
a
p
a
n
2
2
8
C
o
c
a
C
o
l
a
B
u
f
f
a
l
o
1
p
h
i
l
i
p
s
j
o
b
s
n
a
3
2
9
C
o
c
a
C
o
l
a
I
s
r
a
e
l
3
9
p
h
i
l
i
p
s
l
i
g
h
t
2
3
0
C
o
c
a
C
o
l
a
T
v
1
P
h
i
l
i
p
s
L
i
g
h
t
i
n
g
1
3
1
p
h
i
l
i
p
s
n
l
j
o
b
s
1
3
2
P
h
i
l
i
p
s
S
o
u
n
d
P
r
o
j
e
c
t
3
4
3
3
p
h
i
l
i
p
s
s
p
a
i
n
2
3
4
P
h
i
l
i
p
s
T
u
r
k
i
y
e
0
3
5
p
h
i
l
i
p
s
u
k
g
r
a
d
s
1
8
O
w
n
p
o
s
t
s
2
6
1
0
3
9
1
8
8
4
3
1
2
3
9
0
4
6
1
2
5
3
%
T
o
t
a
l
s
a
m
p
l
e
1
%
0
%
0
%
7
%
5
%
1
%
3
%
0
%
1
%
F
i
g
u
r
e
E
-
1
:
S
o
c
i
a
l
M
e
d
i
a
P
o
s
t
s
C
l
a
s
s
i
?
c
a
t
i
o
n
Bibliography
Agichtein, E., Castillo, C., Donato, D., Gionis, A., & Mishne, G. (2008). Finding high-quality content in social
media. In Proceedings of the international conference on web search and web data mining (pp. 183–194).
Association for Computing Machinery.
Araujo, T., & Neijens, P. (2012). Friend me: Which factors in?uence top global brands participation in social
network sites. Internet Research, 22, 626 – 640.
Auditore, P. J. (2012a). The market for social media and business intelligence tools. Asteria Research.
Retrieved July 31, 2012, from http://sandhill.com/article/the-market-for-social-media-and
-business-intelligence-tools/
Auditore, P. J. (2012b). Share-unisphere research study: Adoption of social media bi on the rise.
SHARE-Unisphere. Retrieved July 31, 2012, from http://www.dbta.com/Articles/Editorial/
Trends-and-Applications/SHARE-Unisphere-Research-Study-Adoption-of-Social-Media-BI-on
-the-Rise-83119.aspx
Azvine, B., Cui, Z., & Nauck, D. D. (2005). Towards real-time business intelligence. BT Technology Journal ,
23, 214–225.
Berthon, P. R., Pitt, L. F., McCarthy, I., & Kates, S. M. (2007). When customers get clever: Managerial
approaches to dealing with creative consumers. Business Horizons, 50(1), 39–47.
Bonneau, J., & Preibusch, S. (2010). The privacy jungle: On the market for data protection in social networks.
In T. Moore, D. Pym, & C. Ioannidis (Eds.), Economics of information security and privacy (pp. 121–167).
Springer US.
Bos, W., & Tarnai, C. (1999). Content analysis in empirical social research. International Journal of Educational
Research, 31(8), 659–671.
Bouwman, H., Faber, E., Haaker, T., Kijl, B., & de Reuver, M. (2008). Conceptualizing the stof model. In
H. Bouwman, H. de Vos, & T. Haaker (Eds.), Mobile service innovation and business models (pp. 31 –
69). Berlin, Heidelberg: Springer.
Boyd, D. M., & Ellison, N. B. (2007). Social network sites: De?nition, history, and scholarship. Journal of
Computer-Mediated Communication, 13(1), 210–230.
Bryant, L. (2011). Moving from social media monitoring to social business intelligence. Headshift, Dachis Group.
Retrieved July 25, 2012, from http://www.headshift.com/our-blog/2011/06/15/from-social-media
-monitoring-to-social-business-intelligence/
Bughin, J., Chui, M., & Manyika, J. (2010). Clouds, big data, and smart assets: Ten tech-enabled business
trends to watch. McKinsey & Co. Quarterly. (White Paper).
CBS. (2012). Standaard bedrijfsindeling 2008. versie 2012. Centraal Bureau voor de Statistiek.
Cha?ey, D. (2009). E-business and e-commerce management (4th edition). Upper Saddle River, NJ, USA:
Prentice-Hall, Inc.
Chen, D.-N., Jeng, B., Lee, W.-P., & Chuang, C.-H. (2008). An agent-based model for consumer-to-business
electronic commerce. Expert Systems with Applications, 34(1), 469–481.
DesAutels, P. (2011). Ugis: Understanding the nature of user-generated information systems. Business Horizons,
54(3), 185 - 192.
Dey, L., & Haque, S. K. M. (2008). Opinion mining from noisy text data. In Proceedings of the second workshop
on analytics for noisy unstructured text data (pp. 83–90). New York, NY, USA: ACM.
Dhar, V., & Chang, E. A. (2009). Does chatter matter? the impact of user-generated content on music sales.
Journal of Interactive Marketing, 23(4), 300–307.
Bibliography 119
Doeland, D. (2012). Social media & internet key performance indicators. Retrieved August 10,
2012, from http://ddmca.nl/diensten/social-media-and-internet-monitoring/social-media-key
-performance-indicators/
Dong-Hun, L. (2010). Growing popularity of social media and business strategy. Korean Consumer & Society.,
112–117.
Dugan, L. (2012). Twitter to surpass 500 million registered users on wednesday. Retrieved August 15, 2012,
from http://www.mediabistro.com/alltwitter/500-million-registered-users_b18842
Eccles, R. (1991). The performance measurement manifesto. Harvard Business School , 69, 131–137.
Eckerson, W. W. (2009). Performance management strategies. how to create and deploy e?ective metrics. The
Data Warehousing Institute. (White Paper).
Elbashir, M. Z., Collier, P. A., & Davern, M. J. (2008). Measuring the e?ects of business intelligence systems:
The relationship between business process and organizational performance. International Journal of
Accounting Information Systems, 9(3), 135–153.
evolve24. (2012). Website. Retrieved September 26, 2012, from http://www.evolve24.com/
Fortuin, L. (1988). Performance indicators – why, where and how? European Journal of Operational Research,
34, 1–9.
Gartner. (2009). Gartner reveals ?ve business intelligence predictions for 2009 and beyond. Retrieved June 17,
2012, from http://www.gartner.com/it/page.jsp?id=856714
Gast, A., & Zanini, M. (2012). The social side of strategy. McKinsey Quarterly.
Gates, S. (2001). Aligning strategic performance measures and results. The Conference Board. Balanced
Scorecard Interest Group.
Gessner, G. H., & Volonio, L. (2005). Quick response improves returns on business intelligence investments.
Information Systems Management, 22(3), 66–74.
Ghose, A., & Panagiotos, I. (2010). The economining project at nyu: Studying the economic value of
user-generated content on the internet. Journal of Revenue and Pricing Management, 8, 241–246.
Gianfortoni, P., Adamson, D., & Rosé, C. P. (2011, July). Modeling of stylistic variation in social media with
stretchy patterns. In Proceedings of the ?rst workshop on algorithms and resources for modelling of dialects
and language varieties (pp. 49–59). Edinburgh, Scotland: Association for Computational Linguistics.
Guo, J., Zhang, P., Tan, J., & Guo, L. (2012). Mining hot topics from twitter streams. Procedia Computer
Science, 9, 2008–2011.
Hanna, R., Rohm, A., & Crittenden, V. L. (2011). We’re all connected: The power of the social media
ecosystem. Business Horizons, 54(3), 265–273.
Huang, Q., Li, Q., Li, H., & Yan, Z. (2012). An approach to incremental deep web crawling based on incremental
harvest model. Procedia Engineering, 29(0), 1081–1087.
IBM. (2011). From social media to social crm. reinventing the customer relationship. IBM Institute for Business
Value analysis. (White Paper).
Ittner, C., Larcker, D., & Randall, T. (2003). Performance implications of strategic performance measurement
in ?nancial services ?rms. Accounting, Organizations and Society, 28, 715–741.
Jansen, B. J., Zhang, M., Sobel, K., & Chowdury, A. (2009). Twitter power: Tweets as electronic word of
mouth. Journal of the American Society for Information Science, 60(11), 2169–2188.
Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite! the challenges and opportunities of social
media. Business Horizons, 53(1), 59–68.
Kaplan, A. M., & Haenlein, M. (2012). The britney spears universe: Social media and viral marketing at its
best. Business Horizons, 55(1), 27–31.
Kaplan, R. S., & Norton, D. P. (1992). The balanced scorecard – measures that drive performance. Harvard
Business Review(5), 71–79.
Kaplan, R. S., & Norton, D. P. (1993). Putting the balanced scorecard to work. Harvard Business Review,
71(5), 134–147.
Kaplan, R. S., & Norton, D. P. (1996). Using the balanced scorecard as a strategic management system.
Harvard Business Review(5), 1–13.
Kapow. (2009). Five key lessons for converting social media data into business intelligence. Kapow Software
(White Paper).
Kennerley, M., & Neely, A. (2003). Measuring performance in a changing business environment. International
Journal of Operations and Production Management, 23(2), 213–229.
Kietzmann, J. H., Hermkens, K., McCarthy, I. P., & Silvestre, B. S. (2011). Social media? get serious!
understanding the functional building blocks of social media. Business Horizons, 54(3), 241–251.
Kirtis, A. K., & Karahanb, F. (2011). To be or not to be in social media arena as the most cost-e?cient
marketing strategy after the global recession. Procedia Social and Behavioral Sciences, 24, 260–268.
120 Bibliography
Klassen, A. (2009). Forget twitter; your best online marketing tool is the humble product review. Ad Age, 1.
Kozinets, R. V., de Valck, K., Wojnicki, A. C., & Wilner, S. J. (2010). Networked narratives: Understanding
word-of-mouth marketing in online communities. Journal of Marketing, 74(2), 71–89.
KPMG. (2009). Does your business intelligence tell you the whole story?
Krippendor?, K. (2004). Content analysis: An introduction to its methodology. (2nd, Ed.). Thousand Oaks,
CA: Sage.
Laroche, M., Habibi, M. R., & Richard, M.-O. (2012). To be or not to be in social media: How brand loyalty
is a?ected by social media? International Journal of Information Management(0), -.
Lee, C.-H. (2012). Unsupervised and supervised learning to evaluate event relatedness based on content mining
from social-media streams. Expert Systems with Applications, 39, 13338–13356.
Lonnqvist, A., & Pirttimaki, V. (2006). The measurement of business intelligence. Information Systems
Management, 23(1), 32–40.
Lovejoy, K., Waters, R., & Saxton, G. D. (2012). Engaging stakeholders through twitter: How nonpro?t
organizations are getting more out of 140 characters or less. Public Relations Review, 38, 313–318.
Marwick, A. E., & Boyd, D. (2011). I tweet honestly, i tweet passionately: Twitter users, context collapse, and
the imagined audience. New Media & Society, 13(1), 114–133.
McCorkindale, T. (2010). Can you see the writing on my wall? a content analysis of the fortune 50’s facebook
social networking sites. Public Relations, 4, 1–13.
Molloy, J. C. (2011). The open knowledge foundation: Open data means better science. Public Libary of
Science, 9(12).
Murdough, C. (2009). Social media measurement: It’s not impossible. Journal of Interactive Advertising, 10,
94–99.
Murugesan, S. (2007). Understanding web 2.0. IT Professional , 8, 34–41.
Neely, A., & Adams, C. (2005). Performance prism. In E. in Chief: Kimberly Kempf-Leonard (Ed.), Encyclopedia
of social measurement (pp. 41–48). New York: Elsevier.
Neely, A., Adams, C., & Crowe, P. (2001). The performance prism in practice. In Measuring business excellence
(Vol. 5, pp. 6–12). MCB University Press.
Neely, A., Bourne, B., & Kennerley, M. (2000). Performance measurement system design: Developing and
testing a process-based approach. International Journal of Operations and Production Management, 20,
1119–1145.
Neely, A., Gregory, M., & Platts, K. (1995). Performance measurement system dsign: A literature review and
research agenda. International Journal of Operations and Production Management, 15, 80 – 116.
Netbase. (2012). Website. Retrieved September 26, 2012, from http://www.netbase.com/our-network/
partners/sap-social-media-analytics/
Neuendorf, K. A. (2002). The content analysis guidebook. Thousand Oaks, CA: Sage.
O’reilly, T. (2007). What is web 2.0: Design patterns and business models for the next generation of software.
Communications & Strategies, 65, 17–37.
Osterwalder, A., & Pigneur, Y. (2010). Business model generation. Hoboken, New Jersey: John Wiley & Sons,
Ltd.
Patterson, A. (2012). Social-networkers of the world, unite and take over: A meta-introspective perspective on
the facebook brand. Journal of Business Research, 65(4), 527–534.
Peterson, R. A., & Wilson, W. R. (1992). Measuring customer satisfaction: Fact and artifact. Journal of the
Academy of Marketing Science, 20, 61–71.
Pirttimäki, V., & Hannula, M. (2003). Process models of business intelligence. Frontiers of E-Business Research,
250–260.
PWC. (2007). Guide to key performance indicators. communicating the measures that matter. Price Waterhouse
Coopers. (White Paper).
Radian6. (2012). Website. Retrieved September 26, 2012, from http://www.radian6.com/
Ramakrishnan, T., Jones, M. C., & Sidorova, A. (2012). Factors in?uencing business intelligence (bi) data
collection strategies: An empirical investigation. Decision Support Systems, 52(2), 486–496.
Ratner, B. (2003). Statistical modeling and analysis for database marketing: e?ective techniques for mining big
data. Chapman & Hall / CRC Press.
Reinhold, O., & Alt, R. (2011). Analytical social crm: Concept and tool support. BLED 2011 Proceedings,
Paper 50.
Rouibah, K., & Ould-ali, S. (2002). Puzzle: a concept and prototype for linking business intelligence to business
strategy. The Journal of Strategic Information Systems, 11(2), 133–152.
Shahin, A., & Ali Mahbod, M. (2007). Prioritization of key performance indicators: An integration of analytical
hierarchy process and goal setting. International Journal of Productivity and Performance Management,
Bibliography 121
56, 226 – 240.
Sloan, P. (2012). Facebook earnings match estimates; 955 million users. Retrieved August 15,
2012, from http://news.cnet.com/8301-1023_3-57480865-93/facebook-earnings-match-estimates
-955-million-users/
Smith, A. N., Fischer, E., & Yongjian, C. (2012). How does brand-related user-generated content di?er across
youtube, facebook, and twitter? Journal of Interactive Marketing, 26(2), 102–113.
Smith, R. (2006). Key performance indicators. the basics of maintenance and reliability. Chapter 6.
Stephens, W. (2012). Research and social media: Why & how to do a content analysis. Retrieved
July 20, 2012, from http://www.accountplanningblog.com/2012/05/research-and-social-media
-why-how-to-do.html
Takac, C., Hinz, O., & Spann, M. (2011). The social embeddedness of decision making: opportunities and
challenges. Electronic Markets, 21, 185-195.
Tirunillai, S., & Tellis, G. (2012). Does chatter really matter? dynamics of user-generated content and stock
performance. Marketing Science, 31, 198–215.
Tsai, Y., & Cheng, Y. (2012). Analyzing key performance indicators (kpis) for e-commerce and internet
marketing of elderly products: A review. Archives of Gerontology and Geriatrics, 55(1), 126 - 132.
Turban, E., Lee, J., King, D., & Chung, H. (1999). Electronic commerce: A managerial perspective.
Prentice-Hall, Inc., New Jersey, USA.
Twitter. (2012). Twitter turns six. Retrieved June 17, 2012, from http://blog.twitter.com/2012/03/
twitter-turns-six.html
Van Beek, D. (2006). De intelligente organisatie. prestatieverbetering en organisatie-ontwikkeling met business
intelligence. Tutein Nolthenius: ’s Hertogenbosch.
Watson, H., & Wixom, B. (2007). The current state of business intelligence. Computer, 40(9), 96–99.
Weinberg, B. D., & Pehlivan, E. (2011). Social spending: Managing the social media mix. Business Horizons,
54(3), 275–282.
Wikipedia. (2012). Wikipedia. Retrieved September 13, 2012, from http://en.wikipedia.org/wiki/
Wikipedia
Wisner, J., & Fawcett, S. (1991). Linking ?rm strategy to operating decisions through performance
measurement. Production and Inventory Management Journal , Third Quarter, 5 – 11.
You, W., Xia, M., Liu, L., & Liu, D. (2012). Customer knowledge discovery from online reviews. Electronic
Markets, 1–12.
Zeng, D., Chen, H., Lusch, R., & Li, S.-H. (2010). Social media analytics and intelligence. Intelligent Systems,
IEEE, 25(6), 13–16.
Zhang, M., Jansen, B. J., & Chowdhury, A. (2011). Business engagement on twitter: a path analysis. Electronic
Markets, 21(3), 161–175.
122 Bibliography
doc_586858168.pdf
Social Business Intelligence How and where firms can use social media data for performance measurement, an exploratory study
Faculty of Technology, Policy and Management
Social Business Intelligence
How and where ?rms can use social media data for
performance measurement, an exploratory study
Final
Joeri Heijnen
M
a
s
t
e
r
T
h
e
s
i
s
Social Business Intelligence
How and where ?rms can use social media data for performance
measurement, an exploratory study
Master Thesis
Empty
Joeri Heijnen
[email protected]
1320319
December, 2012
Faculty of Technology, Policy and Management · Delft University of Technology
The work in this thesis was supported by KPMG Advisory N.V. Their cooperation is hereby gratefully
acknowledged.
Copyright c Faculty of Technology, Policy and Management
All rights reserved.
Graduation Committee
Formal Chair First Supervisor
prof. dr. Y.H. (Yao-Hua) Tan dr. ir. G.A. (Mark) de Reuver
Full professor Assistant professor
Delft University of Technology Delft University of Technology
Faculty of Technology, Policy and Management Faculty of Technology, Policy and Management
Section Information- and Communication Technology Section Information- and Communication Technology
Second Supervisor External Supervisor
dr. M.E. (Martijn) Warnier ir. M.H. (Han) Horlings AITAP
Assistant professor Manager Business Intelligence
Delft University of Technology KPMG Advisory N.V.
Faculty of Technology, Policy and Management IT Advisory
Section Systems Engineering Business Intelligence
Abstract
Introduction
Both for individuals and for organisations the ?rst decade of the 21
st
century is characterised by the social
media trend. Social media platforms are increasingly popular, and are amongst others used by individuals
to express their opinions. Also ?rms acknowledge the opportunities o?ered by social media and are therefore
increasingly pursuing to realise their goals through means of social media (Murdough, 2009). The value of the
data produced on these platforms lies in the fact that consumers – i.e. (potential) clients – produce these data.
In addition, the information is created instantly, real-time and by many people. It is therefore not surprisingly
that Dey and Haque (2008) state that data generated from online communication acts as “potential gold mines”
for discovering knowledge.
Next, ?rms are increasingly hungry for information that reveals underlying trends and dependencies a?ecting
the ?rm’s performance. Business intelligence systems are used to obtain such insights (Lonnqvist & Pirttimaki,
2006). The demand for (real-time) business intelligence systems and the popularity of social media o?er room for
synthesis. Systems that are purposed to derive actionable information from social media to support managerial
decision-making are referred to as social business intelligence systems. Thus far, business intelligence systems
particularly derive management information from internal data. With the rise of a new data source – social
media platforms – the question rises how a ?rm should process these external data, what kind of managerial
information could be derived from the new data sources, and whether or not each ?rm is able to apply social
business intelligence. In business intelligence, indicators representing the strategy of a ?rm are established.
These indicators are termed ‘key-performance indicators’. Consequently, data re?ecting the performance of
di?erent processes are linked to these key-performance indicators.
Whereas links between social media data and key-performance indicators may leverage the opportunities of social
media for ?rms, a fundamental prerequisite allowing social business intelligence is the existence of user-generated
social media content. After all, user-generated content that does not exist can not be analysed. Thus, an
organisation is dependent for the generation of content on social media users and needs to determine whether
social media data exists before considering to invest in social business intelligence systems. So far, it is not clear
which organisational characteristics a?ect the existence of social media content. In this research, two general
characteristics describing a ?rm are used to investigate the existence of social media data; (i) industry type and
(ii) customer relation type.
Research Objective
On the one hand social media is a new phenomenon and acknowledged as a source of data of which valuable
information can be derived. On the other hand, it is unclear which ?rms are able to collect social media data
that is related to their ?rm and how ?rms should process these new data in accordance with existing business
intelligence processes. Therefore, the objective of this research has been formulated as:
The objective of this research is to develop a procedure to utilise social media data for business
intelligence, for which the applicability is investigated for ?rms in di?erent industries and for
di?erent customer relations.
Method
Our sample consists of social media messages related to eighteen di?erent ?rms, in seven di?erent industries
performing di?erent customer relations. Because the sample ?rms operate in di?erent industries and execute
iii
di?erent customer relations, it is possible to gain insight in potential di?erences between the social media
messages related to these ?rms. During a period of two weeks, social media messages from various platforms
have been crawled into a local database to allow further analyses. The content in the dataset is sourced from
Twitter, Facebook public pages, Flickr, Newssites, Google+ public pages, (Wordpress) Blogs, Picasa, YouTube
and Friendfeed. These platforms are popular in Western Europe.
To gain insight in the amount of ?rm-related social media messages, the average daily mentions of ?rms served
as a proxy to compare the volume of messages related to di?erent ?rms. Next, using a content analysis, a
portion of the collected messages have manually been classi?ed into di?erent categories based on the messages’
subjects. These categories correspond with generally applied categories of key-performance indicators. As such,
the results of the content analysis are directly linked to ?rms’ key-performance indicators, allowing to draw
conclusions on the relatedness of social media messages to di?erent key-performance indicators.
Incorporating the new external data source requires traditional business intelligence systems to be adjusted. A
social business intelligence procedure should be consistent with these traditional systems, and should additionally
consider the challenges involved when processing social media data. As such, the requirements for a social
business intelligence procedure have been established based on generally applied business intelligence concepts.
Furthermore, the challenges involved in the processing of social media data are discovered by the collection
of social media messages for the content analysis. Based on the traditional BI concepts and the challenges
discovered in the content analysis, a business intelligence procedure is developed. The procedure is veri?ed by
analysing its consistency with existing BI systems and its ability to solve the issues emerging when processing
social media data.
Results
The results of this research are twofold. Firstly, we gained insight in the applicability of social business
intelligence by investigating the existence and content of ?rm-related social media messages. Secondly,
a procedure to collect, process and analyse social media data for business intelligence purposes has been
established.
(ii) Applicability of social business intelligence
The applicability of social business is investigated on two facets. Firstly, the volume of ?rm-related social media
messages is investigated to obtain insight in the amount of data that is available for ?rms. The volume of
?rm-related social media content is however not su?cient to draw conclusions on the applicability of social
business intelligence. Therefore, the second facet on which the sample data is analysed relates to the content
of the social media messages. Especially, the subjects of the messages were analysed.
Volume
The average daily mentions di?ers from ?rm to ?rm. This implies that the applicability of social business
intelligence will not be possible for all ?rms, since not for each ?rm data is generated. Figure 1 illustrates the
average daily mentions of di?erent ?rms in our sample.
Heineken
Coca-Cola
Philips
3.000
3.500
TomTom
KLM
2.500
x
1
/d
a
y
]
1.500
2.000
A
v
e
r
a
g
e
D
a
ily
M
e
n
t
io
n
s
[
x
1
/d
a
y
Bol.com NS
Albert Heijn
C-1000
ABN AMRO Arcelor
Mittal
500
1.000
A
v
e
r
Bol.com NS
PostNL
Blokker
Aegon Unibail
Rodamco
AkzoNobel
Arcelor
Mittal
Arcadis Fugro
0
500
B2B B2C
Figure 1: Average Daily Mentions of Firms, Clustered per Customer Relation Type
Figure 1 shows the daily volume of ?rm-related social media content, in which the ?rms are clustered on their
customer relation type and consequently ordered descending. This ?gure suggests that B2C ?rms – coloured
iv Abstract
in red – are more likely to ?nd social media content that is related to their ?rm than ?rms performing B2B
relations (coloured in blue).
The second dimension on which the volume of ?rm-related social media content is investigated relates to
industries. Our sample consists of eighteen di?erent ?rms active in seven di?erent industries. As a ?rst step to
identify possible di?erences in the volume of daily messages between industries, the ?rms have been clustered
on industry type in ?gure 2, and have consequently been sorted in descending order.
Heineken
Coca-Cola
Philips
3.000
3.500
TomTom
KLM
2.500
x
1
/d
a
y
]
1.500
2.000
A
v
e
r
a
g
e
D
a
ily
M
e
n
t
io
n
s
[
x
1
/d
a
y
Albert Heijn
C-1000
ABN AMRO
NS Bol.com
Arcelor
Mittal
500
1.000
A
v
e
r
NS Bol.com
Arcelor
Mittal
Blokker
Aegon PostNL
AkzoNobel
Unibail
Rodamco Arcadis Fugro
0
500
Information&
Communication
Industry Transport &Storage Wholesale &Retail Financial Institutions Mining&Quarrying
Consultancy, Research
&Other Specialised
Business Services Communication Business Services
Figure 2: Average Daily Mentions of Firms, Clustered per Industry Type
Figure 2 strongly suggests that there exists a di?erence in the amount of user-generated content between di?erent
industries, with industrial ?rms being highly mentioned on social media, while consulting ?rms are the least
mentioned.
Subjects
Next to an assessment of the amount of social media posts that are created on the web, this thesis examined
the subjects of the social media posts in order to link the messages to ?rms’ key-performance indicators. The
social media messages of the ?rms have been classi?ed into categories based on their subject. These categories
are based on ten categories of commonly applied key-performance indicators. Consequently, the collected social
media posts of the ?rms in the sample have manually been classi?ed into one of these categories.
Our analysis shows that the subjects of social media messages di?er from ?rm to ?rm. The majority of social
media messages related to ?rms (41%) express how the external stakeholders of a ?rm perceive the company. In
this thesis, such posts have been classi?ed as community posts. 18% of the social media messages in our dataset
contained the name of a ?rm, but did not contain any valuable information for the ?rm and have consequently
been assigned as unde?ned posts. About 11% of the social media messages relate to ?nancial results, which
consist of ?nancial performance discussions (5%) and stock related discussions (6%).
The content analysis of this research suggests that the subjects of social media messages related to B2B
?rms contain a higher percentage of short term ?nancial results, news and professionals related messages than
messages related to B2C ?rms. Unfortunately for B2B ?rms, such type of information is yet available internally.
Acquiring social media data to gain additional management information is therefore of less value for B2B ?rms.
Next, the analysis indicates that the social media messages related to B2C ?rms contain a higher percentage
of posts related to customer relations, product and service quality and product and service innovation than
messages related to B2B ?rms. It are these types of information that deliver additional value to the ?rm, since
this information is not available at ?rms internally.
In addition, the content analysis of this research suggests that the subjects of social media posts di?er between
industries, but that the majority of the subjects in each industry relates to community, i.e. social media posts
revealing how the community perceives the company. The results indicate that ?rms active in the information
& communication, ?nancial institutions and transport & storage industries are more subjected to social media
messages related to customer relations, while ?rms active in the mining and quarrying and consulting industries
will ?nd messages related to ?nancial performance.
v
(ii) Procedure for social business intelligence
Based on (i) traditional business intelligence frameworks and (ii) the experience we gained in collecting,
processing and analysing social media data in the content analysis, a social business intelligence (“SBI”)
procedure has been developed. Figure 3 schematically shows the social business intelligence procedure.
Reaction based on
social intelligence
Strategic
mapping of
KPIs
Reacting
Search terms
Action plan(s) to
respond to gained
intelligence
g
Collecting
Mapping
insights to
business
units
Data pre
Unstructured
data
Information for
business units
Data pre-
processing
Categorising
Analysing
Structured, combined
(and anonymised) data
Categorised data
Figure 3: Blueprint: Social Business Intelligence Procedure
Our SBI procedure consists of seven main components, being (i) strategic mapping of KPIs, (ii) collecting,
(iii) data pre-processing, (iv) categorising, (v) analysing, (vi) mapping insights to the business units, and (vii)
reacting. The seven steps can be interpreted as a cycle, i.e. the output of the last step in?uences the ?rst step.
The very ?rst step of social business intelligence sets the scene for the objects that are to be collected and
analysed. Namely, in the ?rst step the key-performance indicators that are to be measured by social media data
are selected. Not each type of KPI is to be measured by social media data since there does simply not exist
any related social media data to these types of KPIs. Firms should mainly focus on KPIs related to customer
relations, public image and – to a less extent – on product and service innovation when selecting KPIs that are
to be measured using social media data.
The second step of the SBI procedure relates to data collection. In contradiction to regular BI systems, the
data is to be sourced from external parties in social business intelligence. People create ?rm-related messages
on di?erent platforms, of which the vast majority of publicly accessible messages are created on Twitter. The
search terms that are used to ?lter out the content at which the ?rm is interested should be based on the social
KPIs selected in the previous step.
The social media data has been collected from multiple platforms which adhere to their own data format. The
di?erent format are to be combined into one uniform database, so that – in a later step – data analysis can be
applied on the complete dataset. Furthermore, the ?rm should select those attributes that are necessary for the
analysis, not each platform o?ers the same richness of attributes to a social media post. In addition, the data
should be anonymised to be in compliance with new Regulations regarding data privacy. Finally, spam – i.e.
social media posts that do not relate to the ?rm – should be removed from the collected data.
The data pre-processing step resulted in a structured database in which the social media messages from multiple
platforms are combined. In the categorising step, the messages are clustered on di?erent issues of interest,
depending on the ?rm’s subject of interest. E.g., messages related to certain products can be categorised, or
one can cluster the messages that are created by people with many followers, etc. Again, the criteria at which
the messages are categorised are determined by the selection of the social KPIs in the ?rst step.
So far, the collected data has not provided any insights. It is in this analysis step of the procedure where data
is transformed into information. The categories that were established in the previous step are analysed in this
step. For instance, sentiment analysis can be applied on the categories related to the ?rm’s products in order
to acquire intelligence related to customer experiences of the products. However, the most valuable intelligence
vi Abstract
is gained when social media data is related to internal data. For instance, the volume of social media messages
related to a certain product may be correlated with the sales volume of that product. It is in this phase of the
SBI procedure where such relations are explored.
In the ?rst step of the procedure, KPIs have been selected. These KPIs typically relate to a certain function of
the ?rm, and hence have an “owner”. The intelligence gained in the previous step relates to KPIs, and should
feed back to the owner of the KPI. Generally, it are the people in the ?rm that are responsible for the KPI who
are the ones that can reason how the KPI is in?uenced. Therefore, these people are the ones that can draft an
action plan in case the KPI needs improvement.
The ?nal step of the social intelligence procedures consists of the execution of the action plans that are developed
in collaboration with people from the business lines that are responsible for the respective KPIs. Actions on
the gained intelligence may involve revisions of internal processes or strategies, or external interventions such
as social media engagement.
Contents
Graduation Committee i
Abstract ii
Preface xi
1 Research Problem 1
1-1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1-2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1-3 Research Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1-4 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1-5 Coherence of Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1-6 Research Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1-6-1 Exploratory Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1-6-2 Description of Research Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1-6-3 Data Collection and Research Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1-7 Project Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1-8 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1-9 Scienti?c Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1-10 Societal Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1-11 Project Deliverable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2 Conceptual Frame of Research 15
2-1 Business Intelligence Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2-2 Registering the Right Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2-2-1 Strategy and Business Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2-2-2 Frameworks Supporting the Formulation of Performance Indicators . . . . . . . . . . . . . . . 17
2-2-3 Performance Measurement System Design Process . . . . . . . . . . . . . . . . . . . . . . . 20
2-2-4 Typology of Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2-2-5 KPI Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2-3 Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2-4 Sub Conclusion: How Business Intelligence is Applied . . . . . . . . . . . . . . . . . . . . . . . . . . 25
viii Contents
3 Research Domain 27
3-1 Web 2.0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3-2 Social Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3-2-1 Social Media Platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3-2-2 User-Generated Content on Social Media Platforms . . . . . . . . . . . . . . . . . . . . . . . 31
3-2-3 Current Applications of Social Media in Firms . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3-3 Social Business Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3-3-1 Current State of Social Business Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3-4 EU Legislation on Social Media Data Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3-4-1 What Firms are allowed to do with Public Data . . . . . . . . . . . . . . . . . . . . . . . . . 40
3-5 Sub Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4 Content Analysis 42
4-1 Theoretical Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4-1-1 Hypotheses Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4-1-2 Material to Investigate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4-2 Establishment of Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4-2-1 Operationalising the Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4-2-2 Determining the Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4-2-3 Description of the Measuring Period . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4-3 Pretest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4-3-1 Categories of Social Media Posts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4-3-2 Revised Taxonomy of Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4-4 Data Collection and Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4-4-1 Search Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4-4-2 Scraping Social Media Content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4-4-3 Data Cleaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4-5 Descriptive Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4-5-1 Channel Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4-5-2 Volume of Firm-Related Social Media Messages . . . . . . . . . . . . . . . . . . . . . . . . . 62
4-5-3 Subjects of Social Media Posts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4-6 Interpretation of the Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4-6-1 Volume of Social Media Posts related to Firms . . . . . . . . . . . . . . . . . . . . . . . . . 68
4-6-2 Subjects of Social Media Posts related to Firms . . . . . . . . . . . . . . . . . . . . . . . . . 69
4-7 Sub Conclusion: Social Media Posts that relate to KPI Categories and the Performance Prism
Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5 Blueprint of a Social Business Intelligence Procedure 72
5-1 Requirements Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5-1-1 Description of Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5-1-2 Requirements Check on Business Intelligence Concepts . . . . . . . . . . . . . . . . . . . . . 76
5-2 Social Business Intelligence Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5-2-1 Strategic mapping of KPIs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5-2-2 Collecting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5-2-3 Data Pre-Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5-2-4 Categorising . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5-2-5 Analysing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5-2-6 Mapping insights to Business Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5-2-7 Reacting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5-3 Veri?cation of Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5-4 Real-Time Social Business Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5-5 Social Business Intelligence versus Business Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . 85
5-6 Sub Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Contents ix
6 Conclusions & Discussion 90
6-1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6-2 Contributions to Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6-2-1 Methodological Innovation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
6-3 Implications for Practice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6-4 Re?ection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6-4-1 Twitter Scraper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6-4-2 If I had More Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6-4-3 Stepwise Description of Data Collection Process . . . . . . . . . . . . . . . . . . . . . . . . . 96
6-5 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6-6 Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6-6-1 Classi?er . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6-6-2 Social Media Posts Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6-6-3 The Real Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6-6-4 Case Study: Relations of Social Media Metrics and Key-Performance Indicators . . . . . . . . 98
A Performance Prism Perspectives and Key-Performance Indicators Categories 99
B Classi?cation of Social Media Posts 101
C Social Media Platform Distribution 108
D Descriptive Statistics of Social Media Post Categories 112
E Corporate Engagement 116
Bibliography 118
x Contents
Preface
Social media is a trend in the ?rst decade of this century, and the concept is increasingly incorporated in the
daily lives of people. Scepticism towards the new technology is losing support, and companies are aware that
the new trend cannot be denied. Though the new phenomenon is gaining attention in the scienti?c world, social
media was not yet part of the curriculum at my faculty. I am grateful that I was o?ered the opportunity and the
con?dence to dive in the rather unexplored world of research into social media, and explore the opportunities
for companies o?ered by social business intelligence.
First of all I would like to thank my graduation committee. My ?rst supervisor of Delft University of Technology,
Mark de Reuver, critically reviewed my work on a regular basis. I hereby thank Mark for his constructive
comments and suggestions for improvements, I experienced our meetings as pleasant and useful. Mark’s
experience in scienti?c research and knowledge of ICT and business models contributed to the quality of my
thesis. Martijn Warnier supervised my work as second supervisor from the Systems Engineering section. Martijn
indicated issues concerned with social media (data) that I did not think of in the ?rst place, for which I am
grateful. Harry Bouwman chaired my committee as professor from the ICT section. Harry contributed to this
thesis by critically reviewing my work and o?ering suggestions for improvements, which were mainly related to
scienti?c concepts. Thank you for these comments, the critical notions improved the level of this work.
I would like to express my gratitude to my supervisor at KPMG, Han Horlings. Thanks to weekly meetings with
Han I was driven to progress my thesis. I found a sparring partner to discuss especially business intelligence
related aspects of my work. Han, thanks for your time, contributions and coaching! In addition I would like to
thank all employees and co-interns of KPMG’s Business Intelligence department for their interest in my thesis,
their contributing opinions and ideas on the subject, for being challenging competitors during the karting event
and for the fun at the Amsterdam Parade last summer.
Amstelveen, December 2012
Joeri Heijnen
xii Preface
Chapter 1
Research Problem
1-1 Introduction
More and more, customers are using content sharing sites to express their opinions about almost anything, from
soccer matches to ?nancial statements of large corporations. Examples of platforms where these expressions
are shared to the world are blogs and forums, social network sites and wikis. In 2008, “75% of internet surfers
used social media” (A. M. Kaplan & Haenlein, 2010), and the usage of social media is not limited to teenagers.
Members of generation X, now 35–44 years old, are increasingly active on social media sites (A. M. Kaplan &
Haenlein, 2010). Anno 2012, people express how they feel, what they do, what they think of, and what they
intend to do in over 340 million daily Twitter posts (Twitter, 2012). The value of the information produced
on these platforms lies in the fact that consumers produce these data. In addition, the information is created
instantly, real-time and by many people. Since social media posts are often non-anonymous and directly linked
to a person, ?rm or brand, the content produced on social media platforms can be interpreted as an indicator
of people’s attitude towards a ?rm, product or service. The user-generated content is considered as a driver for
future sales by Dhar and Chang (2009), hence containing economic value for ?rms (Ghose & Panagiotos, 2010).
In the ?rst decade of the 21
st
century, business intelligence (“BI”) has evolved to one of the critical processes
for organisations to provide useful insight, to support decision-making, and to drive organisational performance
(Ramakrishnan, Jones, & Sidorova, 2012). According to Watson and Wixom (2007) BI has become a “strategic
initiative and is regarded as an instrument in driving business e?ectiveness and innovation”. For organisations,
it is increasingly important to quickly respond to changes in the environment (Gessner & Volonio, 2005).
Therefore, BI systems are required to contain a component that allows monitoring the real-time environment.
We de?ne such systems as ‘real-time BI’ systems.
From the above, we can derive two trends in the current business landscape:
(i) an increase in the usage of social media, and,
(ii) an increase in the usage of business intelligence systems.
Trend (i): Social Media in Organisations Organisations are increasingly pursuing to realise their goals
through social media (Murdough, 2009). Social media applications support organisations in creating value
in many of their activities, e.g. in marketing, services, human resource management and customer relationship
management (A. N. Smith, Fischer, & Yongjian, 2012). In addition, ?rms are able to acquire data from social
media at low costs. Dey and Haque (2008) state that data generated from online communication acts as
“potential gold mines” for discovering knowledge. It is therefore that this thesis focuses on the extraction of
information based on the data created by consumers on social media.
The increased application of social media has serious consequences for an organisation’s exposure to the actors in
their environment, which include (potential) customers, suppliers and competitors. It seems that the power has
been taken from the corporate marketing departments by individual consumers that create, share and discuss
online blogs, tweets, Facebook entries, movies, pictures, etc. (Kietzmann, Hermkens, McCarthy, & Silvestre,
2011). With or without permission from the organisation, communication about brands will happen. In an
2 Research Problem
environment where customers gain more and more power, an organisation needs to carefully treat its actions and
control its exposure. Therefore, companies empower employees to talk, listen, and respond to what consumers
post on social media (A. N. Smith et al., 2012).
Though many organisations acknowledge the opportunities in the application of social media, there also exists
a fair degree of uncertainty with respect to allocating marketing e?ort and budget to social media, and “limited
understanding” of the social media platforms (Weinberg & Pehlivan, 2011). Kietzmann et al. (2011) argue that
many executives avoid or ignore social media because they do not understand what it is, how to engage with it
and learn from it. This thesis contributes to a further understanding of social media and discovers opportunities
to leverage the valuable content on these platforms for business purposes.
Trend (ii): Business Intelligence in Organisations Business intelligence systems are applied to obtain a better
understanding of underlying trends and dependencies – often coming from the external context – that a?ect the
business (Lonnqvist & Pirttimaki, 2006). Whereas BI systems were initially perceived as tools that were used
exclusively to support strategic decision-making, organisations have recently commenced to further exploit the
capabilities of BI systems to support wider business activities (Elbashir, Collier, & Davern, 2008).
The scale of recent investments in BI systems re?ects the growing importance and highlights the need for more
attention in research studies. Elbashir et al. (2008) estimated that global spending on BI systems and related
products reached USD 6.1 billion in 2008. A paper by Gartner (2009) predicted that organisations will increase
spending on “packaged analytic applications, including corporate performance management (“CPM”), online
marketing analytics that optimise processes, not just report on them”. Azvine, Cui, and Nauck (2005) predict
that in the future, “business intelligence will be available to everyone in the enterprise, and will be embedded
in many business systems”.
1-2 Problem Statement
The demand for (real-time) business intelligence and the popularity of social media o?er room for synthesis.
The opportunities o?ered by linking both concepts are acknowledged in the literature, e.g by Dey and Haque
(2008) and Lovejoy, Waters, and Saxton (2012). However, search queries
1
related to the subject of this thesis
into the scienti?c databases ScienceDirect and JStore, and the search engine Google Scholar resulted in the
understanding that social media applications for BI purposes are relatively underexposed in the literature.
Generally, research in the area of social media is related to marketing activities, sales, promotions, public
relations and customer relationship management, e.g. by Dong-Hun (2010); Ratner (2003); Klassen (2009);
Kozinets, de Valck, Wojnicki, and Wilner (2010); Kirtis and Karahanb (2011); Hanna, Rohm, and Crittenden
(2011); A. M. Kaplan and Haenlein (2012); You, Xia, Liu, and Liu (2012). The focus of the research conducted
in the literature is mainly focused on the organisation expressing itself to the outside (social media) world,
whereas this thesis focuses on the incoming aspect. A reason for the shallow results discovered in the literature
may be the relatively new character of combining social media and business intelligence.
Zeng, Chen, Lusch, and Li (2010) distinguish social media research between social media analysis and social
media intelligence. Social media analysis is concerned with “developing and evaluating informatic tools and
frameworks to collect, monitor, analyse, summarise, and visualise social media data”. Social media intelligence –
on the other hand – “aims to derive actionable information from social media in context-rich application settings,
develop corresponding decision-making or decision-aiding frameworks, and provide architectural designs and
solution frameworks for existing and new applications that can bene?t from the wisdom of crowds through the
web”.
Many social media monitoring tools, like Socialmention.com, Radian6, RowFeeder, Trackur, uberVU, SAS
Social Media Analytics, Finchline, Sprout Social, etc. mainly reveal the performance of a ?rm on social media
(number of mentions, number of likes, % of positive mentions), and treat the social media component of a ?rm
as a separate business unit executing its own strategy. However, the purpose of business intelligence is to reveal
the underlying parameters that determine the performance of the organisation, that is, not limited to solely
social media performance. In order to understand the in?uence of social media content on a ?rm’s performance,
a link between the company’s key performance indicators (“KPIs”) and social media parameters is required
because KPIs measure the performance of an organisation with respect to its strategy. Some social media
1
(SOCIAL BUSINESS INTELLIGENCE), (BUSINESS INTELLIGENCE 3.0), (SOCIAL MEDIA) AND (ORGANISATION),
(SOCIAL MEDIA) AND (BUSINESS), (SOCIAL MEDIA) AND (BUSINESS INTELLIGENCE), (TWITTER) AND (BUSINESS
INTELLIGENCE), (WEB 2.0) AND (BUSINESS INTELLIGENCE), (SOCIAL MEDIA) AND (STRATEGY)
1-3 Research Objective 3
monitoring tools, like Kapow Software and ListenLogic seem – at a glance – to establish this link. Zeng et al.
(2010) highlight the need for clearly de?ned social media performance measures because much of the research
is conducted in a setting which aims to support decisions in organisations. We argue that the possibilities of
social media for business intelligence purposes reaches further than what is currently o?ered by the social media
analytics tools. This argument is supported by Reinhold and Alt (2011), who state that “existing tools still have
a limited functional scope”. The key bene?ts will be gained whenever the KPIs of an organisation are linked to
the parameters that are measured by social media tools. Only in that case, one can speak about ‘social business
intelligence’. This thesis contributes to a transition from social media ‘monitoring’ towards social ‘business
intelligence’.
Whereas links between organisational performance and social media content can leverage the opportunities
of social media for ?rms, a fundamental prerequisite allowing social business intelligence is the existence of
user-generated social media content. After all, user-generated content that does not exist can not be analysed.
Thus, an organisation is dependent for the generation of content on social media users and needs to determine
whether social media data exists before considering to invest in social business intelligence systems. However,
it is not clear which organisational characteristics a?ect the existence of social media content. The following
section illustrates which factors are to be considered when one tries to categorise the availability of social media
content that is related to ?rms.
Firstly, it is likely that within some industries users express their opinions more often than in other industries.
We expect that one expresses his or her opinion more often about a product that is purchased on a frequent
basis. For example, domestic products are purchased more frequent than a car or a house. Therefore, the
consumer industry is probably discussed more often than the real-estate market. Secondly, the relation with
end-users makes it that people discuss the company on social media, or not. Some ?rms are more visible
for consumers than others. Zhang, Jansen, and Chowdhury (2011) support this factor by concluding that
“business engagement on social media relates directly to consumer’s engagement with online word-of-mouth
communication”. When users experience malfunctions in a mobile network they complain at the ?rm at which
they signed the contract, while the ?rm that delivered the network equipment – which may be responsible for the
errors – remains una?ected. This example illustrates that it is necessary to make a distinction between companies
in the same industry based on their position regarding consumers. Turban, Lee, King, and Chung (1999) classify
e-commerce into either business-to-business (“B2B”), business-to-consumer (“B2C”), consumer-to-consumer
(“C2C”), consumer-to-business (“C2B”), non-business e-commerce, or intra-business e-commerce” (as cited in
Chen, Jeng, Lee, and Chuang (2008)). We will use this classi?cation to assign an organisation’s position
regarding consumers since it clearly illustrates how close an organisation acts to the end consumer. As such the
network service provider can be positioned as a B2C ?rm, while the provider of the equipment performs B2B
relations.
Next, in the case that there exists social media content, an organisation should be able collect and analyse the
data. The unstructured nature of the data, various languages, various data formats, interpretation di?culties,
unveri?ed information and privacy issues are aspects that make the usage of social media data for business
intelligence di?erent from ‘regular’ – i.e. internal management information – BI data.
Knowledge Gap From the previous, we can conclude the following. It is unclear in which industries and for
which type of customer relations ?rms can apply social media data for business intelligence. Secondly, there
is no understanding how organisations should process social media data in relation with business intelligence.
Taking into account the previous, the following knowledge gap is formulated:
It is unclear how ?rms can process social media data for business intelligence, and how the
applicability of social media data for business intelligence varies among di?erent industries and
di?erent customer relation types.
1-3 Research Objective
Social media is a new phenomenon, and increasingly popular for both consumers and organisations. Business
intelligence is applied in organisations to measure organisational performance and to provide managerial
information. The literature agrees that social media posts may contain valuable insights for organisations
that managers can use in their decision-making. Hence, the two concepts o?er room for synthesis. However,
there does not exist a structured procedure that prescribes how organisations should acquire and analyse these
4 Research Problem
social media posts in order to generate managerial information. In addition, it is unknown how (i) di?erent
industries and (ii) di?erent customer relations a?ect the existence of social media data on the web. After
all, if (potential) clients do not generate social media posts related to a ?rm, it will not be possible to derive
information from the posts. Therefore, the objective of this thesis is formulated as:
The objective of this research is to develop a procedure to utilise social media data for business
intelligence, for which the applicability is investigated for ?rms in di?erent industries and for
di?erent relations with end-users.
As such, insight in (i) the suitability of social media for business intelligence for di?erent organisations and (ii)
a procedure prescribing the steps required for social business intelligence is obtained.
Concepts in Research Objective In order to clarify the research objective, the key concepts are listed and
explained below.
• Procedure to utilise social media data for business intelligence
A procedure to utilise social media data for business intelligence prescribes which steps are necessary when
an organisation applies social media data for the measurement of organisational performance. Within business
intelligence procedures, managers endeavour to measure organisational performance based on metrics that re?ect
the performance of organisational activities. Generally, these activities are performed by di?erent departments.
In this thesis we look for performance metrics that are in?uenced by social media data.
• Social media data
Social media data can be quantitative or qualitative in nature. Examples of quantitative social media data are the
number of likes, views or shares of a certain page, the number of followers, friends or retweets through the course
of time. Qualitative social media data contains the text of the posts. In this thesis, we investigate how social
media data can be used for business intelligence.
• Business intel ligence
Business intelligence is a process in which information is derived from data to support decision making. The
acquired information is required to measure organisational performance, at which managers can base their decisions.
Information may for example relate to trends in the level of inventory of a certain product, or the amount of sales
in a certain period.
• Firm contexts
Though there are various ways to de?ne a ?rm’s context, we describe the context of a ?rm based on two dimensions
in this thesis: (i) industry and (ii) relation with end-consumers. We employ this de?nition of context in this thesis
because we are particularly interested in the variations of the applicability of social business intelligence on these
two dimensions. Next, a generic classi?cation of a ?rm’s context on these two dimensions allows the conclusions
of the research to be applicable at a broad range of ?rms.
i. Industry
Organisations can be classi?ed in industries. All organisations in the same industry deliver similar products
/ services. We apply CBS’ (2012) classi?cation to position ?rms in certain industries. Examples of industries
are the telecommunications industry, or the ?nancial industry.
ii. Relation with end-users
Each organisation has di?erent customers. Generally, a distinction between Business-To-Business (“B2B”)
and Business-To-Consumer (“B2C”) is made to described the relation with an organisation’s customer. In
B2C relations, the end-user is part of the relation.
1-4 Research Questions
From the research objective, the following main research question is formulated:
How can ?rms use social media data for business intelligence, taking into account the ?rm’s speci?c
industry and relationship with end-users?
1-5 Coherence of Research Questions 5
In order to describe the domain of this thesis, the ?rst sub question describes the current state of social media,
the role of business intelligence in ?rms and the developments towards social business intelligence. Therefore,
the ?rst sub question is formulated as:
1. What is the current state of social media in relation with business intelligence?
(a) What are social media?
(b) How are social media generally applied within ?rms?
(c) How is business intelligence generally applied within ?rms?
(d) How are key-performance indicators established within ?rms?
(e) How can key-performance indicators be categorised?
(f) What is social business intelligence?
The main research objective contains a component in which we reveal in which contexts – i.e. for which
industries and for which customer relation type – ?rms are able to acquire social media data, and in which not.
This objective follows from the fact that ?rms are dependent on the users of social media whether or not social
media data is available. Therefore, the second sub question investigates for which ?rms social media posts are
available, and to what subjects the posts are related. The subjects of social media posts are consequently used
to assign social media posts to the KPIs of a ?rm. The composition of the second sub question is twofold, sub
questions 2(a) and 2(b) are quantitative in nature and provide insight in the volume of social media posts. On
the other hand, 2(c) and 2(d) are qualitative in nature and provide insight in the content of the social media
posts related to ?rms. The second sub question is formulated as:
2. In which ?rm contexts
2
are ?rms able to acquire social media data for business intelligence?
(a) How does the volume of social media posts related to ?rms vary between di?erent industries?
(b) How does the volume of social media posts related to ?rms vary between di?erent relations with
end-users?
(c) How do subjects of social media posts related to ?rms vary between di?erent industries?
(d) How do subjects of social media posts related to ?rms vary between di?erent relations with
end-users?
Secondly, the research objective contains a component in which we describe how a ?rm can acquire and process
social media data for business intelligence purposes. The third sub question focuses on the development of a
procedure to process social media data so that it can be joined up in business intelligence processes. A key
requirement of this process is that it should ?t within existing business intelligence activities. Therefore, 1(c)
investigates how business intelligence is generally applied in organisations, and will result in requirements for
a procedure in which social media data is applied for business intelligence. As discussed, social media data
di?ers from data that is generally processed in BI systems. Question 3(a) discusses the potential problems and
pitfalls when processing social media data. Consequently, 3(b) provides solutions for these problems. In 3(c),
we determine how social media data can be linked to KPIs. Finally, 3(d) describes how a ?rm can process social
media data while following the generally applied BI approach. The third sub question is de?ned as:
3. Which processes are required to incorporate social media data into general business intelligence
frameworks?
(a) What problems arise when applying social media data for business intelligence?
(b) How can the problems discovered in 3(a) be tackled?
(c) How can social media data be linked to key-performance indicators?
(d) How can social media data be processed in accordance with general business intelligence systems?
1-5 Coherence of Research Questions
Each research question delivers information that is required to answer another question. The coherence of the
research questions is presented in ?gure 1-1. The arrows represent the output of a research question which, in
turn, serve as input to answer an other research question.
2
In this thesis, we de?ne a ?rm context based on the ?rm’s industry and customer relation type.
6 Research Problem
Requirements
for a social
business intelligence
procedure
Main Research Question
How can firms use social media data for business intelligence,
taking into account the firm’s specific industry and relationship with end-users?
Sub Question 2
2. In which contexts are firms able to acquire
social media data for business intelligence?
Sub Question 3
3. Which processes are required to incorporate
social media data into general business
intelligence frameworks?
Procedure prescribing how
to collect, process and analyse
social media data
How does the volume of
social media posts
related to firms vary
between different:
2. (a) industries?
2. (b) relations regarding
end-users?
Quantitative description
of the availability of
social media posts in
different contexts.
Qualitative description
of the subjects of social
media posts in different
contexts.
How do the subjects of
social media posts
related to firms vary
between different:
2. (c) industries?
2. (d) relations regarding
end-users?
1. (c) How is business
intelligence generally
applied within
organisations?
3. (d) How can social
media data be
processed in
accordance with general
business intelligence?
3. (a) What problems
arise when applying
social media data for
business intelligence?
Understanding of how
business intelligence is
applied in firms, and
what the consequences
are when adding
social media data in
this process
3. (b) How can the
problems discovered in
3(a) be tackled?
Understanding of the
pitfalls of social media
data
Solutions for the
pitfalls of processing
and analysing social
media data
Sub Question 1
1. What is the current state of social media in
relation with business intelligence?
Overview of available social media
data in different contexts
Knowledge about current
technologies related to
social business intelligence
3. (c) How can social
media data be linked to
key-performance
indicators?
Understanding of relations
between social media data
and KPIs
1. (a) What are social
media?
1. (b) How are social
media generally applied
within firms?
1. (d) How are key-
performance indicators
established within firms?
Understanding
how business intelligence
is applied in organisations
1. (f) What is social
business intelligence?
Definition of
social business
intelligence
Understanding of
currently exploited
opportunities offered
by social media
Understanding of
generally applied
KPIs in firms
Knowledge about social
media platforms
and the data that is
created on such platforms
1. (e) How can key-
performance indicators
be categorised?
Understanding of the
role of KPIs within
business intelligence
Figure 1-1: Coherence of Research Questions
1-6 Research Method 7
1-6 Research Method
This section describes the type of research (section 1-6-1), the research method (section 1-6-2) and the approach
of the research (section 1-6-3).
1-6-1 Exploratory Research
Exploratory research is conducted for a problem that has not been clearly de?ned. It relies on reviewing
literature and/or data. Often, the results of exploratory research are not usually useful for decision-making by
themselves, but they can provide signi?cant insight into a given situation. The goal is to learn “what is going on
there?”, and to investigate social phenomena without explicit expectations. Mainly, the purposes of exploratory
research are exploratory, descriptive and explanatory in nature. This thesis researches an area that is relatively
unexplored, and of which the functioning is not clearly documented in theories and frameworks. Therefore, this
thesis can be positioned under exploratory research.
1-6-2 Description of Research Methods
The research questions formulated in section 1-4 individually require di?erent research methods in order to be
answered. The research consists of a mix of literature studies, consulting experts, and content analysis on the
acquired data. All methods and the corresponding requirements for data and other resources are discussed in
this section. Table 1-1 schematically lists the corresponding research method for each research question.
Table 1-1: Research Questions versus research Methods
L
i
t
e
r
a
t
u
r
e
r
e
v
i
e
w
C
o
n
t
e
n
t
a
n
a
l
y
s
i
s
C
o
n
s
u
l
t
i
n
g
B
I
e
x
p
e
r
t
s
1. What is the current state of social media in relation with business intelligence?
(a) What are social media?
(b) How are social media generally applied within ?rms?
(c) How is business intelligence generally applied within ?rms?
(d) How are key-performance indicators established within ?rms?
(e) How can key-performance indicators by categorised?
(f) What is social business intelligence?
2. In which contexts are ?rms able to acquire social media data for business intelligence?
(a) How does the volume of social media posts related to ?rms vary between di?erent industries?
(b) How does the volume of social media posts related to ?rms vary between di?erent relations with end-users?
(c) How do subjects of social media posts related to ?rms vary between di?erent industries?
(d) How do subjects of social media posts related to ?rms vary between di?erent relations with end-users?
3. Which processes are required to incorporate social media data into general business
intelligence frameworks?
(a) What problems arise when using social media data for business intelligence?
(b) How can the problems discovered in 3(a) be tackled?
(c) How can social media data be linked to key-performance indicators?
(d) How can social media data be processed in accordance with general business intelligence systems?
Literature Review
Scienti?c articles are studied, mainly in the Journal of Electronic Markets, Journal of Information Systems
Management, Journal of Business Research, Journal of New Media & Society, Business Horizons, Journal of
8 Research Problem
Strategic Information Systems and the Journal of Computer-Mediated Communication. The literature review
was supported by books in the related research context. In addition, reports and white papers by acknowledged
consulting ?rms in the ?eld of information technology have been studied. The novel character of social media
and social business intelligence makes it that especially in these reports social business intelligence is mentioned,
whereas this term is less visible in the scienti?c area. These reports often contain examples from innovations
and practical experiences. A such, a variate overview will be presented about related research and theories to
this thesis.
Consulting Business Intelligence Experts
Firstly, interviewing experts contributes to an understanding of the actual situation of business intelligence in
organisations and the potential role of social media in this ?eld. This allows to scope the research in a topic
that is actual and relevant. Secondly, a part of the research will describe how business intelligence is applied
in organisations. Whereas this is mainly investigated using literature in the ?eld of business intelligence, BI
experts can validate the ?ndings. Thirdly, a procedure prescribing how to execute social business intelligence
will be develop. Such a procedure is required to be applicable in organisations as an integral part of the existing
– regular – BI process.
Content Analysis
Content analysis is appropriate for this research since it o?ers a systematic method to compare content for a
large sample of data. Content analysis is a research technique that can be used to identify what people are
sharing on social media. The research technique is described by Stephens (2012) as an “in-depth look at recorded
information” and as “a means of analysing texts” by Bos and Tarnai (1999). The sources of these texts can be
various, for example newspapers, articles, web sites, or – as in this research – social media posts. Neuendorf
(2002) de?nes content analysis as a “systematic, objective, quantitative analysis of message characteristics”.
As discussed, this thesis purposes to analyse the characteristics of social media posts, and link these posts to
organisational functions. Krippendor? (2004) states that a “content analysis entails a systematic reading of a
body of texts”, and argues that every content analysis requires the following six questions to be considered:
1. Which data are analysed?
2. How are they de?ned?
3. What is the population from which they are drawn?
4. What is the context relative to which the data are analysed?
5. What are the boundaries of the analysis?
6. What is the target of the inferences?
Bos and Tarnai (1999) provide a procedure for analysing content, which is schematically shown in ?gure 1-2. In
the ?rst step, the problem is formulated at the theoretical level, research questions are de?ned and the object of
investigation is determined. Secondly, the unit of analysis is de?ned by establishing categories and determining
the sample. The third step consists of pretesting the reliability of the data, and the validation of the categories
that were established in step 2. Discovered de?ciencies are consequently renovated. In the fourth step of the
content analysis procedure the data is collected and analysed. Finally, the results are interpreted and discussed
on the basis of the problem.
It is the stepwise approach of Bos and Tarnai (1999) that is applied on the content analysis of this thesis. We will
retrieve user-generated content from various social media platforms, store it into a database, and consequently
analyse the collected posts. By analysing the social media content, it is possible to classify the nature of the
content into categories, and ?nd di?erences between posts related to di?erent organisations.
1-6-3 Data Collection and Research Approach
Figure 1-3 illustrates the sequence and the links of the research steps in a schematic manner. A sample consisting
of several organisations across di?erent industries and with di?erent customer relation types will be established.
The selection of the organisations forms the point of departure for the collection of social media data. The
1-7 Project Scope 9
Research outline, research questions,
formulation of hypotheses, material to
investigate
Operationalising the categories,
determining the sample, determining the
unit of analysis
Establishment of categories
Theoretical level
Determining reliability and validating the
categories
Pretest
Appropriate statistical analyses
Data collection and evaluation
Immanent interpretation of the results,
discussion of the results on the basis of
the problem
Interpretation of the results
Figure 1-2: A procedure for analysing content (Bos & Tarnai, 1999).
content analysis requires that the social media data is available in a database. Therefore, social media posts
need to be loaded from the web into our database. This process is called scraping. The selection of the data
will be executed based on keywords corresponding to the selected organisations.
Scraping content from social media platforms results in unstructured data. In addition, the data is expected to
be polluted by e.g. spam or by users who apply nicknames related to the search terms used to scrape the content.
Therefore, the data needs to be cleaned before commencing the analysis. Once the spam and irrelevant posts
are removed from the dataset, the content analysis can start. In this analysis, social media posts are classi?ed
in relation to KPI categories based on the subject of the posts. Once the content analysis has been performed
for the ?rms, it is possible to identify di?erences between the subjects of social media posts across industries
and di?erent positions regarding end-users. Consequently, we can draw conclusions on the applicability of social
media for business intelligence purposes.
The third research question relates to how organisations should execute social business intelligence. For that
reason, a procedure prescribing how to execute social business intelligence will be designed. However, not
before the requirements of a social business intelligence systems are clear, the framework can be designed. The
framework is veri?ed by (i) BI experts and (ii) the ?t in the system that is currently executed in general BI
systems. Finally, conclusions are drawn regarding the applicability of social business intelligence in ?rms.
1-7 Project Scope
Business intelligence and social media are broad concepts. In order to describe the focus of the proposed research,
this section describes the scope of the research. Firstly, the research is scoped by a focus on a particular process
of business intelligence; registering and processing. Next, the research analyses social media activities on a set
of platforms, while others are excluded. Finally, some ?rms are part of our analysis while other are not.
Registering and Processing One possible way to represent BI, is through a cycle. Though many of these
cycles exist in the literature, they do not di?er much from each other (Pirttimäki & Hannula, 2003). Van Beek
(2006) describes BI as a cycle of registering, processing, and reacting on gathered data. Figure 1-4 highlights
the focus of this thesis. The gathering of data, ‘getting the data in’, is the most challenging aspect of BI,
requiring about 80% of the time and e?ort (Watson & Wixom, 2007). The fundamental scope of the proposed
research will be on this part; the gathering and registering of unstructured data generated on social media, and
is highlighted in ?gure 1-4. One of the core activities related to business intelligence, is the formulation of key
performance indicators. Not before these metrics are de?ned, the registering of data can commence. Therefore,
key performance indicators take a central role in this research.
10 Research Problem
Sample selection
Content analysis | Text mining | Data analysis
Literature review
on social media
Literature review
on (real-time)
business
intelligence
Literature review
on big data
Literature study
Analyse related
research
Create framework
to position this
thesis in exististing
theories
Content Analysis
Draw conclusions
on applicability of
social media
content for BI
Literature review
on (e-)business
Determine
knowledge gap
Formulate
research
questions
Relate social
media messages
to KPIs
Analyse
differences in
volume across
industries
Analyse
differences in
content across
industries
Classify
applicability of
social media for BI
Determine
keywords to scrape
Twitter data
Scrape social
media data
Record social
media posts in
database
Determine content
to scrape from
social media data
Scraping
Data cleaning
Analyse
differences in
content across
customer relations
Create sample of
companies to
analyse
Categorise
companies in
customer relation
types
Categorise
companies in
industries
Framework creation
Create social
business
intelligence
framework
Experience
in processing
social media
data
Understanding
‘where’ to apply
social BI
Understanding
‘how’ to apply
social BI
Formulating
requirements for
social BI
framework
Verification of
framework
Analyse
differences in
volume across
customer relations
Figure 1-3: Research Approach
Social Media Platforms Many social media platforms are available, and the range of social media platforms
is vast and growing (A. N. Smith et al., 2012). These platforms di?er in scope, functionality and in culture
(Boyd & Ellison, 2007). “Some sites are for general masses, like Twitter, Hi5 and Facebook. Other sites, like
LinkedIn, are more focused on professional networks. Media sharing sites such as MySpace, YouTube, and Flickr
concentrate on shared videos and photos” (Kietzmann et al., 2011). In addition, there also exist platforms that
are explicitly not purposed to be publicly accessible. An example of such a platform is Yammer, which is used by
1-8 Literature Review 11
G
D
A
Register
React
Process
Combination
Distribution
E
v
a
l
u
a
t
i
o
n
Operational
Tactical
Strategical
C
o
n
t
e
x
t
u
a
l
T
r
a
n
s
a
c
t
i
o
n
a
l
E
n
v
i
r
o
n
m
e
n
t
Figure 1-4: Thesis Scope, visualised in the BI cycle (van Beek, 2006)
organisations for internal communication. The focus of this thesis however is on publicly accessible platforms,
since the key purpose is to investigate what kind of information ?rms can derive from publicly accessible social
media. When investigating the opportunities of social media data for business intelligence purposes it is valuable
to collect data from a great variety of platforms, so that possible di?erences in the nature of the content can be
identi?ed. Our analysis includes 25 platform types that are monitored, among them Facebook’s public pages,
Twitter, Google+’s public pages, Identi.ca, YouTube, Flickr, Vimeo, Picasa, Wordpress based blogs, Blogger,
Typepad, RSS enabled blogs, Yahoo! Answers and Newssites. The di?erent platforms are monitored, implying
that each time a post is generated containing the predetermined keywords (e.g. ‘Albert Heijn’, or ‘Heineken’),
the post is extracted and saved into a database containing all gathered social media posts. Anno 2012, these
platforms are the most popular social media platforms in the Western World. However, there are many other
social network sites in the world. E.g Sina Weibo (the Chinese counterpart of Twitter), Qzone (China), Habbo,
Badoo (Latin America) and many other platforms are not part of our analysis. The selected social media
platforms – as well as the ?rms – are active in Europe.
Selection of Firms 18 ?rms have been selected for the analysis, that are active in di?erent industries. The
starting point of the sample selection has been the list of ?rms that are part of the Amsterdam Exchange Index
(“AEX”). The main reason for this selection criterion is the fact that these organisations are stock listed, and
hence publicise annual reports containing information about strategic initiatives, ?nancial ?gures, etc. In case
the analysis shows inter sector di?erences – e.g. between two comparable ?nancial institutions – the annual
reports may provide company speci?c information (e.g. amount of employees, attitude towards social media,
etc.) clarifying these di?erences. Whenever a sample containing privately owned companies would have been
selected, access to additional information would be limited. In addition, organisations listed in the AEX are
generally well-established, visible to the public and regularly subject to news articles. It is therefore expected
that these ?rms are subject of discussion on social media. The sample is further elaborated in section 4-2-2
(page 45).
1-8 Literature Review
In the following section research that relates to this thesis is presented. The literature related to the topic of
this thesis has been found using search queries (SOCIAL MEDIA), (SOCIAL BUSINESS INTELLIGENCE),
(SOCIAL MEDIA DATA), (SOCIAL MEDIA) AND (DATA EXTRACTION), (SOCIAL MEDIA) AND
(CRAWLING), (WEB 2.0) AND (CRAWLING), (TWITTER) AND (BUSINESS INTELLIGENCE) and (WEB
2.0) AND (BUSINESS INTELLIGENCE) in the scienti?c literature databases ScienceDirect, JStore and the
search engine Google Scholar. Existing research related to this thesis have been found in the scienti?c journals
of Electronic Markets, Computer Science, Journal of the American Society for Information Science, Public
Relations, Journal of Marketing, Public Relations Review, Expert Systems with Applications and the Journal of
Interactive Marketing. Though not all of these studies are explicitly related to business intelligence, the central
theme is the extraction of information from social media sites. We do not limit our review of related work to
12 Research Problem
one social media platform. Instead, the presented research consists of a mix in which Twitter, Facebook, Blogs,
and Questioning and Answering sites served as the data source.
Jansen, Zhang, Sobel, and Chowdury (2009) analysed 150,000 tweets containing branding comments, sentiments,
and opinions. The researchers analysed the content of the tweets, and found that 19% of microblogs’ posts
contain a mention of a brand. Of these branding microblogs, nearly 20% contained some expression of brand
sentiments. Of these, more than 50% were positive and 33% were critical of the company or product. The
research concludes that microblogging is an online tool for customer word of mouth communications, and is
especially suited for brand management activities.
Zhang et al. (2011) – in their quest to uncover the Twitter community dynamics – studied the “in?uences of
business engagement in online word-of-mouth communication” and investigated “the trajectories of a business’
online word-of-mouth message di?usion in the Twitter community”. They studied nine-brands on Twitter, and
concluded that “business engagement on Twitter enhances consumers’ engagement with online word-of-mouth
communication”. Therefore, the authors argue that “businesses must go beyond simply being aware of or taking
into consideration electronic word-of-mouth messages and instead must engage in the communication process
as both initiators and active participants. Next, Zhang et al. (2011) found that “retweeting, as an explicit way
to show consumers’ response to business engagement, only reaches consumers with a second-degree relationship
to the brand” and that the “life cycle of a tweet is generally 1.5 to 4 hours at most”.
McCorkindale (2010) investigated – based on a content analysis – how the Fortune 50 companies used Facebook.
The research studied how many fans an organisation had, what organisational information was included, if they
used photos and videos, if they used discussion boards, whether they generated feedback, etc. She found that
companies are using Facebook extensively, but that most companies were not using the site to “disseminate
news and information about the organisation”. Next, the research indicated that the companies should focus
more on “relationship-building strategies in order to encourage users to revisit the sites”. The content analysis
of McCorkindale (2010) revealed that there are several reasons why people post on Facebook pages. “Some
were current employees who identi?ed where they worked and for how long, while some were former employees
reconnecting with past coworkers. Headhunters posted jobs at competing corporations on the wall, and job
seekers posted they were looking for employment. Customers having product problems, especially in the
technology ?eld, would post their issues on the wall hoping to ?nd solutions. Journalists also posted on pages
requesting interviews”.
Agichtein, Castillo, Donato, Gionis, and Mishne (2008) argue that the “quality of user-generated content varies
drastically from excellent to abuse and spam”, and that the “task of identifying high-quality content in sites
based on user contributions – social media sites – becomes increasingly important”. Therefore, the authors
developed a method to exploit “community feedback to automatically identify high quality content”. Agichtein
et al. (2008) applied their model on a popular questioning and answering site (Yahoo! Answers). The system
of Agichtein et al. (2008) models all user relations, and applies the user ratings on the individual answers. As
such, the system determines high-quality content based on the ratings that users assigned to the content.
Guo, Zhang, Tan, and Guo (2012) developed a system that detects popular topics on Twitter. According to
the authors, the key technology in mining web text includes the modules “text classi?cation, clustering, topic
detection and tracking, opinion tendency identi?cation, and multi-document automatic summarisation”. Guo
et al. (2012) argue that popular topic detection systems should entail these ?ve modules. However, the nature
of tweets – “very short, sparse and spreading rapidly” – is di?erent from regular web text. Therefore, Guo et
al. (2012) propose a more “?exible and practical approach based on frequent pattern mining”.
Kozinets et al. (2010) qualitatively studied 83 blogs in order to understand how marketing departments in?uence
consumer-to-consumer communications. The authors distinguished the strategies of the marketeers into four
categories – evaluation, embracing, endorsement, and explanation.
Araujo and Neijens (2012) researched how top global brands participated in social network sites by investigating
which factors in?uence the presence and level of engagement of these brands on social network sites. The authors
reviewed the corporate websites of 129 brands in di?erent markets, targeting di?erent ages of audience, di?erent
home markets, di?erent web operations and in di?erent countries. Consequently, the authors determined
whether or not the companies refer to their presence at social network sites. The research found that social
network site “presence was signi?cantly higher for information technology and telecommunication brands”
(Araujo & Neijens, 2012), implying that the presence of ?rms on social media di?ers between ?rms in di?erent
industries. Furthermore, Araujo and Neijens (2012) found that “brands targeting younger audiences also engage
at higher levels than brands targeting generic audiences” and that the “country in which the brand operates
plays a signi?cant role in a brand’s likelihood of adopting social network sites”. The ?ndings of Araujo and
1-9 Scienti?c Relevance 13
Neijens (2012) indicate that the applicability of social media for business purposes di?ers between ?rms, which
support the basis of this thesis.
Dey and Haque (2008) acknowledge that “the data generated from online communication acts as potential gold
mines for discovering knowledge”. However, as Dey and Haque (2008) illustrate, “the quality of texts generated
from online sources can be extremely poor and noisy” because the “text data typically comprises spelling errors,
ad-hoc abbreviations and improper casing, incorrect punctuation and malformed sentences”. It is therefore
that text mining techniques based on “pure linguistic strategies fail to extract information from noisy text”.
According to Dey and Haque (2008), “statistical techniques on the other hand which though not as successful
as the linguistic methods, are more suited to extract information from noisy text. However, lack of appropriate
training data often poses as a bottleneck”. Dey and Haque (2008) conclude that – when processing unstructured
social media data – “domain related training sets” are required to the clean the text before the text can be
processed by Natural Language Processing Tools. With such domain related training sets, the word ‘small’ can
be classi?ed as either positive or negative, depending on its context.
Lovejoy et al. (2012) performed a content analysis of the tweets related to 73 non-pro?t organisations to examine
“how these organisations use Twitter to engage stakeholders”. Within that analysis, the researchers looked at
“the organisations’ utilisation of tweet frequency, following behaviour, hyperlinks, hashtags, public messages,
retweets, and multimedia ?les”. Lovejoy et al. (2012) conclude that non-pro?t organisations use social media as
a “one-way communication channel”, and not as a platform for “conversation and community building”.
Lee (2012) acknowledge that the “contents of microblogs preserve valuable information”. In his study, Lee
(2012) focused on real-world o?ine events, and the information that was generated on social media sites related
to those events. With his system, it is possible to detect real-world events through the content on social media
sites. Also Lee (2012) argues that the challenge of automatically classifying social media posts is the informal
structure of the text.
Dhar & Chang’s (2009) research is one of the few that studied the relation between social media activity and
organisational performance. More speci?cally, they employed social media data to predict sales in the music
industry. Using linear and nonlinear regression, Dhar and Chang (2009) found that “(a) the volume of blog
posts about an album is positively correlated with future sales, (b) greater increases in an artist’s Myspace
friends week over week have a weaker correlation to higher future sales, (c) traditional factors are still relevant
– albums released by major labels and albums with a number of reviews from mainstream sources also tended
to have higher future sales”.
Tirunillai and Tellis (2012) studied the relationship between user-generated content and stock market
performance of the ?rm. The authors found that “of all the metrics of UGC, volume of chatter has the
strongest positive e?ect on abnormal returns and trading volume. Whereas negative UGC has a signi?cant
negative e?ect on abnormal returns, positive UGC has no signi?cant e?ects on these metrics. The volume of
chatter and negative chatter have a signi?cant e?ect on trading volume”. In addition, Tirunillai and Tellis
(2012) found that “an increase in o?-line advertising signi?cantly increases the volume of chatter and decreases
negative chatter”.
From the literature review, we can conclude that there is scienti?c attention in the research ?eld of social media
and the relation with organisational performance. However, no research has been found that investigates the
applicability of social media for organisations in (i) di?erent industries and with (ii) di?erent customer relation
types. Next, though some studies individual tackle di?culties that are inherent to the usage of social media data,
no research has been found that integrally describes how social media should be collected and processed within
a ?rm. As illustrated, the opportunities for social media are bene?cial on many aspects. However, managers
are also reluctant to allocate budget to social media activities (Weinberg & Pehlivan, 2011) and incorporate
social media data in the ?rm’s BI process, because they do not fully understand what social media intelligence
may bring to the ?rm. In addition, it is unclear which type of ?rms are subject of discussion on social media
and – if they are – how a ?rm should collect and process these data so that it adds value to the ?rm.
1-9 Scienti?c Relevance
The proposed research touches the world of e-business, which implies “the transformation of key business
processes through the use of internet technologies” (Cha?ey, 2009). The monitoring of opinions, customer
thoughts, etc. by electronic means – for instance by social media sites – can be positioned under the denominator
‘e-business’. Many literature exists in which the world of e-business is described. This research contributes to
existing models and theories by positioning social media content as an external factor in these theories.
14 Research Problem
Next, science is built on data. The more data is available to scientists, the “greater the level of transparency
and reproducibility and hence the more e?cient the scienti?c process becomes” (Molloy, 2011). Historically,
scienti?c data has not been openly available. In recent years, several scientists advocate the application of open
data. The proposed research will be based on publicly accessible data – coming from social media – and will
hence contribute to understanding the opportunities and threats of applying public data for scienti?c purposes.
The literature contains many de?nitions of business intelligence, and provides theories describing how BI
processes internal as well as external data. Data from social media can be positioned under external factors.
This thesis positions explicitly adds social media data into the existing theories of business intelligence.
Next, the research will be executed based on the research methodology of content analysis. Though this method
is yet widely applied in many research areas, the fact that the source of the content in this research is social
media, makes it new. The lessons learned in this research from applying a content analysis on social media data
contribute to the research methodology of content analysis.
Finally, the literature of customer relationship management (“CRM”) describes how organisations interact with
their customers. Recent literature also includes social media solutions into CRM activities. This research reveals
how user-generated content varies between industries, departments and the position regarding consumers. As
such, the conclusions of this thesis contribute to the applicability of CRM through means of social media.
1-10 Societal Relevance
Many executives avoid or ignore social media because they do not understand what it is and how to engage
with it and learn, though they sense that social media is – and will remain – an important “fabric of commerce”
(Kietzmann et al., 2011; Weinberg & Pehlivan, 2011). This thesis contributes to a further understanding of social
media, and to leverage the opportunities of applying the valuable content on these platforms for organisational
e?orts. The social media phenomenon is relatively fresh, Facebook was launched in 2004, Twitter in 2006.
Because of the novelty, the opportunities for organisations’ social BI activities are rather unexplored. This
research also contributes to an understanding for ?rms whether or not social media data can be applied for
business intelligence purposes in which context. In addition, we expect that legacy BI vendors – such as SAP,
Oracle and IBM – are soon asked by their clients to add a social media component to their BI suite. For these
organisations social business intelligence is also a new phenomenon, and social media data can not be directly
applied to their existing systems (Reinhold & Alt, 2011). The conclusions of this thesis support BI vendors in
the development of social media components within their product range.
1-11 Project Deliverable
This research will reveal two central questions that describe (i) where and (ii) how ?rms can derive information
from social media data for their decision-making process. Therefore, the project has two deliverables.
Deliverable 1: Where? This deliverable speci?es in which contexts a ?rm can implement social business
intelligence. For reasons explained in section 1-2, it is expected that a ?rm’s contexts and aspects determine the
applicability of social media data for BI purposes. This deliverable allows organisations to determine whether
or not they are suited for the applicability of social media for business intelligence.
Deliverable 2: How? This deliverable consists of a procedure that prescribes how ?rms can execute a social
business intelligence system. The procedure can be considered as a document prescribing how an organisation
can execute social business intelligence, and which technical and institutional elements are involved in a social
BI system.
Chapter 2
Conceptual Frame of Research
In this chapter, we de?ne and illustrate what business intelligence (“BI”) is, how BI is applied in ?rms, what
the most important elements are and how BI is regarded in this thesis. In a later stadium of this research a
procedure for processing social media data to support business intelligence will be developed. Such a procedure
is required to ?t in the current method that ?rms adhere to in executing BI. Therefore, an understanding of
business intelligence within ?rms is essential. This chapter provides the knowledge of business intelligence that
is required when we develop a procedure to collect and process social media data for BI purposes in a later
stadium.
Section 2-1 starts by a description of the various de?nitions of BI, and consequently formulates the perspective
on BI that is adhered to throughout this research. In section 2-2 the elements relating to the determination
of ‘what to measure?’ are discussed, a key activity in business intelligence. Section 2-2 explains the relation
between a ?rm’s strategy and the ?rm’s performance metrics. As we will see, performance metrics take a central
role in BI. In section 2-3, the processing of data is described. Finally, section 2-4 concludes this chapter by a
description of how business intelligence is applied within ?rms.
2-1 Business Intelligence Perspectives
Business intelligence is a process in which data is translated into information that is required for managerial
decision-making. The literature contains many de?nitions of business intelligence. Elbashir et al. (2008) state
that “business intelligence systems provide the ability to analyse business information in order to support
and improve managerial decision making across a broad range of business activities”. Van Beek (2006)
de?nes business intelligence as “a continuous process that helps organisations gathering and registering data,
analysing it and consequently applying the resulting information and knowledge in decision-making processes
to improve organisational performance”. Rouibah and Ould-ali (2002) describe business intelligence as “a
strategic approach for systematically targeting, tracking, communicating and transforming relevant weak signs
into actionable information on which strategic decision-making is based”. Lonnqvist and Pirttimaki (2006)
de?ne BI as “an organised and systematic process by which organisations acquire, analyse, and disseminate
information from both internal and external information sources signi?cant for their business activities and for
decision making”. Although these de?nitions vary slightly from each other, the common aspect is that business
intelligence is perceived as a process that translates data into interpretable information that supports managerial
decision-making.
Van Beek (2006) visualises business intelligence (“BI”) as a cycle, consisting of three main processes (?gure 2-1).
For the remainder of this thesis, we follow van Beek’s – loosely de?ned – perspective on business intelligence
because it captures the various de?nitions of BI found in the literature. The three main processes – register,
process and react – are discussed in the following paragraphs.
Register The BI cycle starts with carefully listening – registering – to the environment. Within the
environment, a distinction is made between contextual and transactional environments. The contextual
environment consists of aspects that (may) have an e?ect on the organisation. The transactional environment
16 Conceptual Frame of Research
G
D
A
Register
React
Process
Combination
Distribution
E
v
a
l
u
a
t
i
o
n
Operational
Tactical
Strategical
C
o
n
t
e
x
t
u
a
l
T
r
a
n
s
a
c
t
i
o
n
a
l
E
n
v
i
r
o
n
m
e
n
t
Figure 2-1: Business Intelligence Cycle (van Beek, 2006).
consists on the one hand of actors that have a direct relation with the company, like customers, suppliers,
employees and competitors. On the other hand, the transactional environment is made up of institutions
a?ecting the organisation, like new policies or legislation. BI registers signals arising from this environment.
Process Consequently, when data (in whatever format) is registered, it is required to be processed. Processing
the gathered data will reveal trends and provide valuable information. Van Beek (2006) positions a ‘small BI
cycle’ within this process; data is gathered, analysed, and distributed to the right organisational departments.
This part of the BI cycle will be further described in section 2-3.
React Following the results provided by the processing of the data, the company can react. Van Beek
(2006) argues that a company can react on three levels; operational, tactical or strategical. Consequently,
the environment evaluates the companies’ changes in interactions, resulting in new signals for the ?rm’s BI
cycle.
2-2 Registering the Right Indicators
As illustrated in the previous section, the business intelligence process starts with registering. But, what is it
that a ?rm has to register? Not unless a ?rm has clearly set what will be registered, the BI cycle can commence.
The determination of ‘what to measure’ is a process on itself, which is described in this section. In the following
sections, the necessary steps to determine what a ?rm has to register are described. Section 2-2-1 describes that
?rms formulate – or, should formulate – their performance metrics based on their strategy. In section 2-2-2, two
widely accepted frameworks that support ?rms in the formulation of performance metrics are discussed. Section
2-2-3 discusses a framework that illustrates how a ?rm should design and implement a system for measuring
organisational performance. Next, section 2-2-4 describes the various types of performance measures that are
applied by ?rms. Finally, section 2-2-5 describes ten commonly applied categories of key-performance indicators,
which take a central role in the business intelligence process.
2-2-1 Strategy and Business Model
Firms align their business model with their strategy. The determination of ‘what to measure’ initially starts with
the ?rm’s strategy. A strategy consists of a mission, values, vision, goals, objectives and plans. The “mission
and values de?ne why the organisations exists, what it does, and its guiding principles. The vision combines
an overarching purpose with an ideal, future-state competitive positioning. Goals are broad statements that
embody what the organisation would like to achieve in three to ?ve years, while objectives represent short-term
goals of one to three years” (Eckerson, 2009).
Strategies are translated into business models (Bouwman, Faber, Haaker, Kijl, & de Reuver, 2008). There exists
a variety of views on business models in the literature. Osterwalder and Pigneur (2010) state that “a business
model describes the rationale of how an organisation creates, delivers, and captures value”. Bouwman et al.
2-2 Registering the Right Indicators 17
(2008) provide an all-embracing de?nition of a business model for service-oriented organisations by stating that
“a business model is a blueprint for a service to be delivered, describing the service de?nition and the intended
value for the target group, the sources of revenue, and providing an architecture for the service delivery, including
a description of the resources required, and the organisational and ?nancial arrangements between the involved
business actors, including a description of their roles and the division of costs and revenues over the business
actors”. In other words, a business model describes what a ?rm does, why it does that, how it does that, with
whom it does that and for whom it does that.
By aligning the business model with the ?rm’s strategy, managers question themselves “why are we performing
this activity?”, and “what does it contribute to?”. “It is only through consistency of action that strategies
are realised” (Neely, Gregory, & Platts, 1995). Gates (2001) argues that applying value driver maps to
analyse a company’s vision and drivers of performance helps a company aligning its business model with its
strategy. By aligning the business model with the strategy, a ?rm ensures that it performs those activities that
contribute to the intended strategy. It is in this phase of the BI process where managers determine where in
the organisation and by which activities value is added to the main objective of the organisation, and why and
how the performance of these activities are to be measured. There are many frameworks developed to craft
a business model that is based on the ?rm’s strategy, at which – in the end – performance metrics can be
formulated. These frameworks are the topic of the following section.
2-2-2 Frameworks Supporting the Formulation of Performance Indicators
Performance measurement is “the process of quantifying the e?ciency and e?ectiveness of action” (Neely et al.,
1995). To quantify actions – forming the ?rm’s business model – indicators are required that represent these
actions. Such indicators are called performance indicators. Performance indicators should be derived from a
?rm’s strategy (R. S. Kaplan & Norton, 1992, 1993, 1996; Kennerley & Neely, 2003; Tsai & Cheng, 2012;
Fortuin, 1988). The previous section revealed that a business models amongst others speci?es which activities
are required to deliver value for the customer of the ?rm. Hence, the business model prescribes which activities
are to be executed. The activities are consequently required to be measured, so that managers can determine
the organisation’s performance in accordance with its strategy. These activities are measured by performance
indicators, indicating the performance of the individual activities.
But how do we determine the right performance indicators? One of the ?rst widely recognised frameworks
(Neely, Bourne, & Kennerley, 2000) that help managers to decide ‘what to measure’ is the balanced scorecard,
developed by R. S. Kaplan and Norton (1992). R. S. Kaplan and Norton (1992) provide a framework that
supports managers in formulating performance indicators based on four perspectives. Because the balanced
scorecard ensures that managers do not only focus on ?nancial ?gures, it gives managers a “comprehensive view
of the business”, which is required in the competitive environment of the ?rm (R. S. Kaplan & Norton, 1992).
Since the balanced scorecard links a company’s strategy with concrete actions (R. S. Kaplan & Norton, 1996),
it is considered as a tool helping managers to align the company’s business model with its strategy. In response
to the balanced scorecard, Neely, Adams, and Crowe (2001) developed a scorecard that adopts a multi-actor
view in formulating performance metrics, and hence incorporates the perceptions of multiple stakeholders into
the performance metrics formulation process. Both frameworks are discussed in the following sections.
The Balanced Scorecard
Because “you get what you measure” (Kennerley & Neely, 2003), R. S. Kaplan and Norton (1992) advocate
that a ?rm should measure those metrics that contribute to the ?rm’s strategy. In addition, R. S. Kaplan
and Norton (1992) argue that managers should not only focus on ?nancial ?gures, but also on other areas
representing organisational performance. Therefore, R. S. Kaplan and Norton (1992) developed the “balanced
scorecard”. The balanced scorecard is not without reason called a ‘balanced’ scorecard. In essence, it stimulates
managers to not only think in ?nancial ?gures when measuring organisational performance, but also on other
areas. The framework distinguishes organisational activities in the (i) customer-, (ii) internal business-, (iii)
innovation and learning-, and (iv) ?nancial perspective, which are discussed below:
• The customer perspective describes how customers view the ?rm, and ensures that customer’s needs are
ful?lled. R. S. Kaplan and Norton (1992) further categorise the customer’s concerns into time, quality,
performance and service, and cost. Hence, when a manager adopts the customer perspective in formulating
performance metrics, he or she will formulate performance metrics that involve these categories.
18 Conceptual Frame of Research
• The internal business perspective describes where the particular ?rm must excel at and speci?es what a
“company must do internally to meet its customers’ expectations” (R. S. Kaplan & Norton, 1992). It is
in this perspective where managers “attempt to identify and measure their company’s core competencies,
the critical technologies needed to ensure continued market leadership. Companies should decide what
processes and competencies they must excel at and specify measures for each” (R. S. Kaplan & Norton,
1992).
• The third perspective of the balanced scorecard – innovation and learning – ensures that the organisation
continues to improve and create value. Due to competition, “the targets for success keep changing”
(R. S. Kaplan & Norton, 1992). It is therefore that companies must design its organisation in a way
that it can innovate. It is in this perspective where managers consider the development of new products,
entering new markets, etc.
• The ?nancial perspective ensures that the shareholders’ needs are ful?lled. “Financial measures
indicate whether the company’s strategy, implementation, and execution are contributing to bottom-line
improvement” (R. S. Kaplan & Norton, 1992).
The balanced scorecard set the scene for the development of a variety of other performance measurement
frameworks at the beginning of the 1990s.
The Performance Prism
In response to the various types of scorecards that have been developed after Kaplan & Norton’s (1992)
balanced scorecard, Neely et al. (2001) developed a “second generation performance measurement framework”
called the performance prism. According to Neely and Adams (2005), there were three fundamental reasons
why the balanced scorecard was outdated, and why a new framework was required. Firstly, the balanced
scorecard solely focuses on the needs of two groups of (internal) stakeholders; shareholders and customers. In
today’s business environment, ?rms can no longer consider only these two groups of stakeholders. For example,
employees, environmental parties, labour unions, other communities, regulatory bodies, etc. have been fully
denied in the balanced scorecard, while these groups truly in?uence a ?rm in practice. Second, an organisation’s
“strategy, processes, and capabilities have to be aligned and integrated with one another” (Neely & Adams,
2005), e.g. with the processes of the ?rm’s suppliers. Third, ?rms “have to recognise that their relationships are
reciprocal – stakeholders have to contribute to organisations as well as receive something from them” (Neely &
Adams, 2005). Thus, the key innovation that is captured in the performance prism is the fact that it takes a
comprehensive stakeholder orientation, i.e. a multi-actor perspective, whereas former frameworks (such as the
balanced scorecard) adopt a mono-actor perspective. It is necessary to adopt a multi-actor perspective, since
?rms need to have “contributions from their stakeholders – usually capital and credit from investors, loyalty
and pro?t from customers, ideas and skills from employees, materials and services from suppliers, and so on”
(Neely & Adams, 2005).
The performance prism consists of ?ve perspectives, whereas the balanced scorecard comprises of four. The
central element in all these ?ve performance prism perspectives is the stakeholder aspect.
• Stakeholder Satisfaction
The stakeholder satisfaction perspective ensures that managers consider “who the ?rm’s stakeholders are,
what they do, and what they need”. Whereas the balanced scorecard explicitly focuses on two groups
of stakeholders, i.e. on customers through the customer perspective and on shareholders through the
?nancial perspective, the performance prism does not specify stakeholder groups but rather allows for
ambiguity. As such, the ?rm is stimulated to consider all stakeholder groups in its ecosystem, and specify
what these groups want.
• Strategies
The second facet of the performance prism focuses on strategies. With the identi?cation of the needs in the
stakeholder satisfaction perspective, strategies can be developed that ful?l the needs of the stakeholders.
The key question in this facet is: “What strategies should the organisation adopt to ensure that the
wants and the needs of its stakeholders are satis?ed?” (Neely & Adams, 2005). Whereas the balanced
scorecard method of formulating measures starts with the ?rm’s strategy, Neely and Adams (2005) argue
that strategies should be designed in accordance with the needs of the stakeholders.
2-2 Registering the Right Indicators 19
• Processes
In the third facet of the performance prism, the processes are de?ned that are required to ful?l the
strategies de?ned in the second perspective. It is in this perspective where general business processes,
such as product development, demand generation, demand ful?lment, planning, etc. are de?ned. Neely et
al. (2001) stress the importance of measures that re?ect the performance of the processes. Thus, it is in
this perspective where organisations consider which measures are required to determine the organisational
performance.
• Capabilities
Within the fourth aspect of the performance prism, the capabilities required to execute the processes are
identi?ed. Capabilities consist of a combination of “people, practices, technology, infrastructure, etc.”.
• Stakeholder Contribution
Finally, the performance prism “recognises the fact that not only do organisations have to deliver value to
their stakeholders, but also that organisations enter into a relationship with their stakeholders which should
involve the stakeholders contributing to the organisation” (Neely et al., 2001). It is in the stakeholder
contribution perspective where ?rms consider what they want from their stakeholders.
5. Capabilities: What capabilities do we need to put
in place to allow us to operate our processes?
Together, these ?ve perspectives provide a comprehen-
sive and integrated framework for thinking about organi-
zational performance in today’s operating environment
(see Fig. 1).
The performance prismalso seeks to address the short-
comings of the ?rst-generation measurement frameworks
and methodologies, such as the balanced scorecard,
the work on shareholder value, and the various self-
assessment frameworks, such as the Malcolm Baldrige
Award criteria and the business excellence model of
the European Foundation for Quality Management
(EFQM).
The Nature of the Measurement
Problem
Why is this new performance measurement framework
needed? After all, everyone knows that ‘‘you can’t manage
what you don’t measure.’’ And given that people have
been managing organizations for years, then surely by
now they must have perfected their measurement
systems.
Sadly, as in so many walks of life, theory does not re?ect
practice. The number of organizations with weak perfor-
mance measures and measurement systems in place is
immense. Examples abound of organizations that have
introduced performance measures that quite simply
drive entirely the wrong behaviors. There must be
a better way.
There has been a revolution in performance measure-
ment and management during the last few years. Various
frameworks and methodologies—such as the balanced
scorecard, shareholder value added, activity-based cost-
ing, cost of quality, and competitive benchmarking—have
each generated vast interest, activity, and consulting rev-
enues, but not always success. Yet therein lies a paradox.
It might reasonably be asked: how can multiple, and
seemingly inconsistent, business performance frame-
works and measurement methodologies exist? Each
claims to be unique and comprehensive, yet each offers
a different perspective on performance.
Kaplan and Norton’s balanced scorecard, with its four
perspectives, focuses on ?nancials (shareholders), cus-
tomers, internal processes, plus innovation and learning.
In doing so, it downplays the importance of other stake-
holders, such as employees, suppliers, regulators, and
communities. The business excellence model combines
results, whichare readily measurable, withenablers, some
of which are not. Shareholder value frameworks incorpo-
rate the cost of capital into the equation, but ignore ev-
erything (and everyone) else. Both the activity-based
costing and the cost of quality frameworks, on the
other hand, focus on the identi?cation and control of
cost drivers (non-value-adding activities and failures/
non-conformances, respectively), which are themselves
often embedded in the business processes. But this highly
process-focused view ignores any other perspectives on
performance, such as the opinions of investors, custom-
ers, and employees. Conversely, benchmarking tends to
involve taking a largely external perspective, often com-
paring performance with that of competitors or some-
times other ‘‘best practitioners’’ of business processes
or capabilities. However, this kind of activity is frequently
pursued as a one-off exercise toward generating ideas
for—or gaining commitment to—short-term improve-
ment initiatives, rather than the design of a formalized
ongoing performance measurement system.
How can this be? How can multiple, seemingly
con?icting, measurement frameworks and methodologies
exist? The answer is simple: they can exist because they all
add value. They all provide unique perspectives on per-
formance. They all furnishmanagers with a different set of
lenses through which they can assess the performance of
their organizations. The key is to recognize that, despite
the claims of some of the proponents of these various
approaches, there is no one best way to address the mea-
surement and management of business performance. The
reason for this is that business performance is itself
a multifaceted concept, the complexity of which the ex-
isting frameworks only partially address. Essentially, they
provide valuable point solutions.
A Better Solution to the
Measurement Problem
Our solution is the three-dimensional framework that we
call the performance prism. This framework has been
deliberately designed to be highly ?exible so that it can
provide either a broad or a narrow focus. If only a partial
• Stakeholder satisfaction
• Strategies
• Processes
• Capabilities
• Stakeholder contribution
Figure 1
42 Performance Prism
aspect of performance management is required, such as
a single stakeholder focus or a particular business process
agenda, then the performance prism can be applied to
designing a measurement system and appropriate mea-
sures (and their attendant metrics) that address that con-
text. Conversely, if a broad corporate or business unit
performance management improvement initiative is re-
quired, the performance prism is equally capable of sup-
porting that, too. How does it help to achieve these aims?
The performance prism has ?ve perspectives. The top
and bottom perspectives are stakeholder satisfaction and
stakeholder contribution. The three side perspectives are
the organization’s strategies, processes, and capabilities
for addressing those sets of wants and needs. Figure 2
illustrates these ?ve basic perspectives of performance
measurement and management.
Why does the framework look like this and why does it
consist of these constituent components? It is clear that
those organizations aspiring to be successful in the long
termwithin today’s business environment need to have an
exceptionally clear picture of who their key stakeholders
are and what they want or need. But having a clear picture
is not enough. In order to satisfy their own wants and
needs, organizations have to access contributions from
their stakeholders—usually capital and credit from
investors, loyalty and pro?t from customers, ideas and
skills from employees, materials and services from sup-
pliers, and so on. They also need to have de?ned what
strategies they will pursue to ensure that value is delivered
to their stakeholders. In order to implement these strat-
egies, they have to understand what processes the enter-
prise requires and must operate both effectively and
ef?ciently. Processes, in themselves, can only be executed
if the organization has the right capabilities in place—the
right combination of people skill sets, best practices, lead-
ing technologies, and physical infrastructure.
In essence, then, the performance prism provides
a comprehensive yet easily comprehensible framework
that can be used to articulate a given business’s operating
model. Its components are described in the following
sections.
Stakeholder Satisfaction
Where should the measurement design process begin?
One of the great myths (or fallacies) of measurement
design is that performance measures should be derived
from strategy. Listen to any conference speaker on the
subject. Read any management text written about it. Nine
times out of ten the statement ‘‘Derive your measures
from your strategy’’ will be made. This is such a concep-
tually appealing notion that nobody stops to question it.
Yet to derive measures from strategy is to misunderstand
fundamentally the purpose of measurement and the role
of strategy. Performance measures are designed to help
people track whether they are moving in the intended
direction. They help managers establish whether they
are going to reach the destination they set out to reach.
Strategy, however, is not about destination. Instead, it is
about the route that is chosen—how to reach the desired
destination.
At one level, this is a semantic argument. Indeed,
the original work on strategy, carried out in the 1970s
by Andrews, Ansoff, and Mintzberg, asserted that
a strategy shouldexplain both the goals of the organization
and a plan of action to achieve these goals. Today,
Stakeholder satisfaction
Investors
Customers &
Intermediaries
Employees
Regulators &
Communities
Suppliers
What measures? What measures? What measures?
What measures?
Investors
Customers &
Intermediaries
Employees
Regulators &
Communities
Suppliers
Investors
Customers and
intermediaries
Employees
Regulators and
communities
Suppliers
Which
strategies?
What measures? What measures? What measures?
Which
processes?
Which
capabilities?
What measures?
What measures?
Stakeholder satisfaction
Figure 2
Performance Prism 43
Figure 2-2: Five Facets of the Performance Prism. Adopted from Neely & Adams (2005).
Figure 2-2 presents the performance prism. The ?ve areas of the prism represent one of the perspectives.
The triangular – outside facing – surfaces represent the two stakeholder perspectives, which are unique for
the performance prism. The three rectangular – inside facing – surfaces represent the strategy, processes and
capabilities perspectives. The fact that stakeholders are an important element in the performance prism is
also highlighted by the substance of the prism, which mentions groups of stakeholders. As argued by Neely
and Adams (2005), the performance prism “has been deliberately designed to be highly ?exible so that it can
provide either a broad or a narrow focus”. As a result, the perspectives of the performance may seem vague.
The authors explicitly made the perspectives vague so that the framework is broadly applicable.
To conclude, a ?rm should register those metrics that re?ect the performance of the activities that contribute
to the ?rm’s strategy. Since organisational activities are tailored to the ?rm’s strategy, and metrics are derived
from these activities, there exists a link between performance metrics and an organisation’s strategy. Though
the alignment of performance metrics and strategy may sound self-evident, many organisations struggle with
strategic alignment: even at the healthiest companies, about 25% of the employees are unclear about their
company’s direction. KPMG (2009) argues that in many organisations, there is “no explicit linkage between
20 Conceptual Frame of Research
the strategy and the information used to manage the business”, implying that managers are measuring activities
that do not contribute to the ?rm’s strategy. Managing without or the wrong metrics “gives one the feeling of
being lost with no hope”, and leads to a “lack of management control” (R. Smith, 2006). Once the performance
metrics are established and mutual di?erences in importance are assigned, the BI process can commence. The
values of the performance indicators will then reveal the performance of the ?rm. As illustrated, R. S. Kaplan and
Norton (1992) introduced the ?rst framework that considered other metrics than solely ?nancial ?gures. Next,
Neely et al. (2001) developed – in response to the ?aws relating to the mono-actor perspective of the balanced
scorecard – a framework that considers the ?rm’s stakeholders; the performance prism. The performance prism
is deemed as a framework that is – given today’s business landscape – better suited for performance metrics
formulation than the balanced scorecard. Moreover, since the purpose of this thesis is to incorporate social
media data – created by multiple actors – in management information for di?erent ?elds and departments, a
multi-actor perspective is required.
2-2-3 Performance Measurement System Design Process
Wisner and Fawcett (1991) – as quoted in Neely et al. (2000) – developed a “process for performance
measurement system design”. Figure 2-3 shows the nine-step process proposed by Wisner and Fawcett (1991),
which clearly illustrates that performance metrics should be derived from a ?rm’s strategy, and that performance
indicators should be assigned to functional areas – performing individual activities – of the ?rm.
Clearly define the firm’s mission statement.
Identify the firm’s strategic objectives using the mission statement as a guide.
Develop an understanding of each functional area’s role in achieving the various strategic objectives.
For each functional area, develop global performance measures capable of defining the firm’s overall competitive
position to top management.
Communicate strategic objectives and performance goals to lower levels in the organisation. Establish more
specific performance criteria at each level.
Assure consistency with strategic objectives among the performance criteria used at each level.
Assure the compatibility of performance measures used in all functional areas.
Use the performance measurement system to identify competitive position, locate problem areas, assist the firm in
updating strategic objectives and making tactical decisions to achieve these objectives and supply feedback after
the decisions are implemented.
Periodically re-evaluate the appropriateness of the established performance measurement system in view of the
current competitive environment.
Figure 2-3: Performance measurement system design process (Wisner and Fawcett, 1991).
In a joint research, the business intelligence system vendors SAP, IBM, Corda and Pentaho examined how
organisations formulate performance metrics. Also they found that managers determine performance metrics
based on the ?rm’s strategy. Next, they concluded that performance metrics should be “tailored to every
individual and role in the organisation” (Eckerson, 2009). As a result, departments and individuals consequently
understand how their activities contribute to the company’s strategy, which is often stated in generic and vague
terms. Consequently, employees will focus on those activities that are important, because “what’s get measured,
gets done” (Kennerley & Neely, 2003). This typical management quote illustrates the imposing consequences
that descent from the determination of performance metrics, that is, “what’s not get measured, gets not done”.
As illustrated by Eckerson (2009), “if the metrics do not accurately translate the company’s strategy, the
organisation will ?ounder”. It is therefore that determining performance metrics is a critical activity of business
intelligence.
2-2 Registering the Right Indicators 21
2-2-4 Typology of Performance Indicators
Performance indicators are key elements in business intelligence, since they re?ect the performance of the
activities that contribute to the ?rm’s strategy. In this section, we elaborate more about performance indicators
and the types of metrics that exist.
Leading and Lagging Indicators There are two fundamental types of indicators; leading indicators and lagging
indicators. Leading indicators lead to results, and are also referred to as ‘(value) drivers’. Lagging indicators
are the results that measure the output of past activities, and are also known as ‘outcomes’ (R. Smith, 2006).
Leading indicators are used to manage, while lagging indicators measure how well has been managed.
With leading indicators it is possible to respond directly when poor results are found. With lagging
indicators, “we get value from knowing how well we performed but have little opportunity to immediately
a?ect under-performance” (R. Smith, 2006). Hence, leading indicators are more powerful, and can be perceived
as short-term indicators of an organisations’ results. It is therefore that ?rms manage by leading indicators.
Illustratively, table 2-1 lists some examples of leading and lagging indicators. It is noteworthy that among
di?erent departments and individuals in organisations there could exist pluriformity in the perception of the
type of indicators, “one man’s outcome measure can be another man’s value driver” (Eckerson, 2009).
Table 2-1: Examples of leading and lagging indicators
Leading indicators Lagging indicators
New sales today Revenues
Planned rework today Cost
Customer cases currently open Capacity
Contracts in negotiation for Q2 Return on equity
Identi?ed software bugs Customer satisfaction
Employee retention
Margins
Reliability
Failures
Downtime
Quantitative and Qualitative Indicators Another distinction between metrics is the di?erence between
quantitative or qualitative based indicators. Quantitative indicators measure processes by counting, adding,
averaging, etc. numbers. Examples of quantitative measures are inventories, number of orders, number of clients,
delivery time of goods, sales, other ?nancial ?gures, etc. In contrast with qualitative indicators, quantitative
indicators are relatively easy to measure.
However, there are many other criteria to judge performance than solely on (?nancial) quantitative indicators
(Neely et al., 2000; Eccles, 1991). Other metrics are qualitative in nature and require a proxy to be measured.
An example of a qualitative measure is customer satisfaction. The measurement of customer satisfaction
results in quantitative data, but is primarily based on subjective interpretation of customers’ opinions.
Customer satisfaction is therefore traditionally measured by surveys (Peterson & Wilson, 1992). “Traditionally,
performance evaluation has depended to a great extent on ?nancial indicators. However, given the current
environmental uncertainties, ?nancial indicators can no longer give a complete view of business operations”
(Tsai & Cheng, 2012). It is therefore that qualitative measures are as much as important as quantitative
measures. The trick is to identify the links between qualitative measures and ?nancial measures. Firms can for
instance conduct statistical analyses to correlate qualitative indicators with ?nancial performance. Regression
analysis can be applied to identify the key drivers that impact sales, pro?tability, etc. The performance prism
framework allows for the incorporation of qualitative indicators next to quantitative indicators.
Key Performance Indicators To distinguish between performance indicators that are more important than
others, some indicators are termed ‘key-performance indicators’ (“KPIs”). But what is it that makes a
performance indicator ‘key’? PWC (2007) argues that the performance indicators that are key to a ?rm are
those that are used to manage the business. According to Tsai and Cheng (2012), KPIs “are the groundwork of
the performance system which turns the strategic goals of a company into long-term objectives”. The addition of
22 Conceptual Frame of Research
the word ‘key’ to a performance indicator indicates that these metrics are assigned more attention than others.
Thus, it are the KPIs that represent processes that are paramount for the success of a ?rm. Table 2-2 lists the
elements that a key-performance indicator should ful?l, it should be speci?c, measurable, attainable, realistic
and time-sensitive (“SMART”) (Shahin & Ali Mahbod, 2007).
Table 2-2: Requirements of a key-performance indicator (Shahin & Ali Mahbod, 2007).
Requirement Description
Speci?c KPIs should be detailed and as speci?c as possible.
Measurable A KPI should be measurable against a standard of performance and a standard of
expectation.
Attainable The goal of a KPI should not be out of reach. They should be reasonable and attainable.
Realistic A goal should be realistic taking into account the particular working environment.
Time sensitive Goals should have a time frame for completion, to monitor the progress.
2-2-5 KPI Categories
The previous sections illustrated that ?rms manage their business by measuring key-performance indicators, and
that these indicators should represent – whether indirectly – the ?rm’s strategy and stakeholders’ needs. Because
not every ?rm executes the same strategy and not each ?rm has the same stakeholder groups, di?erent ?rms will
apply di?erent KPIs for performance measurement (Shahin & Ali Mahbod, 2007). Generally, managers apply
value driver maps to determine the performance metrics that correspond with the ?rm’s speci?c strategy (Gates,
2001). A value driver map is a break-down of the ?rm’s strategy into activities – drivers – that are required
to achieve the ?rm’s strategy. On the top level of a value driver map, drivers are generic and for many ?rms
identical. Examples of generic performance metrics are net result, operating result, operating expenditures and
operating margin. These high-level, mostly ?nancial metrics, are generally applied within ?rms. As indicated
by R. S. Kaplan and Norton (1993), all ?rms should focus on the four perspectives; ?nancial, customer, internal
business and innovation and learning when de?ning performance metrics. However, the authors also note that
“speci?c measures within these categories should be tailored to the ?rm’s strategy” (Ittner, Larcker, & Randall,
2003). Thus, ?rms with di?erent strategies require di?erent metrics. How can we categorise metrics that are
speci?c for each ?rm?
Table 2-3: Categories of Key-Performance Indicators (Ittner et al., 2003).
KPI Category Example KPI
1 Short-term ?nancial results Annual earnings, return on assets, cost reduction
2 Customer relations Market share, customer satisfaction, customer retention
3 Employee relations Employee satisfaction, turnover, workforce capabilities
4 Operational performance Productivity, safety, cycle time
5 Product and service quality Defect rates, quality awards
6 Alliances Joint marketing or product design, joint ventures
7 Supplier relations On-time delivery, input into product/service design
8 Environmental performance Government citations, environmental compliance or certi?cation
9 Product and service innovation New product or service development success, development cycle time
10 Community Public image, community involvement
In order to categorise the many performance indicators that one can think of, Ittner et al. (2003) reviewed
literature in the ?eld of the balanced scorecard, intangible assets, intellectual capital, and value-based
management to ?nd the most applied categories of KPIs. Based on the models and frameworks that have been
developed in these research areas, Ittner et al. (2003) distinguish ten performance categories, being short-term
?nancial results, customer relations, employee relations, operational performance, product and service quality,
alliances, supplier relations, environmental performance, product and service innovation, and community. These
categories are listed in table 2-3. The ?nal column of the table shows example metrics. The classi?cation
of Ittner et al. (2003) clearly takes a multi-actor perspective into account, and is therefore considered as
an appropriate classi?cation in line with the performance prism. The search terms (PERFORMANCE
INDICATORS CATEGORIES), (KPI CATEGORIES), (KPI CLASSIFICATION), (KEY PERFORMANCE
2-2 Registering the Right Indicators 23
INDICATORS) AND (CATEGORISATION) in the scienti?c databases ScienceDirect and JStore, and the
scienti?c search engine Google Scholar did not result in literature containing other KPI classi?cations than
Ittner et al.’s (2003) categories. Business intelligence professionals from KPMG have acknowledged that the ten
categories are representative for the KPIs that are actually used by ?rms in practice.
Stakeholder
Satisfaction
Strategies
Processes
Capabilities
Stakeholder
Contribution
1. Short-term financial results
2. Customer relations 3. Employee relations
4. Operational performance
5. Product and service quality
6. Alliances
7. Supplier relations
8. Environmental performance
9. Product and service innovation
10. Community
Figure 2-4: The ?ve Performance Prism Perspectives and corresponding Key-Performance Indicator Categories
As described earlier, performance indicators, and especially key-performance indicators should contribute to
a ?rm’s strategy. And, as illustrated by R. S. Kaplan and Norton (1992), performance metrics should be
established from multiple perspectives, that is, not only ?nancial ?gures. Furthermore, Neely and Adams
(2005) argued that ?rms should formulate performance metrics from a multi-actor perspective. Therefore,
the performance metrics categories established by Ittner et al. (2003) must somehow relate to one of the ?ve
performance prism perspectives. Figure 2-4 schematically shows the relations between the ?ve performance
prism perspectives and the key performance categories. For readability issues, the ?ve perspectives have been
visualised in a pie chart, rather than in a prism. For an explanation of the assignment of the KPI categories to
the ?ve performance prism perspective, see appendix A. The performance categories of Ittner et al. (2003) allow
us to systematically assign social media posts to one of the ten categories. As a result, we can draw conclusions
from the applicability of social media data for certain KPI categories. For the remainder of this thesis, we will
apply the categorisation of Ittner et al. (2003) to categorise key-performance indicators.
24 Conceptual Frame of Research
2-3 Processing: From Data to Information
The registering of signals results in raw data which needs to be processed before it represents information.
Figure 2-1 showed the business intelligence cycle. In the second phase of BI, registered signals are processed.
Van Beek (2006) describes this process as a cycle on itself, which is discussed in this section. It is important
to understand the theory underlying the processing of signals when considering to apply social media data for
BI purposes, because an organisation usually applies business intelligence already. A social media component
should hence be consistent with the existing system(s) and process(es). Van Beek (2006) distinguishes 15
activities making up the processing of gathered data and turning it into information. Figure 2-5 shows the
activities in the BI cycle.
Register
React
Process
Combination
Distribution
E
v
a
lu
a
t
io
n
Operational
Tactical
Strategical
C
o
n
t
e
x
t
u
a
l
T
r
a
n
s
a
c
t
io
n
a
l
E
n
v
ir
o
n
m
e
n
t
1. Collecting
2. Filtering
3. Combine
4. Aggregating
5. Visualising
6. Interpret
7. Internalise
8. Revise &
Recalibrate
9. Verify
10. Enrich
11. Share &
Communicate
12.
Remember
13. Decide
14. Distribute
15. Anticipate
on Changes
Figure 2-5: Processing data to information. Based on van Beek (2006).
The ?fteen steps are discussed in the following section.
1. Collecting
Signals, which are stored as data in di?erent systems are collected in a separate system. When considering
social business intelligence, access to di?erent social media platforms – like Twitter and Facebook – is
required in order to collect the data. Each social media platform has its own method of storing data, and
not each platform is as publicly accessible as the other.
2. Filtering
Only signals that contribute to the deduction of information pass the ?ltering process. Data that are
outdated or of poor quality are removed. Especially when applying social media data for BI purposes,
this step requires much e?ort. The data can consist of spam, polluting the data.
3. Combine
The data that is collected and ?ltered on separate systems are combined and integrated into one single
source, so that analyses are based on one ‘version of the truth’.
4. Aggregating
Detailed data are aggregated to a level so that users can quickly understand the data and ?nd information.
5. Visualising
In order to make the data quickly interpretable for the users, the data is visualised.
The ?rst ?ve steps consist of automated activities that convert signals into information. So far, the signals are
translated into information that is now interpretable for the users. The next ten steps of the process consist of
non-automated activities that involve humans to interpret the information, and act on it.
2-4 Sub Conclusion: How Business Intelligence is Applied 25
6. Interpreting
The information generated by the ?rst ?ve steps are interpreted by humans. For example, the automated
process generated a graph showing the amount of sales in a given region over a given time span. The
meaning of the graph is interpreted by the user.
7. Internalise
In this step, the information derived from the interpreting step is combined with other information in the
problem’s context. It is in this step that the real underlying trends and explanatory factors are analysed
so that the information is embedded in one’s cognitive understanding of the system.
8. Revise & Recalibrate
The new information may a?ect existing information. This step ensures that existing information is
revised and adjusted based on the new information.
9. Verify
This step veri?es the new information with other mechanisms. For example, a decrease in market share
is compared with the companies’ turnover development. Whenever turnover increases while market share
decreases, it may indicate an increase of the overall market. If such mechanisms contradict, the process
of turning the data into information has to be checked for errors.
10. Enrich
In this step, the information – graphs, ?gures, numbers, etc. – are enriched by textual explanation of the
information. A decrease in market share, which is visible in e.g. a pie chart, may be enriched by a textual
explanation of two new competitors on the market.
11. Share & Communicate
By sharing and communicating about the information with other members of the organisation, the
information is brought under submission of various perceptions and views.
12. Remember
Some information do not require immediate action. However, the information may be relevant whenever
future information is acquired. It is therefore important that the information is remembered.
13. Decide
This step involves the reaction on the information. Managers decide how they act on the information, for
instance by launching an advertisement campaign, or to sell a part of the organisation.
14. Distribute
The decision in the previous step is generally taken by managers on higher levels of the organisation. The
new information and decisions following from that information are distributed to the right persons in the
organisation in this step.
15. Anticipate on Changes
The new information may be of a negative character, requiring (structural) organisational change. An
organisation should adopt a positive attitude to change according to the new information.
These ?fteen steps describe how a signal is generally translated into information at which managers can act.
When an organisation intends to implement a (sub) system that extracts signals from social media platforms
to derive information, it should be designed according to this method of processing signals.
2-4 Sub Conclusion: How Business Intelligence is Applied
With the rise of a new data source – social media platforms – for ?rms to access customer perceptions, the
question rises how a ?rm should process these data. The process should in any case correspond with existing
business intelligence processes in ?rms. Therefore, it is essential to understand the general business intelligence
process that ?rms adhere to. This chapter reviewed literature in the ?eld of business intelligence, of which the
conclusions are presented in this section.
Though there exist many views on business intelligence, the common aspect is that BI collects and translates
data into information that supports managerial decision-making. BI can be regarded as a process of three
steps; registering data, processing the data into information and reacting on the conclusions derived from
26 Conceptual Frame of Research
G
D
A
Register
React
Process
STRATEGY
BUSINESS MODEL
KEY PERFORMANCE
INDICATORS
Values, mission,
vision, objectives,
goals, plans
Performance
metrics
Revise
strategy
External data (e.g. social media data)
Internal data (e.g. level of inventory)
Stakeholders beliefs,
perceptions, values
External factors
Business Intelligence Process
Strategy Alignment
S
h
o
r
t
-
t
e
r
m
f
i
n
a
n
c
i
a
l
r
e
s
u
l
t
s
C
u
s
t
o
m
e
r
r
e
l
a
t
i
o
n
s
E
m
p
l
o
y
e
e
r
e
l
a
t
i
o
n
s
O
p
e
r
a
t
i
o
n
a
l
p
e
r
f
o
r
m
a
n
c
e
P
r
o
d
u
c
t
a
n
d
s
e
r
v
i
c
e
q
u
a
l
i
t
y
A
l
l
i
a
n
c
e
s
S
u
p
p
l
i
e
r
r
e
l
a
t
i
o
n
s
E
n
v
i
r
o
n
m
e
n
t
a
l
p
e
r
f
o
r
m
a
n
c
e
P
r
o
d
u
c
t
a
n
d
s
e
r
v
i
c
e
i
n
n
o
v
a
t
i
o
n
C
o
m
m
u
n
i
t
y
Figure 2-6: Strategy and Business Intelligence
that information. It is essential that ?rms measure those activities that contribute to its business model and
corresponding strategy, because ‘what’s get measured, get’s done’. The key notion that should be concluded
from this chapter is that an organisation’s strategy should be based on the needs and preferences of the ?rm’s
stakeholders, and that a company’s strategy drives the values, objectives, goals and plans of company, which, in
turn, determine the (key) performance indicators. Therefore, a link between the BI process and an organisation’s
strategy is required. This link is established by de?ning key-performance indicators that are based on the ?rm’s
business model. In turn, the business model should re?ect the organisation’s strategy. This perspective on
business intelligence is schematically presented in ?gure 2-6.
Though each ?rm executes its own speci?c strategy and will consequently measure organisational performance on
speci?c KPIs, we can classify KPIs into ten categories; short-term ?nancial results, customer relations, employee
relations, operational performance, product and service quality, alliances, supplier relations, environmental
performance, product and service innovation and community. Furthermore, ?gure 2-6 illustrates that the BI
cycle can not commence before the ?rm has determined ‘what it should measure’.
Chapter 3
Research Domain
The purpose of this chapter is to describe the research domain. As illustrated in chapter 1, this thesis intends
to develop a procedure that allows ?rms to process social media data for business intelligence purposes. It is
therefore that this chapter explains what social media is, how ?rms currently apply social media, for which
purposes, what social business intelligence is and what developments currently take place in the world of social
business intelligence. Since social media is one of the many applications enabled by Web 2.0, we start with a
description of Web 2.0.
3-1 Web 2.0
Web 2.0 is the generation of web pages that not only provide information, but additionally allow users to interact
with these web pages. In contradiction with the ?rst phase of the web’s evolution, Web 2.0 allows anyone to
create and share content. The content-creating feature makes it that Web 2.0 is also referred to as “the wisdom
web, people-centric web, participative web, and read/write web” (Murugesan, 2007). Web 2.0 allows users to
do more than just retrieve information. Whereas the internet was traditionally applied to read, watch, and buy
products in Web 1.0, it is increasingly utilised to create, modify, share, and discuss content in the Web 2.0 era.
It is the enabling of the creation of user-generated content that distinguishes Web 2.0 from Web 1.0. O’reilly
(2007) – who sees Web 2.0 as “the web as platform” – indicates that the power of Web 2.0 is “collective
intelligence”, and turns the web into a kind of global brain”. The best-known example is probably Wikipedia
(launched in 2001), an online encyclopedia created by the Internet users that contains now 23 million (Wikipedia,
2012) articles. The existence of Wikipedia illustrates that people are willing to share their knowledge with others.
For example, people that searched for ‘Hasbro’s Easy-Bake Oven’ in Web 1.0 would have found a static web
page promoting the product, while in the Web 2.0 era, people also ?nd in the top 5 of search results a warning
that the Easy-Bake Oven may lead to serious burns on hands due to a poorly-designed oven door (A. M. Kaplan
& Haenlein, 2010). These warnings have been written by users feeling that they had to share their (negative)
experiences and product knowledge. Other examples of collective intelligence are ask and answering sites,
demonstration videos, and product reviewing sites. Not only knowledge is shared with the rest of the world
through Web 2.0 solutions, people also are willing to share their opinions about various topics, their favourite
dining places, what they think about the new election candidate, etc. Especially these topics, that we can
position under the denominator of opinions and private events, are shared through means of social media.
3-2 Social Media
Social media allow users to connect and share content with each other through Web 2.0 based platforms. The
rise of social media in the ?rst decade of the 21
st
century is a natural consequence of Web 2.0. The relation
with Web 2.0 is highlighted by Kaplan & Haenlein’s (2010) de?nition of social media, stating that “social media
is a group of internet-based applications that build on the ideological and technological foundations of Web 2.0,
and that allow the creation and exchange of user-generated content”. However, with this de?nition we cannot
distinguish social media from Web 2.0 sites, since user-generated content is a fundamental element in Web 2.0
28 Research Domain
anyhow. Why is it that we refer to social media when we talk about Facebook or Twitter, but to Web 2.0 when
considering Wikipedia? Kietzmann et al. (2011) state that social media “allow individuals and communities to
share, co-create, discuss, and modify user-generated content”. Again, we see the importance of user-generated
content when de?ning social media. But Kietzmann et al.’s (2011) de?nition contains an important additional
component that allows us to distinguish social media from Web 2.0; communities.
Within social media – and especially in social networking sites – users can connect with other users, so that they
can share (personal) information. It is the aspect that allows people to connect with each other that distinguishes
social media from Web 2.0. The social element, connecting with other users, is much more existent in social
media than in Web 2.0. Whereas in Web 2.0, user-generated content is accessible to anyone, in social media
people can restrict this accessibility to people they have selected beforehand. Therefore, we de?ne social media
as Web 2.0 based applications that allow users to create and share user-generated content with pre-selected
users and communities.
3-2-1 Social Media Platforms
The web applications through which users can connect and share content with each other are called social media
platforms. There are many social media platforms available, and the range of social media platforms is vast and
growing (A. N. Smith et al., 2012; A. M. Kaplan & Haenlein, 2010; Hanna et al., 2011). These platforms di?er
in scope and functionality. In turn, there is variation in how people use these platforms. “Some sites are for
general masses, like Friendster, Hi5 and Facebook. Other sites, like LinkedIn, are more focused on professional
networks. Media sharing sites such as MySpace, YouTube, and Flickr concentrate on shared videos and photos
(Kietzmann et al., 2011).
In order to support managers in understanding social media, and to select the right platform for the ?rm’s
purpose, researchers have tried to classify the di?erences between the social media platforms. Weinberg and
Pehlivan (2011) distinguish social media platforms based on two dimensions; (i) half-life of information and (ii)
information depth. The half-life of information refers to the “longevity of the information in terms of availability
/ appearance on the screen and interest in a topic. Depth of information refers to the richness of the content,
and the number of diversity of perspectives” (Weinberg & Pehlivan, 2011). As such, Weinberg and Pehlivan
(2011) positioned popular social media platforms in their framework (?gure 3-1). Micro-blogs, such as Twitter,
allow users to use a limited number of character in each post and therefore have a shallow information depth. On
the other hand, community sites purposed to extensively discuss topics among users have a higher information
depth.
Blogs
(e.g. WordPress)
Communities
(e.g. MacRumors)
Micro-Blogs
(e.g. Twitter)
Social Networks
(e.g. Facebook)
Long
Short
Shallow Deep
Half-life of information
Information
Depth
Figure 3-1: Social media by information half-life and information depth (Weinberg & Pehlivan, 2011).
Kietzmann et al. (2011) created a framework that distinguishes social media platforms based on seven building
blocks. These blocks are “constructs that allow us to make sense of how di?erent levels of social media
functionality can be con?gured”. The functional building blocks are shortly discussed below, and are applied
3-2 Social Media 29
on Facebook, LinkedIn and Twitter in ?gure 3-2 (page 30) to illustrate di?erent focus points on di?erent social
media platforms.
1. Identity
This block represents the extent to which users reveal their identities in a social media setting. Especially
on self-branding platforms, such as LinkedIn, identity is a strong aspect.
2. Conversations
This block represents the extent to which users communicate with other users in a social media setting.
Some sites are much more intended to facilitate conversations – like Twitter – than others.
3. Sharing
Sharing represents the extent to which users exchange, distribute, and receive content. Especially on
Twitter, people share what they are doing, what they think of, etc.
4. Presence
The presence block represents the extent to which users can know if other users are accessible. It includes
knowing where others are, like ‘check-ins’ at Facebook or Foursquare.
5. Relationships
This block represents the “extent to which users can be related to other users”. Related implies “some
form of association that leads them to converse, share objects of sociality, meet up, or simply just list each
other as a friend or fan”.
6. Reputation
Reputation is the extent to which users can identify the standing of others.
7. Groups
The groups functional block represents the extent to which users can form communities and
sub-communities.
In the following sections we discuss three important social media platforms that are part of the analysis in this
thesis, Twitter, Facebook and Blogs.
Twitter is recognised as being the site on which users ask information and complain. Twitter is a micro-blogging
site, designed to let people post short – 140 character – text updates called ‘tweets’ to others. Twitter prompts
users to answer the question ‘what are you doing?’, leading to a constantly updated timeline of short messages
that range from humor, opinions, musings on life to links and breaking news. Kietzmann et al. (2011) argues
that Twitter posts are “mostly short status updates of what users are doing, where they are, how they are
feeling, or links to other sites”. Participants choose Twitter accounts to ‘follow’ in their stream, and they each
have their own group of ‘followers’. Unlike social networks like Facebook and LinkedIn, where a connection is
bidirectional, Twitter has an asymmetric network infrastructure of followers. The site was launched in 2006,
and broke into the mainstream in 2008 – 2009, when accounts and media attention grew exponentially (Marwick
& Boyd, 2011)”. In February 2012, Dugan (2012) announced that Twitter had over 500 million users registered.
Twitter is an important phenomenon from the standpoint of its incredibly high number of users.
According to Jansen et al. (2009), around 20% of all tweets contain mention of a brand. Of these brand-related
tweets, nearly 20% express a brand sentiment, of which 50% were positive, and 33% were critically. In 2010, the
number of Twitter followers per ?rm increased by 241% over the year (Kirtis & Karahanb, 2011). Acknowledged
by A. N. Smith et al. (2012), Twitter posts contain more brand-related information than Facebook and YouTube.
Since the purpose of this thesis is to contribute to the development of social business intelligence in ?rms, Twitter
is a social media platform that is part of the analysis.
Facebook is the absolute number one social networking site. Though it was only founded in 2004, it is ranking
second in the most popular websites in the world these days. In July 2012, the website reported to have 955
million monthly active users (Sloan, 2012) who log on at least once every 30 days. Half of these active users
30 Research Domain
PRESENCE
RELATIONSHIPS SHARING
IDENTITY
REPUTATION CONVERSATIONS
GROUPS
PRESENCE
RELATIONSHIPS SHARING
IDENTITY
REPUTATION CONVERSATIONS
GROUPS
PRESENCE
RELATIONSHIPS SHARING
IDENTITY
REPUTATION CONVERSATIONS
GROUPS
Figure 3-2: Di?erent social media serve di?erent purposes. Based on Kietzmann et al. (2011).
3-2 Social Media 31
log on every day (Laroche, Habibi, & Richard, 2012). Facebook users can “create pro?les featuring personal
information, interests, photos, and the like, and can “friend” other site users. They can also participate in a wide
range of activities such as writing on friends’ walls, commenting on links, participating in forum discussions,
and “liking” brands. Facebook allows people to build or maintain social capital, communicate with others, keep
up with other peoples’ lives, and discover rumours and gossip” (A. N. Smith et al., 2012).
Over 2010, the percentage of corporations active on Facebook increased by 13%, with the number of ‘likes’
per page rising by 115% globally (Kirtis & Karahanb, 2011). Where Twitter is considered as a platform for
companies to communicate instantly with stakeholders, Facebook is suitable for creating communities among
stakeholders.
Blogs
While Twitter and Facebook conversations are often unstructured and brief, blogs may be a source for
organisations to discover structured customer opinions. After a slow start in the late 1990s, weblogs (“blogs”)
– such as Blogspot and Wordpress based websites – have become very popular, because they are easy to create
and to maintain (Kietzmann et al., 2011). Blogs are often designed as product review sites, where customers
can share their experiences with their products. Generally these blogs are publicly accessible and there are no
restrictions to the amount of characters per post.
3-2-2 User-Generated Content on Social Media Platforms
Social media platforms exist by the virtue of user-generated content (“UGC”). UGC is content that is “publicly
accessible, created outside of professional practices and shows a certain amount of creative e?ort” (A. M. Kaplan
& Haenlein, 2010). UGC can take many forms, such as pictures on Facebook, videos on YouTube, statements
on Twitter, product experiences on blogs, etc. Without users that create and share content, social media sites
are like an empty shell. Therefore, UGC is considered as the fundamental element underlying social media
(A. N. Smith et al., 2012; Boyd & Ellison, 2007). Enabled by Web 2.0, user-generated content has become
increasingly popular on the internet since the early 2000s: more and more users participate in content creation,
rather than just consumption (Agichtein et al., 2008). In China, the percentage of internet content that is
user-produced now exceeds that what is professionally produced (A. N. Smith et al., 2012). According to
A. M. Kaplan and Haenlein (2010) it are not only the technical developments of Web 2.0 and an “increased
broadband availability and hardware capacity” that has contributed to the popularity of UGC on social media
these days, but also “the rise of a generation of ‘digital natives’ and ‘screenagers’ with substantial technical
knowledge and willingness to engage online”.
Though UGC varies in nature between di?erent social media platforms (Kietzmann et al., 2011), A. N. Smith
et al. (2012) and Jansen et al. (2009) indicate that much user-generated content – around 20% – on the internet
contains a brand name. It is here where the opportunities for ?rms materialise, ?rms can inspect the posts that
contain a brand name to discover customer opinions related to their brands. The internet’s accessibility, reach,
and transparency have empowered ?rms that are interested in consumers opinions (Kozinets et al., 2010). User
opinions were not that easy to be gathered before the social media era, while they are now accessible at low
costs (Kirtis & Karahanb, 2011).
As illustrated in the previous paragraph, UGC on social media contains opportunities for ?rms. However,
user-generated content also contains disadvantages. Especially issues related to variance, cohesion and
veri?cation are at stake when processing user-generated content from social media sites. The three issues
are discussed in the following paragraphs.
Variance Firstly, the variance in the quality of UGC is high, “any data can contain information ranging from
excellent to spam” (Agichtein et al., 2008). This makes the tasks of ?ltering and ranking the importance of
social media posts more complex than non user-generated content.
Cohesion Secondly, professionals that base decisions on social media content should be aware of the negative
e?ects arising from cohesion; one negative message about a ?rm – which may not even be true – can snowball
over the internet, reaching many people, and may eventually harm the performance of the company. “Cohesion
describes the phenomenon that evaluations of cost and bene?t associated with prospective behaviour are
aligned via strong communication relationships. People thus become more homogeneous as a result of direct
32 Research Domain
contact via social networking links from node to node. This type of social contagion is typically referred to as
word-of-mouth” (Takac, Hinz, & Spann, 2011).
Cohesion can work out positive as well as negative for an organisation. The negative characteristic materialises
whenever a negative message circulates along the social network. This message is likely to in?uence the
perceptions of the organisation in a negative manner. On the other hand, cohesion may o?er opportunities.
Whenever an organisation intentionally in?uences the discussions on social networks in a positive manner, the
message is likely to be adopted by a large community.
Veri?cation Thirdly, the providers of the information in the social media world generally spread information
without veri?cation, unlike the traditional mass media. Dong-Hun (2010) argues that social media is not yet
capable of replacing the traditional media because of the credibility problem. However, Wikipedia is a successful
example of a website that is based on trust, and established and maintained by the crowd. Information posted
on the website is veri?ed among other users, that directly renovate incorrect information.
The identi?ed threats related to social media data should be considered when developing and implementing a
social business intelligence system.
3-2-3 Current Applications of Social Media in Firms
Both large and small organisations are increasingly visible on social media platforms. In addition, managers
“sense that social media is and will remain an important fabric of commerce” (Weinberg & Pehlivan, 2011).
A Burson-Marsteller research investigated the application of the platforms Twitter, Facebook, YouTube and
Corporate blogs, and found that 25% of the ?rms actively use all four social media platforms, 84% uses at least
one of them (Kirtis & Karahanb, 2011). Of the Fortune 2000 companies, 69% currently use social networking
sites, while 37% planned to use more of them over the next ?ve years (McCorkindale, 2010).
What does it mean when ?rms ‘use’ social media? IBM (2011) researched for which activities ?rms applied social
media, the results are shown in ?gure 3-3. Although IBM (2011) provides detailed insight in the many social
media activities that ?rms employ, we can conclude that ?rms generally apply social media to communicate
with customers, promote activities, monitor the brand name and inspect customer ideas. These four activities
are discussed in the following paragraphs in more detail.
27%
35%
37%
38%
40%
40%
41%
43%
43%
46%
46%
48%
50%
52%
60%
65%
74%
Vendor or partner communications
Customer-to-customer interactions
Training/education
Experts insights/though leadership
Solicit customer ideas
Provide support
Employee-to-employee interactions
Customer research
Recruit employees
Capture customer data
Brand monitoring
Solicit customer reviews
Sell products/services
Generate sales leads
Promote events
Respond to customers questions
Communicate with customers
27%
35%
0% 10% 20% 30% 40% 50% 60% 70% 80%
Vendor or partner communications
Customer-to-customer interactions
% of Respondents applying the social media activity n =351
Figure 3-3: Applications of social media (IBM, 2011)
Marketing
Web 2.0, and especially social media, has empowered the ‘voice of the customer’. Consumers are no longer merely
passive recipients in the marketing exchange process (Hanna et al., 2011). In the past, marketing campaigns
3-2 Social Media 33
were typically developed by companies in-house, without interference of (potential) consumers. Campaigns had
the character of ‘here is the advert, please absorb it’, or ‘here is the product, we hope you like it’. This ‘we talk,
you listen’ approach has been replaced by ‘you talk, we listen’ as a result of the possibilities o?ered by social
media (e.g., Patterson, 2012; Klassen, 2009).
Research indicates that marketing through social media is e?ective for ?rms. “70% of the consumers that have
used social media websites to take product or brand information, 49% of these consumers made a purchase
decision based on the information they pound through social media sites” (Kirtis & Karahanb, 2011).
Firms are also turning into social media marketing to lower the ?rms’ expenditures. The cost reduction aspect of
online marketing as compared to traditional marketing is one of the main reasons why companies are nowadays
applying social media for marketing purposes (Kirtis & Karahanb, 2011). Cost reduction is mainly achieved by
the elimination of the distribution phase, which is required in traditional mass media. In addition, marketing
through means of social media is less expensive than the regular channels because most social media applications
are free of charge. As such, the biggest expenditures related to the execution of a social media strategy represents
the time employees spend on posting messages, responding to comments and blogging. Social media allows
marketers to speci?cally target on client groups, and distinguish between products / services case by case. In
comparison with traditional marketing channels, social media shows also on this aspect lower costs. Driven by
the global recession, many ?rms are in a cost-reducing mode. Because of the economic turmoil social media is
applied as a survival tool by many ?rms, so the economic recession has increased the rate of shift change from
traditional media to social media (Kietzmann et al., 2011).
For the bene?ciary reasons of marketing through social media described in the previous, A. N. Smith et al.
(2012) estimate that the percentage of companies using social media for marketing is expected to reach 88% by
2012, up from 42% in 2008.
Customer Relations Management
Firms also use social media for customer relations management (“CRM”), also referred to as social CRM. Thanks
to social media, “the nature of public relations and how organisations engage their public has changed a great
deal in the past several years” (McCorkindale, 2010). “An environment in which control of the relationship
has shifted to the customer, who has the power to in?uence his or her social network” (IBM, 2011) drives
organisations willing to participate in the online conversations.
Social media platforms hold unprecedented opportunities for companies to get closer to customers, allowing ?rms
to communicate directly with customers, for instance to provide support when customers encounter problems
with products / services. According to Patterson (2012), ?rms “have made progress in conversing with their
customers”.
A recent study by Laroche et al. (2012) indicates that it is bene?ciary for ?rms to establish online communities
in which both ?rms and customers communicate with each other, “brand communities established on social
media have e?ects on customer/product, customer/brand, customer/company and customer/other customer
relationships, which in turn have positive e?ects on brand trust, and trust has positive e?ects on brand loyalty”.
Reputation Management
A failing social media engagement strategy can signi?cantly impact a ?rm’s reputation and sales (Kietzmann
et al., 2011). The increased application of social media has serious consequences for an organisation’s exposure
to its environment. It seems that the power has been taken from the corporate marketing departments by
individual consumers that create, share and discuss online blogs, tweets, Facebook entries, movies, pictures, etc.
With or without permission, communication about brands will happen. In an environment where customers gain
more and more power, organisations need to carefully tread their actions and control its’ exposure. Therefore,
companies more and more empower employees to talk, listen, and respond to what consumers post on social
media (A. N. Smith et al., 2012) in order to control the ?rm’s (online) reputation.
One negative message about an organisation – created by one single person – can snowball over the internet,
reaching many people, and may eventually harm the performance of the company. In 2008, Canadian singer
Dave Caroll wrote a song about United Airlines’ luggage handling employees recklessly throwing his guitar,
which caused a break in his guitar. Frustrated by bad customer experience, he uploaded his ‘United breaks
guitars’ song on YouTube. Consequently, United Airlines experienced a 10% drop (Patterson, 2012) in its share
value and su?ered damage to its reputation. The YouTube clip has been viewed over 12 million times. This is
34 Research Domain
one of the examples that show how powerful the force of social media can be, when a company does not act
according to the preferences of the community. As such, social media platforms may be a source of both threats
and opportunities for brands experiencing unfavourable exposure (A. N. Smith et al., 2012).
Co-creation & Pro-sumers
Today, consumers “are taking an increasingly active role in co-creating everything from product design to
promotional messages” (Berthon, Pitt, McCarthy, & Kates, 2007). This phenomenon is known as co-creation,
and more recently termed as “prosuming” (DesAutels, 2011), illustrating that people are not only consumers
but at the same time producers. Firms are much more required to perceive consumers as partners in the process
of creating products, whereas this was – before the social media era – formerly an activity for solely the ?rm.
An example of such a process is Lay’s recent campaign to decide the new ?avour of their potato crisps. In
Lay’s campaign, consumers were stimulated to contemplate new ?avours and to post these ideas on the web.
Other users consequently rated the ideas that were send it. The winning ?avours have actually been brought to
production. Another example related to co-creation is Samsung, which ‘listened’ closely to the user-generated
content on blogs, and, after hearing complaints that the speakers on the side of the TV were too wide for many
customers’ entertainment cabinets, it redesigned the product (Klassen, 2009).
The co-creation opportunities for ?rms o?ered by social media reach even further. A growing number
of organisations, among them 3M, AEGON, HCL Technologies, Red Hat and Rite-Solutions have recently
experienced with crowdsourcing their strategies (Gast & Zanini, 2012). The organisations o?ered the public the
possibility to provide input in the form of proposals for the company’s future directions. The e?ects resulting
from strategy crowdsourcing is twofold. In the ?rst place, the company gathers information from the external
environment, including perceptions from important actors that would normally be overlooked. An organisation
can consequently craft its strategy with a higher quality. Secondly, the organisation creates “enthusiasm and
alignment behind a company’s direction” (Gast & Zanini, 2012).
Though the previous sections illustrate that social media is widely applied for di?erent purposes in organisations,
many executives eschew or ignore this form of media because they “don’t understand what it is, the various
forms it can take, and how to engage with it and learn” (Kietzmann et al., 2011). Also A. M. Kaplan and
Haenlein (2010) state that the reluctant attitude of some managers towards social media is due to “a lack
of understanding regarding what social media are”. Many organisations acknowledge the opportunities in the
application of social media, while, on the other hand, there also exists a fair degree of uncertainty with respect
to allocating e?ort and budget to social media, and “limited understanding of the distinctions between various
social media platforms” (Weinberg & Pehlivan, 2011).
3-3 Social Business Intelligence
Firms should measure the e?ects of social media activities on organisational performance. As illustrated in
chapter 2, the process of business intelligence requires key-performance indicators to be de?ned so that the
performance of the ?rm can be measured against its strategy. This value-based management approach is
generally applied within ?rms, implying that when a ?rm pursues to perform social media activities, it should
measure the e?ects of these activities in relation with organisational performance.
Existing social media monitoring tools mainly reveal the performance of the organisation on social media
(number of mentions, number of likes, % of positive mentions), and treat the social media component of a ?rm
as a separate business unit executing its own strategy. However, the purpose of business intelligence is to reveal
the underlying parameters that determine the performance of the organisation, that is, not limited to solely
social media performance. In order to understand the in?uence of social media content on an organisation’s
performance, a link between the company’s key-performance indicators and clear social media parameters is
required.
It is argued that the possibilities of social media for business intelligence purposes reaches further than what is
currently o?ered by the social media analytics tools. The key bene?ts will be gained whenever the KPIs of an
organisation are linked to the parameters that are measured by social media tools. Only in that case, one can
speak about ‘social business intelligence’. In social business intelligence, the social media activities related to a
?rm are translated to organisational performance.
3-3 Social Business Intelligence 35
3-3-1 The Current State of Social Business Intelligence: Early Adoption
Software developers acknowledge the opportunities generated on social media platforms for ?rms. With the
rise of social media, and the popularity of BI within organisations, software solutions o?ering social media
‘intelligence’ are emerging rapidly. As a result, tools for analysing information become widely available at
ever-lower prices (Bughin, Chui, & Manyika, 2010), some are even o?ered for free.
Auditore (2012a) – the former head of SAP’s Business In?uencer Group and now researcher at Asterias research
– investigated the market for social business intelligence and found that the top four emerging SBI platforms
consists of Radian 6, Kapow, evolve24 and NetBase. According to Kapow (2009), a provider of business
intelligence software, we are at a point in time where social media can be integrated into enterprise business
intelligence platforms. Not only small software development ?rms are on a discovery journey, well-established
companies o?ering total business intelligence solutions are also embracing social media data. For example, SAP
collaborates with NetBase to o?er social media analytics. IBM’s Cognos provides social network capabilities.
Oracle recently acquired Involver, Vitrue and Collective Intellect to add social media analytics to their portfolio
of services. SAS incorporated social media analytics in its platform, and QVSource allows QlikView users to
extend their BI platform with social media intelligence. Table 3-1 lists the top existing, new and emerging
vendors of (social) business intelligence solutions.
Table 3-1: Top (Social) Business Intelligence Vendors (Auditore, 2012a).
Legacy BI vendors New social media BI vendors Emerging social media BI vendors
1. IBM 1. Google 1. Radian6
2. Oracle 2. SAS 2. Kapow
3. SAS Institute 3. IBM 3. evolve 24
4. SAP 4. NetBase
Emerging Social Media Business Intelligence Vendors
Companies that apply social media in their organisation generally apply a cycle consisting of three steps;
(i) monitoring, (ii) analysing, and (iii) engaging (Kapow, 2009; Bryant, 2011) using social media monitoring
platforms. The objective of these platforms is to ‘listen’, in order to monitor the brand(s). Generally,
“automated scripts monitor a handful of keywords from targeted web sites” (Kapow, 2009). The gathered
data in the listening phase is generally continued by mapping customer perceptions, sentiment measuring and
an indication of the company’s reach respecting social media. In general, these tools are solely based on “simple
quantitative counts of how many times a brand has been mentioned” (Patterson, 2012). Some exceptions exist
that provide a general mood of the brand, often based on large datasets. These functions are referred to as
analytics by the software o?erers. Clients receive weekly or daily reports containing ?gures representing the
amount of last week mentions on social media platforms, the number of likes, the sentiment related to that, the
number of shares, retweets, a distribution of the locations, gender distribution, etc. In addition, the software
platforms generally scan all social media platforms continuously and present all relevant content to the user(s),
via dashboards and/or automatic generated reports. Companies’ managers can consequently engage with the
social media users via one portal. The nature of the engagement of companies is often related to customer
relationship management, e.g. a customer-service department explaining to an individual why his or her credit
card is not functioning, or why the company’s website is not presented properly in the customer’s browser.
Other social media posts made by organisations are often marketing related, e.g. an announcement of a new
product release or an o?er.
The emerging social media intelligence tools – Radian6, Kapow, evolve24 and Netbase – and their features are
presented in the following section.
• Radian 6
Salesforce’s Radian6 provides social media monitoring tools, social media engagement software and social
customer relationship management and marketing software. It provides companies with social analytics
comprising of social media metrics and sentiment analysis. Radian6 provides ?rms with a dashboard
illustrating their performance on social media. Consequently, ?rms can engage in online discussions.
Clients of Radian6 include Fuji?lm, Commerce Bank, KLM, Pepsi, L’Oreal, Baker Tilly and Activision
(Radian6, 2012). Radian6 o?ers clients di?erent packages with di?erent features, ranging from EUR 750
per month to EUR 12,000 per month.
36 Research Domain
• Kapow
Kapow o?ers solutions for accessing, extracting and enriching web data (Kapow, 2009). The software
developer illustrates the applicability of public web data for business intelligence, it explicitly mentions
that it o?ers software that structures social media data and turns it into interpretable information. One of
the many tools o?ered by Kapow is the monitoring of social media platforms. Users of Kapow’s software
include AT&T, Intel, Cisco, Vodafone, Morgan Stanley, P&G, DHL, Barclays, Lenovo and Audi.
• evolve24
evolve24 mines, priorities and scores online conversations so that relevant intelligence is provided to the
manager of a ?rm. The software of evolve24 allows users to create custom dashboards to present those
social media metrics that are relevant to the ?rm. Next, it allows predictive modelling to predict the
impact of certain issues, so that the decision process in the organisation is supported by this information
(evolve24, 2012).
• NetBase
NetBase allows users to track social media issues related to the topics of interest. It processes billions
of social media posts to extract structured insights that enterprises can use to quickly discover market
needs and trends, quantify market perceptions about products, services, and companies (Netbase, 2012).
SAP collaborates with NetBase to for social media analytics solutions. Hence, SAP users can easily
integrate the social media analytics provided by NetBase in their existing BI platforms. Amongst others,
Tupperware, Hewlett Packard, Coca-Cola and Kraft Foods are users of NetBase.
Intelligence Provided by Social Media Monitoring Tools
Social media tools, whether they refer to themselves as monitoring, analysis or intelligence tools, o?er a variety of
insights in the performance of ?rms on social media. The novelty of social media and the even more unexplored
applicability of business intelligence on the new phenomenon makes in that there is little scienti?c literature
available in this ?eld. Instead, both large ?rms as IBM and small organisations present white papers and blogs
in which they describe their view on social media metrics. Many of these documents refer to the same variables
that they measure, though they generally adhere to their own ‘invented’ name. Common social media metrics
and intelligence that are provided by the social media monitoring tools are presented below.
• Volume of Posts
The volume of posts measures number of messages or articles that have been created on social media for
a speci?c topic over a given period. The volume of created social media posts containing a ?rm’s name
(or product / service name) illustrates to what extent a company is subject of discussion of social media.
The volume of social media posts can vary from day to day or even from hour to hour. A sudden upwards
deviation in the average volume is a signal for a ?rm that people are paying attention to the ?rm, whether
positive or negative. Firms can relate this ?gure to marketing campaigns or other organised events to
determine the success of their reach.
• Engagement
Engagement represents the involvement of users in the brand. Often, companies measure engagement by
the amount of likes, followers, shares, retweets, etc. However, solely looking at this ?gure is not enough.
After all, it is relatively easy to in?uence this ?gure, for instance by organising a lottery in which users can
win an iPad. Doeland (2012) distinguishes engagement metrics into distribution metrics and interaction
metrics. Distribution metrics describe how well the organisation is visible to the social media public, while
interaction metrics represent how well the public engages in the brand.
• Sentiment
Almost all software tools – even those available for free – o?er sentiment analysis, a measure that represents
the attitude of the content generated by the social media users. Generally, social media posts are classi?ed
as either positive, neutral or negative by linguistic algorithms. These algorithms ‘simply’ textmine each
post associated to the organisation and consequently connect words and phrases like ‘great’, ‘wow’, ‘good’,
‘
’, ‘super’, etc. with a positive attitude. Posts containing words like ‘bad’, ‘dumb’, ‘worthless’, etc. areclassi?ed as negative posts. As such, an indication of the sentiment under the social media users is
generated. Figure 3-4 illustrates the output of a sentiment analysis as it is provided to a user of a social
media monitoring tool, in this case uberVU, one of the popular social media monitoring tools.
Sentiment analysis is a complex activity. Not only because each language requires its own meta data to
classify words and phrases in di?erent languages, but also because most of the sentiment analysis tools
3-3 Social Business Intelligence 37
Table 3-2: Examples of Engagement Metrics
Distribution Metrics Interaction Metrics
Followers Retweets
Fans Forwarding
Mentions Sharing
Reach Comments
Bookmarks Likes
Inbound links Rates
Blog subscribers Reviews
Contributors
Tra?c generated
Time spent on site
Response time
Sentiment
Sentiment refers to whether the tone of the conversation where the keyword was mentioned was
positive, neutral or negative. As a simplified example, "I love Apple" is considered positive towards
"Apple" whereas "I bought an IPad yesterday" is neutral. We use one of the most powerful fully
automated sentiment engines on the market.
Daily Sentiment breakdown shows the number of positive, neutral and negative mentions each day.
Main Negative Themes are the topics that people talk about negatively when mentioning the
keyword.
So for the mention "I hate Apple support" a negative theme is "support"
AVERAGE SENTIMENT
Slightly POSITIVE
27% positive
SENTIMENT BREAK-DOWN
43.3 43.3% positive % positive 40.3 40.3% neutral % neutral 16.2 16.2% negative % negative
DAILY SENTIMENT BREAKDOWN
Jul 27 Jul 28 Jul 29 Jul 30 Jul 31 Aug 01
2k
4k
Figure 3-4: Sentiment Analysis Example
use Natural Language Processing techniques. These techniques assume that the underlying text is “clean
and correct” (Dey & Haque, 2008), a requirement that is not always present in social media posts. Social
media posts comprise spelling errors, ad-hoc abbreviations and improper casing, incorrect punctuation
and malformed sentences. These features pollute the outcome of the algorithms. However, “interest in
noisy text analytics has increased signi?cantly in the recent past” (Dey & Haque, 2008). The systems
that are currently developed also take phrases into account (Agichtein et al., 2008), turning “Wow, the
new product of ABC is really great.. NOT!” into a negative sentiment post. As such, the accuracy of
sentiment analysis is expected to increase by new methods that are currently developed. Most platforms
are commercial and do not disclose full details of their internal feature set.
• Geography
Whenever a person registers itself for a social media platform, he or she is required to ?ll up some personal
information, including the person’s residence. Though it is not guaranteed that users provide legitimate
personal details, social media monitoring tools use this information to determine the location of where
the posts has been made. Next, mobile devices including a GPS component can – if allowed by the
user – provide the social media post with more accurate geographic information. As such, social media
monitoring tools provide details about the geography of the social media posts of a ?rm in a given period.
Figure 3-5 shows an example of the output of a social media geography analysis.
• Topic and theme detection
Social media monitoring tools provide details in the primary topics and themes that consist in the dataset
related to the ?rm. Generally, a list of the ten most ‘trending topics’ is presented. Topic and theme
detection allows ?rms to quickly grasp an understanding of the most discussed topics that consist in the
social media posts related to the ?rm.
• In?uencer ranking
38 Research Domain
Geolocation
Geolocation represents the countries where people talked about the keyword during the selected
time period and the respective share of the conversation. The location of a person is determined by
using mostly Twitter and Facebook data and other profile or location data where available. The
darker the green, the more conversations have taken place about this keyword in that region of the
world.
GEOLOCATION HEAT MAP
TOP LOCATIONS
JAPAN 17%
UNITED STATES 11%
BRAZIL 8%
INDONESIA 7%
ITALY 6%
TURKEY 4%
SPAIN 4%
MEXICO 3%
SOUTH KOREA 3%
UNITED KINGDOM 3%
TOP LANGUAGES
SPANISH 35%
ENGLISH 22%
PORTUGUESE 10%
JAPANESE 8%
ITALIAN 4%
1 6637
Figure 3-5: Geographic Analysis Example
Almost all social media platforms – and especially social networking sites – provide the possibility to
follow other users. As a consequence, the messages that have been created by people with many followers
will reach many other users. Social media monitoring tools provide insight in the amount of followers
of the people that posted a message containing the ?rms names. The users with the most followers are
considered the strongest in?uencers.
It would be wise to combine the sentiment of the posts with the posts made by the strongest in?uencers.
A negative message created by a strong in?uencer is likely to reach many people, which may result in an
overall decrease of the sentiment towards the ?rm. On the contrary, strong in?uencers posting positive
messages may increase the overall sentiment. With this information, web care teams can focus on the
people with many followers in order to have the strongest e?ect on the desired result, which may be an
increase in the overall sentiment.
• Channel distribution
In order to understand on which social media platforms ?rms are subject of discussion, social media
monitoring tools provide insight in the distribution of the posts related to the ?rm across the di?erent
platforms. With this insight, ?rms can decide to focus on those platforms where their ?rm is subject
of discussion. Figure 3-6 shows an example of the output provided by a social media monitoring tool
(uberVU) illustrating the distribution of social media posts in a certain period across various channels.
A 2012 research by IBM and SHARE-Unisphere amongst 711 business and IT managers from across the
world revealed that “72% of the respondents – ?rms – are monitoring social media networks, re?ecting great
awareness of the importance of understanding information ?ow and engaging social media networks”. The
most mentioned business functions employing social media include “sales and marketing (64%), public relations
and communications (38%), IT (37%) and customer service (37%)” (Auditore, 2012b). Table 3-3 lists the top
business initiatives and the parameters that were measured (Auditore, 2012a). The research also indicated that
investments in the area of social business intelligence “continue to trend upward ... 60% of the respondents
indicated that they expect to increase social media monitoring over the next 1–2 years, while 21% indicated it
would be 3–5 years. However, the study shows, managers are unclear about the ultimate usefulness of social
media. This re?ects that social business intelligence is still in an ‘early adopter’ phase. The study concluded
that “social media based business intelligence represents the next great frontier of data management, promising
decision makers vast vistas of new knowledge gleaned from exabytes of data generated by customers, employees,
and business partners” (Auditore, 2012b).
3-4 EU Legislation on Social Media Data Processing 39
Share of voice
This metric represents the breakdown of mentions about the keyword by specific platforms. The
breakdown is based on total number of mentions per platform. This is important when you're trying
to figure out where most of the conversation is happening and where you should focus your
listening and engagement efforts.
Top stories on the top platforms provides a sense of what people are mostly talking about regarding
the keyword on each individual platform.
PLATFORM DISTRIBUTION
Flickr
3407 mentions
1439 mentions
Blogs
60 mentions
59 mentions
News
9 mentions
Youtube
3 mentions
Boards
1 mention
Vimeo
1 mention
Figure 3-6: Channel Distribution Example
Table 3-3: Top Business Initiatives for Social Media and Measured Parameters (Auditore, 2012b).
Top business initiatives Top metrics employed
1. Brand-reputation management 1. Customer satisfaction
2. Marketing communications 2. Overall chatter
3. Customer service 3. Brand experience
4. Customer experience management 4. Advertising campaign performance
5. Sales
6. CRM
3-4 EU Legislation on Social Media Data Processing
Social networks have obtained a “poor reputation for protection users’ privacy due to a continual ?ow of media
stories discussing privacy problems” (Bonneau & Preibusch, 2010). Examples of such stories include “disclosure
of embarrassing personal information to employers and universities, blackmail using photos found online and
social scams” (Bonneau & Preibusch, 2010). The European Commission is of the opinion that social networks
are a useful tool for staying in touch with friends, family and colleagues, but that these networks also present
a risk that personal information, photos and comments might be viewed more widely than people realise. The
Commission also states that in some cases this can have ?nancial, reputational, and psychological consequences.
Currently, legislation in the European Union’s member states on data privacy is based on the Data Protection
Directive 95/46/EC. This Directive has been established in 1995, a period in which Web 2.0 and social networks
did not exist. The technological developments and the scale of data sharing and collecting have increased in
recent years. Given the advances in IT, the Commission deems Directive 95/46/EC outdated. In addition,
as with any Directive, all member states have composed national legislation based on the Data Protection
Directive, implying that each member state applies its own Data Privacy Policy. E.g, in the Netherlands this
resulted in the Personal Data Protection Act
1
in 2001. It is therefore that the Commission drafted a proposal
in January 2012 for Regulation on the protection of individuals with regard to the processing of personal data.
This new legislation takes the social media era into consideration, and is directly applicable in all EU member
states.
Directive 95/46/EC provides the basis for the de?nition of personal data, which may be contained in social
media messages. Personal data are de?ned as “any information relating to an identi?ed or identi?able
1
In Dutch: Wet Bescherming Persoonsgegevens.
40 Research Domain
natural person; an identi?able person is one who can be identi?ed, directly or indirectly, in particular by
reference to an identi?cation number or to one or more factors speci?c to his physical, physiological, mental,
economic, cultural or social identity”
2
. The processing of personal data is de?ned as “any operation or set
of operations which is performed upon personal data, whether or not by automatic means, such as collection,
recording, organisation, storage, adaption or alteration, retrieval, consultation, use, disclosure by transmission,
dissemination or otherwise making available, alignment or combination, blocking, erasure or destruction”
3
. The
Data Protection Directive is only applicable when the data can be marked as personal data.
The newly proposed Data Protection Regulation adheres to the personal data de?nition of Directive 95/46/EC.
Thus, any data that provides one the possibility to retrace a natural person from that data, is personal data.
The Commission introduces the ‘right to be forgotten’, implying that a social network user can request – if there
is to legitimate reason to store it – to remove all data related to the person from their system. Personal data
can only be collected after explicit consent of the person that provides the information. Furthermore, providers
of social media should adopt the principle of ‘privacy by default’, implying that the default settings should
be those that provide the most privacy. Social media sites should also inform users about how the personal
data will be used. The new legislation is expected to come into force in 2014, with penalties up to one million
Euro or 2% of the ?rm’s global revenue in case of a breach. In the following paragraph, we discuss how the
new legislation a?ects the possibilities o?ered by publicly accessible data for ?rms and what procedures are
necessary to be in compliance with the new legislation.
3-4-1 What Firms are allowed to do with Public Data
Firms are allowed to process data whenever these data are not personal data or whenever the creator of the
data has given prior consent to process the data. In order to avoid data to be legally labelled as personal, the
data should be pre-processed in a way that it is not possible to retrace a natural person from the data. Thus,
the data should be made anonymous. E.g., attributes containing the name of the users should be removed.
Though it is not guaranteed that persons actually use their o?cial name on social media, it is advised that ?rms
remove those attributes that may contain information allowing one to retrace a natural person from these data.
Furthermore, ?rms can aggregate the data to a level at which the individual message is not considered for their
analyses. The ‘right to be forgotten’ has consequences for the way in which social media data is stored and
distributed. With the new Regulation, any person can withdraw his or her information from a social media site.
However, social network sites distribute – by means of APIs or trough other ways – the social media messages
to third parties. It will be the social network providers that will become responsible to communicate to its third
parties that a certain user has requested to delete its content.
In order to avoid suspicions, it is advised that ?rms intending to process social media data carefully document
the steps that they undertake to make the data anonymous, and how the ‘right the be forgotten’ is enabled in
the processing of the social media data. Such procedures are to be designed so that privacy is embedded in the
procedure, known as the ‘privacy by design’ principle in the new Data Protection Regulation.
3-5 Sub Conclusion
Social media platforms are Web 2.0 based applications that allow users to create and share user-generated
content with pre-selected users and communities. There exists a variety of social media platforms, some are
aimed at relations between the users, while others are developed to share media like photos and videos. Research
indicates that of all the user-generated content on the internet, about 20% is brand-related. Users e.g. write
their opinion about a new product, complain about a service, discuss new ideas, etc. Therefore, it is interesting
to investigate the opportunities for ?rms to analyse the content that contains their brand name. However,
though user-generated social media content may be valuable for a ?rm, there also exist pitfalls in collecting,
analysing and drawing conclusions from these data. Firstly, the variance in the quality of UGC is high; any
data can contain information ranging from perfectly true to spam. Secondly, cohesion may lead to homogeneous
content; implying that one user adopts the opinion of another. Thirdly, one should be aware of the fact that
users post their messages on social media sites without veri?cation.
Firms are increasingly visible on social media. This trend is even ampli?ed by the current global recession,
bringing ?rms in cost-reduction mode. Firms generally apply social media for marketing e?orts, customer
2
Directive 95/46/EC, O?cial Journal of the European Communities. L 281/31, Article 2(a).
3
Directive 95/46/EC, O?cial Journal of the European Communities. L 281/31, Article 2(b).
3-5 Sub Conclusion 41
relations management, reputation management and to stimulate co-creation. For all these four aspects, social
media engagement is a more e?cient and inexpensive activity than the traditional channels.
G
D
A
Register
React
Process
STRATEGY
BUSINESS MODEL
KEY PERFORMANCE
INDICATORS
Values, mission,
vision, objectives,
goals, plans
Performance
metrics
Revise
strategy
Stakeholders beliefs,
perceptions, values
External factors
Business Intelligence Process
Strategy Alignment
S
h
o
r
t
-
t
e
r
m
f
i
n
a
n
c
i
a
l
r
e
s
u
l
t
s
C
u
s
t
o
m
e
r
r
e
l
a
t
i
o
n
s
E
m
p
l
o
y
e
e
r
e
l
a
t
i
o
n
s
O
p
e
r
a
t
i
o
n
a
l
p
e
r
f
o
r
m
a
n
c
e
P
r
o
d
u
c
t
a
n
d
s
e
r
v
i
c
e
q
u
a
l
i
t
y
A
l
l
i
a
n
c
e
s
S
u
p
p
l
i
e
r
r
e
l
a
t
i
o
n
s
E
n
v
i
r
o
n
m
e
n
t
a
l
p
e
r
f
o
r
m
a
n
c
e
P
r
o
d
u
c
t
a
n
d
s
e
r
v
i
c
e
i
n
n
o
v
a
t
i
o
n
C
o
m
m
u
n
i
t
y
Raw messages
Volume of messages
Sentiment
Location of messages
Trending topics
Influencer ranking
Channel distribution
Multiple sources of social media data
Social media data
Unstructured, non-
personal, external data
Engage on
Social media
Personal Data Pre-Processing
Remove or make personal data
anonymous
Figure 3-7: Strategy, Business Intelligence Process and Social Media Data
As with any activity that is performed in a professional organisation, the performance of the activity and its
contributing value to the ?rm’s overall objective is to be measured. Current social media monitoring tools
– which are evolving rapidly – do not o?er insight in the e?ects on organisational performance due to the
organisation’s social media undertakings. Rather, these tools provide the number of brand mentions on di?erent
platforms, the locations of where the posts were made, gender classi?cation, language, (unreliable) sentiment
analysis, etc. and thereby treating the ?rms’ social media activities as a separate, isolated, business activity.
It is therefore argued that ?rms need to establish a clear link between social media metrics (such as number of
likes, shares, sentiment) and the ?rm’s key performance indicators. Figure 3-7 illustrates the concept of social
business intelligence, with the types of social media data ?owing in the business intelligence process.
Chapter 4
Content Analysis
This thesis examines the applicability of social business intelligence for ?rms in di?erent contexts. More speci?c,
the applicability is investigated for ?rms in (i) di?erent industries and (ii) for di?erent customer relation types.
The purpose of this chapter is to reveal di?erences in social media content related to di?erent ?rms. As such,
social media data related to di?erent ?rms will be collected from social media platforms, after which these
data are analysed on these two dimensions. As described in section 1-6-2, we will apply the content analysis
procedure developed by Bos and Tarnai (1999) as a guidance for this analysis. A content analysis “entails a
systematic reading of a body of texts” (Krippendor?, 2004), which is required to analyse the social media posts
related to di?erent ?rms. The structure of this chapter corresponds with the ?ve step procedure, as shown in
?gure 4-1.
Research outline, research questions,
formulation of hypotheses, material to
investigate
Operationalising the categories,
determining the sample, determining the
unit of analysis
Establishment of categories
Theoretical level
Determining reliability and validating the
categories
Pretest
Appropriate statistical analyses
Data collection and evaluation
Immanent interpretation of the results,
discussion of the results on the basis of
the problem
Interpretation of the results
Section 4-1
Section 4-2
Section 4-3
Section 4-4, 4-5
Section 4-6
Figure 4-1: Content Analysis (Bos & Tarnai, 1999) and Outline Chapter 4.
Section 4-1 formulates the research questions that are to be answered by the content analysis, and it describes
the material that will be investigated. Next, section 4-2 establishes the categories that are to be analysed, it
describes the sample ?rms and the sample period. Thirdly, in section 4-3 a pretest is performed to validate
whether it is possible to collect the data and classify it into one of the established categories. If necessary, the
categories will be adjusted. Furthermore, section 4-3 describes the categories in more detail and presents the
taxonomy – or coherence – of the categories. Fourth, in section 4-4 the data – social media messages related
to di?erent ?rms – is collected and the collection process is evaluated. In section 4-5, the data is analysed and
visualised. Section 4-6 interprets the results of the content analysis. Finally, section 4-7 concludes the ?ndings
of this chapter.
4-1 Theoretical Level 43
4-1 Theoretical Level
The ?rst step in the content analysis contains theoretic notions. Hypotheses and the material to be investigated
are determined. Hypotheses are established to explicitly specify what will be researched in the analysis. The
hypotheses and the material to investigate are described in this section.
4-1-1 Hypotheses Formulation
This thesis answers three sub research questions (see page 4), of which the second question will be answered in
this chapter. From the research questions that are to be answered by means of the content analysis, hypotheses
are formulated. These hypotheses explicitly state what will be investigated. In total, four hypotheses are
formulated that correspond to the second research question of this thesis, i.e. In which contexts are ?rms able
to acquire social media data for business intelligence? As discussed in chapter 1 we specify a ?rm’s context on
two dimensions. Firstly, a ?rm’s context is described based on its customer relation type, consisting of either
B2C or B2B. Secondly, a ?rm’s context is speci?ed by the industry in which it is active. The speci?cation of the
?rm’s context is incorporated in the hypotheses that are established in this section. Two groups of hypotheses
are established; volume-related and subject-related hypotheses.
Hypotheses related to the Volume of Social Media Messages
The ?rst topic of investigation relates to the volume of ?rm-related social media messages. Particularly, the
volume of messages related to ?rms performing di?erent customer relations are compared. Throughout this
thesis, we distinguish two types of customer relations; B2C and B2B. As illustrated in section 1-2, it is expected
that ?rms performing B2C relations are more often subject of discussion on social media than B2B ?rms.
Firms performing B2C relations generally have more customers than B2B ?rms and are more visible to the
end-consumer than B2C ?rms. We expect that these aspects in?uence the amount of social media messages
related to a ?rm. The reason why the volume of social media messages related to di?erent ?rms is important
to investigate is the fact that social business intelligence will only be possible for a ?rm in case that there are
actually messages created that are related to the ?rm. Hence, the ?rst hypothesis is formulated as:
H
1
: The volume of ?rm-related social media messages is higher for B2C ?rms than for B2B ?rms.
Secondly, it is expected that the volume of ?rm-related messages di?ers among ?rms active in di?erent industries.
As illustrated in section 1-2, the nature of the products and/or services traded in di?erent industries a?ects
the rate at which products/services are sold. Therefore, we expect that ?rms in some industries are more often
subject of discussion on social media than ?rms in other industries. For instance, retail products are more
frequently bought by people than houses. Taken into account this rational reasoning, the second hypothesis is
formulated as:
H
2
: The volume of ?rm-related social media messages di?ers between industries.
Whereas the customer relation types have yet been operationalised by two groups (B2B or B2C), the industry
types have not yet been established. In section 4-2, the industry categories will be established based on a generic
classi?cation.
Hypotheses related to the Subjects of Social Media Messages
The third aspect that is investigated in the content analysis relates to the subjects of social media messages.
Whereas the ?rst two hypotheses provide insight in the existence of ?rm-related social media messages, this
insight is not su?cient to draw conclusions on the applicability of social business intelligence. The subjects
of social media messages are also to be included in the analysis. It is important to research the subjects of
social media messages related to ?rms since the subjects of the messages are used to assign the messages to
key-performance indicators. As such, the subjects contained in the messages determine – in combination with
the volume – the applicability of social business intelligence for ?rms.
The subjects of ?rm-related social media messages are investigated on the same two dimensions as the volume
of the messages. The contexts of B2C ?rms di?er from B2B contexts. It is therefore likely that the subjects
44 Content Analysis
of the messages related to B2C ?rms di?er from the subjects of B2B related messages. So far, we do not
have strong signals that certain subjects are more often discussed in one customer category than in the other.
In order to understand which type of ?rms can ?nd messages related to di?erent KPIs, the third hypothesis
investigates whether or not the subjects of ?rm-related social media messages di?er between ?rms performing
di?erent customer relation types. Hence, the third hypothesis is formulated as:
H
3
: The subjects of ?rm-related social media messages di?er between ?rms performing B2B and
B2C relations.
For similar reasons that are concerned with the second hypothesis, the subjects of messages related to ?rms
in di?erent industries are expected to vary. For example, it is likely that messages related to user experiences
are more frequent created in an industry that creates electronic consumer products as compared to an industry
that delivers consulting services. The subjects contained in social media messages – combined with the volume
of these messages – a?ect the opportunities for social business intelligence for ?rms. Since this thesis examines
the opportunities for social business intelligence for di?erent ?rms, it is necessary to investigate variations in
the subjects related to ?rms in di?erent industries. Consequently, the last hypothesis is formulated as:
H
4
: The subjects of ?rm-related social media messages di?er between industries.
The content analysis will be designed according to these hypotheses, and the results of the content analysis
allow us to con?rm or reject the four hypotheses. As such, we can draw conclusions on the applicability of social
business intelligence for ?rms in di?erent contexts.
4-1-2 Material to Investigate
The second step in the theoretical level consists of a description of the material to be investigated. In this
thesis, we investigate social media posts that are related to ?rms. As described in chapter 3, di?erent social
media platforms serve di?erent purposes, leading to di?erent type of posts. A 140 character tweet has a lower
information depth than e.g. a product review site. Since this research is exploratory in nature, it is valuable to
gain as much understanding as possible from the content posted on di?erent social media platforms. Therefore,
the material to investigate is sourced from popular social media sites in Western Europe. The content in the
dataset is sourced from Twitter, Facebook’s public pages, Flickr, Newssites, Google+ public pages, (Wordpress)
Blogs, Picasa, YouTube and Friendfeed.
4-2 Establishment of Categories
In the second step of the content analysis, the categories to be analysed are established and the sample set is
determined. Depending on the research, the categories will di?er. In this thesis, the categories that are to be
analysed consist of ?rms in di?erent industries and with di?erent customer relations. Next, the content analysis
of this thesis requires categories of social media posts to draw conclusions on the applicability of social media
posts for business intelligence.
4-2-1 Operationalising the Categories
We examine di?erences in the volume and subjects of social media messages related to ?rms on two nominal
dimensions; (i) industries and (ii) the customer relation type. The industry dimension is operationalised by
categorising ?rms in di?erent industries. The relation with end-users dimension is operationalised through
means of a distinction between either Business-To-Business ?rms or Business-To-Consumer ?rms.
(i) Industry Classi?cation
CBS (2012) – Statistics Netherlands – provides a hierarchical classi?cation of economic activities, called SBI
1
.
The European Union also has a classi?cation, called NACE
2
, on which SBI is based. SBI allows to classify
?rms based on their economic activities. SBI distinguishes multiple levels, of which the most aggregate level
distinguishes twenty main activities. These activities are listed in table 4-1, and are engaged to classify the
?rms that have been selected in the analysis of this thesis.
1
Standard Industry Classi?cation, in Dutch Standaard Bedrijfsindeling.
2
Statistical Classi?cation of Economic Activities in the European Community, in French Nomenclature statistique des Activités
économiques dans la Communauté Européenne.
4-2 Establishment of Categories 45
Table 4-1: General Classi?cation of Firms (CBS, 2012)
Industry
A Agriculture, forestry and ?shery
B Mining and quarrying
C Industry
D Production and distribution of and trade in electricity, gas, steam and air
E Extraction and distribution of water, sewerage, waste management and remediation
F Construction
G Wholesale and retail
H Transport and storage
I Accommodation, meals and drink provision
J Information and communication
K Financial institutions
L Real estate
M Consultancy, research and other specialised business services
O Public administration
P Education
Q Health and welfare
R Culture, sport and recreation
S Other services
T Households as employers
U Extraterritorial organisations and bodies
(ii) Customer Relation Type Classi?cation
As illustrated in chapter 1, the type of customer relations is likely to have an e?ect on the availability of social
media data related to a ?rm. Therefore, a category describing the type of customer relation is established.
Based on Turban et al.’s (1999) classi?cation of e-commerce, we classify ?rms in either Business-To-Business
(“B2B”) or Business-To-Consumer (“B2C”). This classi?cation will provide insight in the availability of social
media data related to ?rms based on the customer relation type.
Categories of Key-Performance Indicators
The key purpose of this thesis is to link social media posts to organisational performance. As described in
chapter 2, ?rms measure organisational performance based on key-performance indicators. In order to draw
generic conclusions of the applicability of social media data for the purpose of organisational performance,
a generic classi?cation of key-performance indicators is required. Section 2-2-5 illustrated that KPIs can be
classi?ed into ten categories. We will use these categories for the classi?cation of social media posts based on
the subjects of the messages. Table 4-2 shows the categories that are pursued in this research. In section 4-3-1
these categories are described.
4-2-2 Determining the Sample
One of the objectives of this research is to investigate possible di?erences in the user-generated content on
social media related to di?erent ?rms. In order to spot di?erences, our sample exists of ?rms that are active in
di?erent industries and take di?erent positions regarding end-users. This section describes the sample and the
industries that are part of the sample.
Selection of Firms
Based on the industry classi?cation presented in table 4-1, eighteen ?rms have been selected. The starting point
of the sample selection has been the list of ?rms that are part of the Amsterdam stock Exchange (“AEX”). The
main reason for this selection criterion is the fact that these ?rms are stock listed, and hence publicise annual
reports containing information about strategic initiatives, ?nancial ?gures, etc. In case the analysis shows inter
industry di?erences – e.g. between two comparable ?nancial institutions – the annual reports may provide
company speci?c information (e.g. amount of employees, attitude towards social media, etc.) clarifying these
46 Content Analysis
Table 4-2: Categories of Social Media Posts
Category Social media posts . . .
Short-term ?nancial results related to the ?rm’s ?nancial performance
Customer relations from individuals purposed to contact the ?rm, or from the ?rm purposed
to contact an individual
Employee relations related to employees of the ?rm
Operational performance related to the ?rm’s productivity, fact-based statements
Product and service quality related to the experience of products and services
Alliances related to joint-ventures or other cooperations
Supplier relations related to the suppliers of the ?rm
Environmental performance related to environmental / sustainability compliance
Product and service innovation related to innovation
Community revealing the community’s perception of the ?rm (not purposed to
contact the ?rm), chatter
Unde?ned that could not be de?ned in one of the categories
Spam that are not related to the ?rm
di?erences. If a sample containing privately owned companies would have been selected, access to additional
information would have been limited. In addition, ?rms that are listed in the AEX are generally well-established,
visible to the public and regularly subject to news articles. It is therefore expected that these ?rms are subject
of discussion on social media. Furthermore, the ?rm that sponsors this research – KPMG – requested to apply
the analysis on this list of corporations. Table 4-3 lists the selected ?rms, their corresponding industries and
main customer relation. For a description of the individual ?rms and their activities, see appendix B.
In order to design a uniform sample, ?rms of over-represented industries have been replaced by non AEX ?rms
which are also well-established corporations. The distribution of ?rms in the di?erent industries is shown in the
third column of table 4-3. The ?rms have also been classi?ed based on their type of customer relation. Though
the split between B2B and B2C is hard to make for some ?rms because they have B2C as well as B2B relations,
the motivation for the classi?cation of the ?rms is based on its main activities. The main activities, on which
the industry classi?cation as well as the customer relation classi?cation is based, of each ?rm are described in
appendix B. The ?nal column of table 4-3 shows the distribution of B2B versus B2C ?rms in the sample.
Table 4-3: Sample
Firm Industry CBS Relation
1 Akzo Nobel Mining and quarrying B B2B
2 ArcelorMittal Mining and quarrying B B2B
3 Unibail-Rodamco Financial institutions K B2B
4 Arcadis Consultancy, research and other specialised business services M B2B
5 Fugro Consultancy, research and other specialised business services M B2B
6 Coca-Cola Industry C B2C
7 Heineken Industry C B2C
8 Philips Industry C B2C
9 Albert Heijn Wholesale and retail G B2C
10 Blokker Wholesale and retail G B2C
11 C-1000 Wholesale and retail G B2C
12 KLM Transport and storage H B2C
13 NS Transport and storage H B2C
14 PostNL Transport and storage H B2C
15 Bol.com Information and communication J B2C
16 TomTom Information and communication J B2C
17 ABN AMRO Financial institutions K B2C
18 Aegon Financial institutions K B2C
4-2 Establishment of Categories 47
Description of the Industries
The sample consists of ?rms active in seven di?erent industries. As indicated, the industry classi?cation is
based on CBS’ (2012) Standard Industry Classi?cation. A description of the industries that are part of the
sample is presented in this section.
1. Mining and Quarrying
The activities of ?rms in the mining and quarrying industry are concerned with the extraction of oil, gas
and/or minerals such as sand, gravel and clay.
2. Industry
Industry ?rms are producers of food, beverages, tobacco, textile, chemical products, pharmaceutical raw
materials, metal, electric products, machines, cars, other transport modes, furniture and other products.
3. Wholesale and Retail
The industry wholesale and retail consists of ?rms trading in cars, food, machinery, agricultural products,
textile, books, and other consumer products. In addition, ?rms in this industry operate shops in which
consumers can buy their products.
4. Transport and Storage
The activities of ?rms that are active in the transport and storage industry transport persons or products
across land, water, air or other transport modes. Next, ?rms in this industry store products. Also, mail
related activities belong to the transport and storage industry.
5. Information and Communication
Firms in the information and communication industry are publishers of books, papers, magazines, software
and computer games. In addition, the production and distribution of ?lms, music and television shows
are assigned to the information and communication industry. Also telecommunication activities, whether
through wires, wireless, satellite or other mediums are assigned to the information and communication
industry. All software related activities required for telecommunications are part of the information and
communication industry.
6. Financial Institutions
The industry ?nancial institutions consists of banks, ?nancial holdings, investment institutions, insurance
companies, pension companies, asset management companies and other ?nancial ?rms.
7. Consultancy, Research and Other Specialised Business Services
The activities of ?rms in the consultancy, research and other specialised business services relate to advisory
services on di?erent domains. Examples of ?rm types in this industry are law ?rms, accountancy ?rms,
engineering ?rms, architects, marketing ?rms, research ?rms, etcetera.
4-2-3 Description of the Measuring Period
uberVU, one of the emerging social media monitoring and analysis tools has granted access to their tool for a
period of 14 consecutive days, i.e. from Friday 20 July to Thursday 2 August. This tool is further described in
section 4-4-2. During the measuring period the eighteen ?rms have been monitored, resulting in the collection
of 224.687 social media posts related to di?erent ?rms. This amount of messages is deemed su?cient to analyse
what the subjects of social media messages related to ?rms are, and how these subjects di?er in volume from
each other, which is the purpose of this chapter. During 7 of the 14 days in the period, the 2012 Olympic Games
took place. As a consequence, the social media posts of ?rms that are for some reason – e.g. as a sponsor –
involved in the Olympic Games often have the Olympic Games as a subject of the post. It is common for ?rms
to sponsor events. In case the social media messages would have been collected during another period, it is likely
that some ?rms were sponsoring an event as well during the measurement period. Furthermore, the summer
holidays took place during the measurement period. Though people undertake other activities during their
holidays, and hence may show di?erent activities on social media as well, it is not likely that large corporations
– i.e. the ones in our sample – are not subject of discussion during this period. On the contrary, some ?rms will
precisely be mentioned during this period. However, when interpreting the results after the analyses, one should
be aware of the fact that the social media messages at which the conclusions are based have been created during
a holiday period. Furthermore, conclusions related to the procedure of social business intelligence will not be
a?ected by the fact that the data are created during a holiday period, since the way of collecting, processing
and analysing the data will be the same in any period.
48 Content Analysis
4-3 Pretest
In step 3 of the content analysis – see ?gure 4-1 –, the data collection is tested to ensure that the established
categories can be ?lled with data from the selected sample. To validate the categories that were established to
classify social media posts, we pretest the categories that are presented in table 4-2 by classifying the ?rst 100
messages of each respondent. Each of the pretest social media posts have been read and consequently assigned
to one of the categories. The sample illustrated that some of the categories were too generic. Therefore, we
added sub categories to some of main categories to gain more insight in the nature of the social media posts.
The categories of social media posts are discussed in the following section.
4-3-1 Categories of Social Media Posts
The purpose of this thesis is to assign social media posts to key-performance indicators. To achieve this, we
apply the classi?cation scheme of Ittner et al. (2003) to distinguish key-performance indicators from each other.
This classi?cation scheme distinguishes ten key-performance indicator categories. Basically, we adhere to these
ten categories. However, as we have experienced in the pretest, social media posts within one category are too
heterogeneous to simply assign the social media post to the high-level classi?cation that distinguishes between
ten categories. Therefore, an additional level of detail has been assigned to some of the main categories. This
additional level of detail has been established based on the pretest of the categories. Thus, the empirical data
has driven the establishment of these categories. The naming of these – more detailed – sub categories represent
the nature of the social media posts as well as possible. In the following section we discuss the categories of the
social media posts, which are based on the the KPI classi?cation of Ittner et al. (2003).
1. Short-term ?nancial results
The ?rst KPI category consists of indicators representing the (short-term) ?nancial performance of the ?rm.
Financial indicators are typically measured by ?rms using internal systems. In other words, there is no external
in?uence required to measure these metrics. Though the added value of the information of social media posts
related to ?nancial results may be of little value for the ?rm (management has ?nancial results earlier available
than the ?rm’s environment), it is wise to classify these social media posts nevertheless to provide an as
complete overview as possible of the type of social media posts that are available for ?rms. Typical examples of
?nancial indicators are the number of sales in a certain period, amount of debt on a certain moment, operating
expenditures in a certain period, etc. Social media posts that can be classi?ed into this category relate to the
?rm’s ?nancial performance. As we have experienced in the pretest, social media posts related to ?nancial
aspects of a ?rm are either related to discussions of the performance of a ?rm, or related to the ?rm’s share
prices. Accordingly, these two categories have been added as sub categories. These two sub categories are
discussed in the following sections.
1.1 Financial performance discussions
Social media are used to discuss the ?nancial performance of a ?rm. Often, posts related to ?nancial performance
contain a factual statement of the ?rm’s performance, which is sometimes followed by an opinion of the creator
of the post. In addition, these type of social media posts often contain a hyperlink to a website at which the
?nancial performance is further analysed. The following two example posts that consist in our sample illustrate
the type of posts that are classi?ed as ?nancial performance discussions posts:
“Akzo Nobel Q2 Pro?t Takes 21.5% Hit on Restructuring Charges http://t.co/b1518IDQ”.
“STEEL RESULTS: #ArcelorMittal Flat Carbon #Europe reports Q2 earnings fall http://t.co/0nPq7VjM
#steel”.
The existence of ?nancial performance related discussions in our sample data is due to the selection of the
?rms in our sample. As discussed in section 4-2-2, the sample ?rms are based on stock listed companies.
These companies are public limited liability ?rms, hence required to publicise their ?nancial performance on a
regular basis by law. Since the ?nancial performance is publicly available, it makes that these ?rms’ ?nancial
performance are subject of discussion on social media. In case that our sample would have consisted of limited
companies, which are not required to publicise their ?nancial position, we would probably not have found social
media posts related to ?nancial performance in the dataset.
4-3 Pretest 49
1.2 Shares related discussions
Another substantial part of the social media posts related to the ?nancial aspects of a ?rm has the ?rm’s shares
(prices) as subject. Often, these posts are made by analysts specialised in stock markets. The following two
social media posts illustrate discussions related to the shares of a ?rm:
“TomTom: Ster van de week: TomTom maakte vorige week bekend navigatieproducten en -diensten te gaan
leveren... http://t.co/8QRPxJmj #beleggen”.
“Stijgers 1: Wereldhave (4,05%) ; 2: TomTom (2,43%) ; 3: Bal last Nedam (2,25%) ; 4: VastNed
Retail (1,94%) ; 5: Heineken (1,76%)”.
The character of the second example post is typically found in the dataset, it shows the top ?ve funds of that day.
These type of posts are generated automatically by computers, often referred to as ‘bots’. For similar reasons
with the ?nancial performance discussions posts, we expect that shares related discussions are particularly found
in our dataset because the ?rms in our sample are stock listed.
2. Customer relations
The second KPI category de?ned by Ittner et al. (2003) consists of performance indicators related to customer
relations. As indicated in section 1-2, many researches acknowledge the opportunities for customer relationship
management using social media. Not surprisingly, social media posts in our test sample could be assigned under
the umbrella of customer relations. In our opinion, this category required a higher level of detail to distinguish
the social media posts made by the ?rm’s own web team from the posts that were directed towards the ?rm’s
web team. Seven categories have been added under the customer relations category to distinguish the nature
of the social media posts made by or directed to the web care teams. From a BI perspective it is valuable to
gain insight in the type of social media posts that are made by web care teams, since these posts may in?uence
other factors like customer satisfaction, sales, or the costs related to customer relations.
The social media posts related to customer relations typically show a conversation between a ?rm and a customer.
Social media posts related to customer relations represent a direction that is either from the customer to the
?rm, or from the ?rm to the customer. Social media posts from customer to customer containing a ?rm’s name
are categorised in a di?erent category, which will be discussed later. The identi?cation of social media posts
from customer to ?rm, or the other way around, is relatively easy because people begin their message with the
name of the receiver, preceded with an “@”. As such, social media posts starting with e.g. @ABN AMRO have
been classi?ed as a customer to ?rm post. Social media posts that were made by a webcare team could also be
easily recognised, because the creator of the message generally contains the name of the ?rm. To illustrate the
direction of the customer relations posts, the naming of the categories represent the direction of the post.
2.1 Customer questioning the Firm
Social media are used by customers to ask questions. The nature of the questions di?er, some people ask speci?c
questions about a product or service, how to use it or how it’s made while other questions are very broad and
relate to the company’s strategy or position in the market. Social media posts that were made by customers,
directed towards a ?rm and illustrating a question have been classi?ed under the category questioning customer.
Figure 4-2 illustrates the direction of these posts; from customer to ?rm.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-2: Questioning customer
Two example posts of the category questioning customer that have been found are illustratively shown below.
“@albertheijn Wat zijn de ramadanproducten? Ken je me die ? tweeten?
alvast bedankt.”“@KLM Hi, booked ?ights with you via @lastminute.com, wondering how we check in online? Saying that
option isn’t available?”
50 Content Analysis
Obviously, customer relationship teams that are active on the Web monitor social media posts in which customer
ask questions that are related to the ?rm, and consequently respond to these questions. Often, a questioning
customer post is followed by an explaining ?rm post.
2.2 Firm explaining the Customer
Clearly, one of the purposes of a web care team is to help customers with problems they experience. Many social
media posts that are made by web care teams contain an explanation of problems or questions that customers
posted on social media platforms. These posts have been classi?ed as explaining ?rm, illustrating that the social
media post has been written by a ?rm’s web care team to explain a certain customer something in response to
an earlier post made by the customer. Figure 4-3 illustrates the direction of the social media posts that have
been categorised as explaining ?rm.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-3: Explaining ?rm
Two example posts of the category explaining ?rm are shown below:
“@kiimberley94 Dag Kim. Als een bedrag dmv een automische incasso is afgeschreven, wordt het bedrag met
max 2 werkdagen teruggeboekt. Elvira.”
“@mepe176 Bij geldautomaten met een Maestro-logo is dit zeker mogelijk. Sommige winkels bieden
ook deze mogelijkheid. Suzanne.”
2.3 Customer complaining to the Firm
Customers employ social media as a means to complain. Plenty examples that have reached the newspapers
in recent years exist. It is therefore not surprising that our data set shows social media posts that represent
a complaint. Customers complain about product experiences, how they have been treated in their complaints
procedure, etc. Social media posts that have been made by customers and illustrating a complaint have been
classi?ed into the category complaining customer. Figure 4-4 illustrates the direction of these social media
posts; from customer to ?rm.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-4: Complaining customer
Two example posts illustrating complaining customer posts are shown below:
“@ABNAMRO Blijkbaar is de enige manier om met jul lie een probleem op te lossen om NIET TE
BETALEN. Want telefonisch sta je in de kou! HELP”
“@PostNL @PostNLWebcare de zoveelste keer dus dat de bezorgers de aangetekende stukken niet
laten tekenen ..........”
It is the purpose of a ?rm’s web care team to respond to the complaining posts. Therefore, a complaining
customer post is often followed by an understanding ?rm or complaining ?rm post.
2.4 Firm showing feeling of understanding to the Customer
Next, as the sample data shows, a ?rm’s web care team is also purposed to show a customer a feeling of
understanding of his or her experienced problem. The social media posts that were made by a ?rm’s web
4-3 Pretest 51
care team and represent a feeling of understanding with the customer’s complaint have been classi?ed into the
category understanding ?rm. Figure 4-5 schematically shows the direction of these posts. Understanding ?rm
posts di?er from explaining posts since understanding ?rm posts do not o?er the customer a solution to the
experienced problem, but rather show a feeling of understanding.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-5: Understanding ?rm
Two example posts that were found in the test sample and clearly illustrate a sense of understanding of the
customer’s experienced problems are illustrated below.
“@normanwil lems Dag Norman, vervelend te horen dat je reis niet doorgaat. Als je belt met 0900-0024
kunnen we je verder helpen. Margot.”
“@noni1967 Dag Nanette, dat is erg vervelend om te horen. Ik hoop dat het snel verwerkt wordt.
Margot.”
2.5 Customer thanking the Firm
As discussed, web care teams are amongst others purposed to help customers with problems that they experience,
thereby replacing the traditional telephone help desks. The social media posts in our sample clearly show a
conversation, where a ?rm replies to a social media post made by a customer. Once a customer has been assisted
by a company’s web care team, some customers take the e?ort to thank a ?rm for their assistance. Social media
posts that have been made by customers that are directed towards a ?rm and illustrating gratitude towards
the ?rm, have been classi?ed into the category named thanking customer. Figure 4-6 illustrates the direction
of thanking customer posts; from customer to ?rm.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-6: Thanking customer
Two example posts illustrating thanking customer posts are shown below:
“@ABNAMRO Ok, thanks voor de snelle reactie en ?jne dag!
.”“@albertheijn Oke bedankt! Dan ga ik Valkeniersplein proberen
.”2.6 Firm thanking the Customer
The sixth sub category that was added to the customer relations umbrella has been assigned the name thanking
?rm. As we experienced, web care teams often thank the customer for mentioning a de?ciency of a product /
service, or the web care teams thanks a customer for a compliment made on the side of the customer. Figure 4-7
illustrates the direction of these type of posts; from ?rm to customer.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-7: Thanking ?rm
Below, two example posts that were found in the sample and clearly present that the web care team is grateful
towards the customer’s earlier post are shown.
52 Content Analysis
“@Birdy_Fly Bedankt voor de tip Tom, ik zal deze door gaan zetten naar de betre?ende afdeling. Fijn
weekend. Martijn.”
“@LindaWestenberg Bedankt Linda! Ik wens je een ?jne dag toe. Elvira.”
2.7 Firm informing many Customers
Finally, ?rms also use social media to inform their customers on certain topics. Social media posts made by a
?rm and purposed to inform customers on a certain topic are classi?ed into the category named informing ?rm.
Whereas social media posts of the category explaining ?rm also inform customers, informing ?rm posts di?er
because they are directed to anyone. Figure 4-8 illustrates the “one-to-many” relation of the informing ?rm
posts. Thus, informing ?rm posts are not speci?cally directed towards an individual and hence – as opposed
to explaining ?rm posts – do not start with an “@”.
Customer Firm Explaining
Customer Firm Understanding
Customer Firm Thanking
Customers Firm Informing
Firm Customer Questioning
Firm Customer Complaining
Firm Customer Thanking
Figure 4-8: Informing ?rm
Two example social media posts found in the sample and classi?ed as informing ?rm are shown below:
“#NS Deventer-Zutphen (overwegstoring) Tussen Zutphen-Deventer geen treinen door overwegstoring.Extra
reistijd 30/60 min.Tot +/- 14:30... .”
“#NS Zwolle-Amersfoort: defecte trein: Tussen Amersfoort en Zwol le minder treinen.. Extra reistijd 15 -
30 min. ( Tot +/- 22:00 ).”
3. Employee relations
The third category of key-performance indicators comprises indicators related to employee relations. In this
thesis we seek for social media posts that relate to KPIs. Social media posts that are related to employee
relations will be classi?ed under this umbrella. However, as the pretest illustrated, there exists a variety in the
social media posts that could be assigned to the employee relations category. Therefore, two sub categories have
been established; recruitment and employee posts. These ?ndings are in line with McCorkindale (2010).
3.1 Recruitment
The ?rst sub category related to employee relations contains social media messages that involve employee
recruitment processes. Our sample shows messages where people write about vacancies in a ?rm, students
asking companies for an internship place, human resource managers wishing new employees a good start at
their ?rst day of work in the company, etcetera. Social media messages that are related to a company’s
recruitment process have been classi?ed into the recruitment category. Illustratively, two example posts of the
category recruitment are shown below.
“#nieuwe #vacature: Medewerker Verkoopklaar / Rotterdam / Albert Heijn #VCW #banen
http://t.co/bEDnvCFw.”
“Job opportunity: Deputy Program Manager - Trenchless Tech at ARCADIS - Washington D.C.
Metro Area #jobs http://t.co/RRSw5RtV.”
3.2 Employee posts
The second sub category related to employee relations contains social media messages made by the ?rm’s
employees. As our dataset illustrates, employees use social media to share their work experiences or indicate
that they are working at the ?rm. Social media messages that have been made by employees have been classi?ed
under the category employee posts. Illustratively, two posts existing in our data set that have been classi?ed as
employee posts are shown below.
4-3 Pretest 53
“O?cial ly 10 years working @ ABN Amro Bank... Uno?cially 13 years.. .”
“@PostNL Overuren niet betaald, geen bevestiging van gevraagde vakantie en ?etsdeclaratie wordt
niet uitbetaald. Lekker motiverend! #postnl.”
4. Operational performance
The fourth main category of key-performance indicators relates to operational performance. Social media
messages that relate to the operational performance of a ?rm are classi?ed into this category. As discussed,
category 2.3 contains social media messages in which users complain about the ?rm’s product / or service. It
is possible that customers complain about the ?rm’s operational performance, for instance about the delivery
time of a product. However, messages that have been classi?ed into the operational performance category
re?ect facts, while customer-to-?rm posts classi?ed as complaining customer (category 2.3) are more subjective
in nature and directed towards a ?rm. Below, two posts in our sample that have been classi?ed as operational
performance posts are shown.
“RT @ Webwereld: Derde keer in korte tijd storing ABN Amro http://t.co/v8Y953kC.”
“Vrijdag 27-7 kaart verstuurd uit Assen, 27-7 gestempeld in Zwol le. 31-7 al aangekomen in Tynaarlo.
Bravo @postnl. Niet gek voor 15 km.”
5. Product and service quality
The ?fth KPI category that ?rms apply contains indicators related to product and service quality. As our
sample illustrates, customers share their product and/or service experiences through social media. Social media
messages that represent product and service experiences of customers have been classi?ed in the category product
and service quality. Two posts existing in our sample and classi?ed as product and service quality posts are
shown below.
“Ik haat die Heineken met draaidop, altijd snij ik m’n hand er mee open.”
“Ik had een albert heijn tas toen ik thuis kwam waren myn handen helemaal blauw.”
6. Alliances
The sixth main category of key-performance indicators relates to the ?rm’s alliances. Social media messages
that are related to the ?rm’s partnerships / alliances have been classi?ed into the alliances category. Two posts
that have been classi?ed as alliances posts are shown below.
“In what has to be one of the strangest collaborations ever, military scientists from the UK’s Defence
Science and Technology Laboratory (DSTL) have been working with global paint and coating company
AkzoNobel to develop an anti-chemical weapon paint that can absorb harmful chemicals from enemy...
http://inhabitat.com/uk-military-develops-paint-that-absorbs-fal lout-from-chemical-attacks.”
“Op weg naar #Atrium MC om te spreken met nieuw bestuur #vereniging #artsassistenten en samenwerking
met #ABNAMRO.”
7. Supplier relations
The seventh main category of KPIs involves indicators related to a ?rm’s supplier relationships. Since this
thesis seeks to link social media data to KPIs, social media messages related to this category of KPIs have been
classi?ed into the supplier relations category. Apparently, suppliers of a company post social media messages
indicating that they supply the ?rm with products / services. Illustratively, two social media messages existing
in our sample which have been classi?ed as supplier relations posts are shown below.
“Onze koks ?etsen door het Hol land Heineken House, al les loop op rol letjes! #hhh2012
http://t.co/DrbbbXwl”
“Bezig met een nieuwe klus, het #vormgeven van een #advertentie deze keer voor de #albertheijn
#AH Valkeniersplein te #Breda.”
54 Content Analysis
8. Environmental performance
The next main category of key-performance indicators involves indicators re?ecting the ?rm’s environmental
performance. Social media messages relating to the environmental performance of ?rms have been classi?ed
into this category. The dataset shows social media messages in which consumers discuss the environmental
responsibility of the company. Below, two posts existing in our dataset that have been classi?ed as environmental
performance posts are shown.
“We kunnen heel veel bijdragen aan ontwikkelingen op het gebied van duurzaamheid... aldus Albert Heijn.
Ze verkopen uien uit Australië #AH.”
“@GreenpeaceNL - olijfolie v @albertheijn zit tegenw. in plastic “samen meer doen voor het milieu”
-> is t idd beter? http://t.co/2vLBda61.”
9. Product and service innovation
The ninth main category of key-performance indicators involves indicators related to product and service
innovation. As illustrated in section 3-2-3, social media are used by consumers to share product experiences
and to suggest innovations. The process of co-creation and prosuming is shown in our dataset as well. Social
media messages that re?ect people’s attitude to new products or services or contain suggestions for innovations
have been classi?ed as product and service innovations, of which two posts are illustratively shown below.
“Best Reviews - Philips Sonicare HX6732/02 HealthyWhite R732 Rechargeable Electric Toothbrush -
http://maxtodaystore.info/today-p... .”
“Moe worden van #ABN-AMRO bank geld overmaken steeds weer die achterlijke reader nodig , neem een
voorbeeld aan tan- codes van #ING!!”
10. Community
The tenth and ?nal main category of key-performance indicators related to indicators related to the ?rm’s
community. Social media posts belonging to the category community reveal how the community, that is,
external actors, perceive the ?rm. Many social media posts in the pretest could be assigned under the community
category. However, to provide more detail in the type of social media posts related to the community category,
we established ?ve sub categories to the community category. These sub categories are discussed in the following
sections.
10.1 Promotion
Some social media posts in our sample were created by ?rms themselves, and are hence not perceived as
user-generated content from a ?rm point of view. Social media messages that were made by ?rms themselves
and were purposed to promote the ?rm to the environment, were assigned to the sub category named promotion.
The following two posts illustrate promotion activities of a ?rm:
“What do you want to do more of in retirement? Travel, spend time with family, pursue hobbies, or more
education? Check out the results from the quick pol l here!”
“Albert Heijn - Kom op 18,19,25 en 26 aug. naar de Open Dagen van onze boeren en telers.
http://t.co/Saiv2aWY http://t.co/bmP1CePw.”
10.2 News
The test sample illustrated that many social media posts are (simply) noti?cations of news articles. We
positioned these posts under the category community since news messages determine the ?rm’s exposure to
the ?rm’s community. Posts belonging to the sub category news are written by professionals, which are mostly
journalists promoting their news article. Below, two examples of news categorised social media posts are shown.
“En verder in de serie vakantiebaantjes vandaag: Gerrit Zalm, waarmee verdiende de baas van #abnamro
zijn eerste centjes? #BNR.”
“@huizenprijzen: Han de Jong (Chief Economist ABN AMRO) : "Er is weinig in dit leven zo gevaarlijk als
schuld" : http://t.co/VOTreLdg via @youtube #schuld.”
4-3 Pretest 55
10.3 Public image
Thirdly, the sample test illustrated that individuals share their attitude towards a ?rm through means of social
media. We classify these posts into a sub category named public image. Social media posts classi?ed into the
public image category are not directed towards a ?rm, or, not purposed to get in contact with the ?rm. Rather,
social media messages assigned in the public image category represent discussions and “chatter” amongst the
social media users, in which the topic of discussion is the ?rm or its products / services. Public image posts are
written by non-professionals, while – as we will see in the next sub category – posts created by professionals
are assigned into a separate sub category. Below, two example posts that were found in the sample set and
represent the public’s image towards the ?rm are shown.
“Even kijken of blokker kruimeltje de ?lm heeft liggen want ik heb m al leen nog maar op video band!”
“Je moet staatsbanken ABNAMRO en ING 15% betalen als je rood staat op je betaalrekening en je
krijgt 2% als je + staat #Schurkenbanken.”
10.4 Professionals
Fourth, our pretest indicated that there are social media messages created by professionals. Messages that have
been created on social media by external professionals – not from the company – and talking about the ?rm
have been classi?ed into the category labelled professionals. Two example posts of this category are:
“Presentatie @ jaccooudhof van #KPMGmkb op de "Kengetal lenbijeenkomst" van @ Ful lFinance @
ABNAMRO en NOVAK”
“Sarah Harding interviews Arcadis at the A&WMA conference in San Antonio.”
10.5 Distributors
Finally, the data showed that social media are also used to promote products. However, the ?rm that produces
does not have to be necessarily the one that promotes the product. Our dataset contains social media posts
made by distributors of the product. These posts have been classi?ed into the category labelled distributors.
Two example posts of distributors are shown below.
“Macco Akzo Nobel Pai DWP24 Liquid Nails Drywal l Construction Adhesive: Special ly formulated latex
product for in... http://t.co/Vk3PdXLR”
“Best O?er - Philips Norelco AT830 PowerTouch Rechargeable Cordless Razor, Gray/Silver/Black -
http://maxtodaystore.info/weekly-...”
Unde?ned
Based on the 26 social media post categories (including main categories) that have been established in the
previous sections, we can not classify all social media posts. As discussed before, social media data is
unstructured and the interpretation of a social media post is not always easy. Therefore, messages that –
despite of the mentioning of the ?rm’s name in the post – could not be assigned to one of the categories have
been assigned into the unde?ned category. Two examples of posts that were unde?ned are:
“Volg ons (Unicum) @ ABN AMRO bij zuidplein om 13 : 00 !!!!! RT RT.”
“Nu naar ?etsenwinkel, blokker en c1000 met pap en mam!”
Spam
Unfortunately, the search queries that were used to scrape the social media content did still result in the
collection of data that is totally unrelated to the ?rm. This is due to the fact that people’s names or IDs are
similar to the ?rm’s name. Social media posts that were totally unrelated to the ?rm have been classi?ed as
spam, of which two examples are shown below:
“@Klm_babe Okay wel l maybe sometime next week then
”“@Jack_Heineken meen je dat nou? -_-
een korte broek aan naar de zaak ”56 Content Analysis
4-3-2 Revised Taxonomy of Categories
The categories of the social media posts are based on the KPI classi?cation scheme of Ittner et al. (2003), which
allows classi?cation of performance metrics into one of the ten categories. Our addition of sub categories does
not a?ect the structure of Ittner et al.’s (2003) classi?cation scheme, but rather adds a layer of detail to the
categories. An overview of the revised taxonomy – after addition of the sub categories – and a short description
of the social media posts of the corresponding categories is presented in table 4-4. Figure 4-9 schematically
shows the taxonomy of the key-performance categories and the social media post categories.
Table 4-4: Taxonomy of Categories of Social Media Posts
KPI Category Social media posts . . .
1. Short-term ?nancial results related to the ?rm’s ?nancial performance
1.1 Financial performance discussions related to the ?rm’s ?nancial performance
1.2 Stock related discussions made by professionals/individuals analysing the ?rm’s stock price
2. Customer relations from individuals purposed to contact the ?rm
2.1 Questioning customer posts from a customer asking a question to the ?rm
2.2 Explaining ?rm from the ?rm purposed to explain the customer something
2.3 Complaining customer from a customer complaining about the ?rm / ?rm’s products or
services
2.4 Understanding ?rm from the ?rm purposed to show the customer shared
understanding
2.5 Thanking customer from individuals purposed to thank the ?rm
2.6 Thanking ?rm from the ?rm purposed to thank the customer for an earlier post
2.7 Informing ?rm from the ?rm purposed to inform customers (not responding to
an individual)
3. Employee relations related to employees of the ?rm
3.1 Recruitment related to recruitment of new employees
3.2 Employee posts made the ?rm’s employees
4. Operational performance related to the ?rm’s productivity
5. Product and service quality related to the experience of products and services
6. Alliances related to joint-ventures or other cooperations
7. Supplier relations related to the suppliers of the ?rm
8. Environmental performance related to environmental compliance
9. Product and service
innovation
related to innovation
10. Community revealing the community’s perception of the ?rm (not
purposed to contact the ?rm)
10.1 Promotion made by the ?rm for promotion activities
10.2 News made by external professionals (journalism)
10.3 Public image made by non-professionals, individuals (‘chatter’)
10.4 Professionals made by professionals talking about the ?rm
10.5 Distributors made by distributors of the ?rm’s product/service
Unde?ned that could not be de?ned in one of the categories
Spam that are not related to the ?rm
4-4 Data Collection and Evaluation
The fourth step of the content analysis comprises the data collection and analysis. The purpose of this section is
to collect social media posts related to the ?rms of the sample, and to analyse these data to identify di?erences
in the content related to di?erent ?rms. Moreover, the experiences that we encounter in the data collection
4-4 Data Collection and Evaluation 57
1. Short-term financial results
4. Operational performance
5. Product and service quality
6. Aliances
7. Supplier relations
8. Environmental performance
9. Product and service innovation
10. Community
10.2 News
10.1 Promotion
1.1 Financial performance discussions
10.3 Public image
2. Customer relations
Customer to Firm
Firm to Customer
2.3 Complaining customer
2.1 Questioning customer
2.5 Thanking customer
2.6 Thanking firm
2.4 Understanding firm
2.7 Informing firm
Main KPI Categories Social Media Post Categories
3. Employee relations
3.1 Recruitment
3.2 Employee posts
10.4 Professionals
10.5 Distributors
2.2 Explaining firm
1.2 Stock related discussions
Figure 4-9: Taxonomy of Social Media Post Categories
The ?gure illustrates the ten main categories of key-performance indicators that have been found in the literature.
Additionally, sub categories have been established at which the social media messages could be assigned. These
additional categories have been constructed based on the empirical data.
phase as well in the data analysis phase will serve as a baseline in formulating requirements for a social business
intelligence procedure that we develop in a later stage of this thesis.
Watson and Wixom (2007) illustrate that data collection requires about 80% of the “time and e?ort” related to
business intelligence, and that data collection is responsible for “50% of the unexpected costs” in BI projects.
Social media platforms are new sources for ?rms to collect data. The experiences from the collection of social
media data for this research contain valuable lessons learned for ?rms willing to utilise social media data for
58 Content Analysis
business intelligence. Therefore, we pay much attention to describing the steps that are necessary to collect
social media data.
In section 4-4-1 we discuss the search queries that are used to ?lter out those social media posts that relate to
the ?rms in our sample. Section 4-4-2 describes how the social media posts have been extracted from the web,
and how these posts have been placed in a database allowing to be analysed. Although we will use proper search
terms, it is expected that many social media posts contain unrelated information. Therefore, the data will be
cleaned in section 4-4-3. Once the data is cleaned, section 4-5 provides descriptive statistics about the amount
of social media posts available for ?rms in di?erent industries and for di?erent positions regarding end-users.
4-4-1 Search Terms
As on the regular web, search terms are used on social media to ?lter out information that is the subject of
interest. In social media, and especially on Twitter, users place a # (‘hashtag’) before a word to indicate the
subject of the particular social media post. Hashtags can be considered as meta data tags indicating the subject
of the social media post. When each user uses the same hashtags about a certain topic, it becomes easy to track
the stream of social media posts related to that subject. We will use the strength of hashtags to ?lter out the
social media posts that are related to the ?rms in our sample. Another widely used symbol to indicate that a
social media post is direct to a person, or a ?rm, is the @ (‘at symbol’). As with the hashtag, this symbol is
positioned before one’s (nick)name to illustrate that a post is direct towards this person or organisation. We
will use the at symbol in our search terms, because social media posts containing this symbol are directed to a
receiver, the ?rm.
Additionally, because not everyone and not each social media platform adheres accurately to the usage of
hashtags, we add a search term containing the name of the ?rm without a hashtag to the search terms. Next,
because some ?rms have name that can be written in multiple forms, we also search on di?erent names. An
overview of the search terms used to ?lter out the social media posts that are related to the ?rms in our sample
is presented in table 4-5.
Table 4-5: Firms and Corresponding Search Terms
Firm Search Terms
1 ABN AMRO #abnamro, #abn amro, @abnamro, @abn amro, abnamro, abn amro
2 Aegon #aegon, @aegon, aegon
3 Akzo Nobel #akzonobel, #akzo nobel, @akzonobel, @akzo nobel, akzo nobel, akzonobel
4 Albert Heijn #albertheijn, #albert heijn, @albertheijn, @albert heijn, albertheijn, albert heijn
5 Arcadis #arcadis, @arcadis, arcadis
6 ArcelorMittal #arcelor mittal, #arcelormittal, @arcelor mittal, @arcelormittal, arcelor mittal,
arcelormittal
7 Blokker #blokker, @blokker, blokker
8 Bol.com #bol.com, @bol.com, bol.com
9 C-1000 #c1000, @c1000, c1000
10 Coca-Cola #coca-cola, #cocacola, @coca-cola, @cocacola, coca-cola, cocacola
11 Fugro #fugro, @fugro, fugro
12 Heineken #heineken, @heineken, heineken
13 KLM #klm, @klm, klm
14 NS #ns, @ns, ns
15 Philips #philips, @philips, philips
16 PostNL #postNL, @postNL, postNL
17 TomTom #tomtom, @tomtom, tomtom
18 Unibail-Rodamco #unibail-rodamco, @unibail-rodamco, unibail-rodamco
4-4-2 Scraping Social Media Content
Web scraping – also known as web crawling – is the excavation of data from web pages into a local structured
database, so that these data can be analysed (Huang, Li, Li, & Yan, 2012). Figure 4-10 visualises this process.
The web scraper is provided with keywords, so that it can detect those particular web pages or social media
posts related to the topic of interest. All content that ful?ls the keywords are consequently stored into whatever
4-4 Data Collection and Evaluation 59
form the person prefers, which is often a database, a web page, another application or – as in this thesis – a
spreadsheet. Scraping web pages allows one to extract that particular information from the web that one is the
topic of interest, and consequently process the data for its own purpose.
Web
Spreadsheet
Browser
Database
Application
Scraper
Figure 4-10: Web Scraper
Content from the web (e.g. a social network site) is ?ltered, extracted and stored to into di?erent types of media,
e.g. in a database, or a spreadsheet.
Software tools purposed to monitor social media are emerging rapidly. uberVU is one of such tools that
are available on the market. In addition to social media monitoring uberVU allows the extraction of social
media posts into comma-separated value (“CSV”) format, and is therefore regarded as a social media platform
scraper. It is this aspect of the tool that was decisive for the selection of a social media extraction tool that we
used to scrape the content. uberVU was established in 2008, their software is amongst others used by NBC,
Microsoft, Audi, Nestle, T-Mobile, Thomas Cook, 3M, PayPal, BASF and The World Bank. The software
indexes multiple social network platforms, including Facebook, Twitter, YouTube, Flickr, Vimeo, Picasa. In
addition, traditional media like news sites and (Wordpress) blogs are monitored. Consequently, the software
presents metrics including number of mentions over the last period, number of likes, number of shares, platform
distribution, sentiment of the online posts, gender distribution, language of the posts and the countries where
the posts originated.
The eighteen selected ?rms have been monitored for a period of 14 days using search queries based on keywords
containing the name of the ?rms. Please see table 4-5 for the list of keywords used to ?lter out content from the
social media platforms. As such, all posts that were publicly available have been scraped from the social media
websites. uberVU allowed the exportation of maximum 10.000 posts in CSV format per request. Therefore,
the ?rms were subjected to a request on a daily basis. As a consequence, the search request on day t contained
content that existed yet in the search request of day t ? 1. Figure 4-11 illustrates the overlap in the scraped
content. Before the individual daily search requests were consolidated, the records that existed were removed.
The CSV exportation has been executed on a daily basis, and each addition to the database (except for the
?rst) resulted in the notion that there existed yet certain posts in the database. Therefore, we can conclude
that the social media messages in our database provide a complete overview of the messages created in the
measuring period and related to the sample ?rms.
Day Scraped content
1
2
3
4
5
…
Overlap
Overlap
Overlap
Figure 4-11: Overlap in Scraped Content
The ?gure illustrates that the daily search runs for new social media messages resulted in content that did yet exist
in the database. Messages that existed in the database were removed.
When scraping the social media posts in the database, the following attributes of the posts were recorded: date,
platform, username, content, language, sentiment, gender, followers, pro?le, country, region, city and URL to
the post. The URL is the only attribute that is unique for a post, and is used to determine whether or not a
post existed yet in the database, before it was recorded. It is common that social media messages are copied
from one platform to the other. With our scraping method, identical social media messages created on two
60 Content Analysis
platforms are regarded as two individual posts, and therefore consist two times in our database. Table 4-6
shows a cross-cut of one of the monitored ?rms, showing one scraped post and the corresponding attributes.
As illustrated, the scraper is not able to determine all attributes of a posts. For the example post presented in
table 4-6, the tool was unable to determine the sentiment while is it very easy for a Dutch speaking person to
determine that the content is obviously negative. Probably, this is due to the fact that the content is written
in Dutch, while meta data required to determine the sentiment of the post about this language is not (yet)
available. Also the country and the region are unknown, which is due to the fact that the user that has written
the posts did not agree to share his or her location with to the social media platform.
Table 4-6: Example of Scraped Data
Attribute Example
Date 23-7-2012
Platform twitter
Username Anne
Content @ABNAMRO De internet site doet het nog steeds niet... Ik kan
dus geen geld overmaken nu. Dat is mijn probleem nu.
Language dutch
Sentiment unknown
Gender f
Followers S
Country unknown
Region unknown
Pro?le http://twitter.com/AnneXD_
URL http://twitter.com/AnneXD_/statuses/227776320254382080
4-4-3 Data Cleaning
Before the social media data is ready for analysis, it requires cleaning. Though we used proper search terms, not
all these posts actually relate to the ?rms. The search terms used to monitor the selected companies resulted
in posts that did not have any relation with the selected companies. For instance, Twitter user names like
@klm_klm_klm, @KLM_350, @KLM_2013, @KLM_babe and @klm_luvsya existed in the dataset belonging
to KLM, though these Twitter accounts do not have any relation with KLM (the company). This so-called
‘noise’ has been classi?ed as spam. The existence of noise in datasets is especially applicable on social media
data. Therefore, any organisation that using social media data should ?lter out the valuable content from the
noise.
4-5 Descriptive Statistical Analysis
This section describes the statistics that are acquired by the collection of the social media messages. More
speci?cally, three topics are discussed in the following section. First, the distribution of the sources of the social
media messages are presented (section 4-5-1). This distribution will reveal – taken into account the publicly
accessible social media posts – which social media platforms are mostly used by customers to discuss ?rms
and ?rms’ products and services, hence providing ?rms insights in ‘where to look’ for ?rm-related social media
content. Second, the volume of ?rm-related social media posts are examined (section 4-5-2). The volume of
these messages is analysed per industry. As such, we gain insight in the amount of user-generated content that
is created in di?erent industries. The average daily mentions are also analysed per customer relation type.
The ?nal topic that is discussed in this section describes the statistics of the classi?ed social media posts into
categories (section 4-5-3). These categories are related to di?erent sort of KPIs. Therefore this analysis will
reveal which sort of KPIs are likely to be in?uenced by social media activities.
4-5-1 Channel Distribution
The social media messages have been collected from a variety of social media platforms. Figure 4-12 presents
the distribution of all collected posts along the social media channels. As can be concluded, the largest share
4-5 Descriptive Statistical Analysis 61
(83%) of the collected posts have been created on Twitter. These ?ndings are in line with A. N. Smith et al.
(2012). While Facebook and other social media channels are responsible for a much smaller portion of the
posts according to this dataset, one should place a note to these data. Facebook pro?les may namely be set
unattainable for non-friends. Therefore, the scraper – like any other web scraper – was unable to extract data
from private pro?les. Although one can argue that this distribution may provide an incomplete view of the
situation, it is representative for a real situation in which a ?rm would collect social media posts from the
popular sites because it holds also for a ?rm that it cannot access private social media pro?les of people.
8% 3%
5%
n=224.687
83%
Facebook Twitter Blogs News Other Platforms
Figure 4-12: Social Media Channel Distribution
Illustrates the sources of the social media messages in our sample.
The channel distribution di?ers per ?rm, as shown in table 4-7. For each ?rm, the table shows from which
channels the messages have been collected. As can be concluded, it holds for all ?rms that the majority of the
publicly accessible messages are created on Twitter. The table shows one remarkable value, the collected social
media posts of ABN AMRO are for 55% created from Picasa, a photo sharing platform. This ?gure can be
perceived as a one-time event, because ABN AMRO has uploaded pictures from marketing events (KLM Open,
World Tennis Tournament) that the ?rm organises to their Picasa pro?le. The web scraper recognised each
individual picture as a separate social media post. Appendix C shows the channel distribution on a ?rm to ?rm
basis graphically.
Table 4-7: Social Media Channel Distribution
Shows the absolute number of messages that have been collected from the various platforms. Furthermore, it shows
the percentages of the platforms from which the messages have been collected.
Platform Facebook Twitter Blogs News Other Total
Abs % Abs % Abs % Abs % Abs % Abs
ABN AMRO 124 2% 3.000 42% 70 1% 15 0% 3.858 55% 7.067
Aegon 110 8% 1.173 81% 79 5% 20 1% 67 5% 1.449
Akzo Nobel 30 3% 806 87% 43 5% 25 3% 18 2% 922
Albert Heijn 328 3% 11.116 96% 77 1% 1 0% 59 1% 11.581
Arcadis 8 2% 422 93% 9 2% 10 2% 6 1% 455
ArcelorMittal 439 8% 4.569 83% 296 5% 89 2% 139 3% 5.532
Blokker 155 6% 2.526 91% 71 3% 3 0% 14 1% 2.769
Bol.com 472 8% 5.124 89% 115 2% - 0% 71 1% 5.782
C-1000 362 3% 10.583 96% 81 1% 5 0% 33 0% 11.064
Coca-Cola 1.653 5% 29.347 89% 999 3% 69 0% 885 3% 32.953
Fugro 6 1% 385 90% 20 5% 15 4% 2 0% 428
Heineken 5.726 15% 32.332 82% 494 1% 122 0% 751 2% 39.425
KLM 2.316 9% 22.601 86% 617 2% 90 0% 740 3% 26.364
NS 703 12% 4.970 85% 103 2% - 0% 87 1% 5.863
Philips 4.641 12% 26.260 68% 3.404 9% 138 0% 4.007 10% 38.450
PostNL 77 6% 1.207 91% 27 2% - 0% 12 1% 1.323
TomTom 1.308 4% 29.787 91% 630 2% 67 0% 956 3% 32.748
Unibail-Rodamco 3 1% 487 95% 9 2% 12 2% 1 0% 512
Total 18.461 8% 186.695 83% 7.144 3% 681 0% 11.706 5% 224.687
62 Content Analysis
4-5-2 Volume of Firm-Related Social Media Messages
As described in chapter 1, this thesis analyses the availability of user-generated social media content on two
dimensions. These dimensions are customer relation type and industry type. The availability of user-generated
content has empirically been measured during a period of time. In this thesis, the variable called average daily
mentions serves as a measure to describe the amount – or volume – of generated ?rm-related user-generated
social media content.
The eighteen ?rms have been monitored for a period of two weeks. Each time that a social media post that
contained one of the ?rms’ names and that was publicly accessible has been downloaded. Some ?rms were
mentioned more extensive than others. As such, it is possible to gain insight in the number of social media
messages that are daily generated on the web per ?rm. Table 4-8 shows for each ?rm how many posts have
been collected. The ?nal column shows how many times – on average – the ?rm has been mentioned on a daily
basis. The average daily mentions for each ?rm i have been calculated based on formula 4-1.
Average_Daily_Mentions
i
=
Total_Collected_Posts
i
Measured_Days
i
=
Total_Collected_Posts
i
MAX_Date
i
? MIN_Date
i
(4-1)
Table 4-8: Average Daily Mentions per Firm
The second column of the table indicates the total amount of messages that have been collected in relation with
the ?rm. In the third column, this number is divided by the number of days at which messages have been found,
hence representing the average daily mentions of the ?rms.
Firm Collected posts Average daily mentions
1 ABN AMRO 7.067 544
2 Aegon 1.449 111
3 Akzo Nobel 922 77
4 Albert Heijn 11.581 965
5 Arcadis 455 38
6 ArcelorMittal 5.532 461
7 Blokker 2.769 231
8 Bol.com 5.782 482
9 C-1000 11.064 922
10 Coca-Cola 32.953 2.996
11 Fugro 428 36
12 Heineken 39.425 3.285
13 KLM 26.364 2.197
14 NS 5.863 489
15 Philips 38.450 2.958
16 PostNL 1.323 102
17 TomTom 32.748 2.519
18 Unibail-Rodamco 512 39
? 224.687
Volume per Firm
As can be concluded from the ?nal column in table 4-8, the average daily mentions di?ers from ?rm to ?rm.
From this table, we can conclude that the available user-generated content di?ers from ?rm to ?rm, and that
the applicability of social media data for business intelligence purposes will not be possible for all ?rms, since
not for each ?rm UGC is generated. Figure 4-13 illustrates the average daily mentions of di?erent ?rms in
our sample. The ?gure as been ordered from highly mentioned ?rms to less mentioned ?rms. In the following
paragraph, the volume of social media messages is investigated from a customer relation type perspective.
Volume per Customer Relation Type
Our sample consists of a mix of ?rms that pursue a B2C or a B2B relation. One of the objectives of this thesis
is to investigate whether and to what extent B2C ?rms are more often subject of discussion on social media
than B2B ?rms. With the collected data we can analyse this topic.
4-5 Descriptive Statistical Analysis 63
Heineken
Coca-Cola
Philips
3.000
3.500
TomTom
KLM
2.000
2.500
t
i
o
n
s
[
x
1
/
d
a
y
]
Albert Heijn
1 000
1.500
2.000
A
v
e
r
a
g
e
D
a
i
l
y
M
e
n
t
i
o
n
s
[
Albert Heijn
C-1000
ABN AMRO
NS Bol.com
Arcelor
Mittal
Blokker
Aegon P tNL Unibail
500
1.000
Blokker
Aegon PostNL
AkzoNobel
Unibail
Rodamco Arcadis Fugro
0
Firm
Figure 4-13: Average Daily Mentions of Firms
Bar chart illustrating the variation in the volume of ?rm-related social media posts. Firms have been ordered
descending.
Heineken
Coca-Cola
Philips
3.000
3.500
TomTom
KLM
2.500
x
1
/d
a
y
]
1.500
2.000
A
v
e
r
a
g
e
D
a
ily
M
e
n
t
io
n
s
[
x
1
/d
a
y
Bol.com NS
Albert Heijn
C-1000
ABN AMRO Arcelor
Mittal
500
1.000
A
v
e
r
Bol.com NS
PostNL
Blokker
Aegon Unibail
Rodamco
AkzoNobel
Arcelor
Mittal
Arcadis Fugro
0
500
B2B B2C
Figure 4-14: Average Daily Mentions of Firms
An overview of the average daily produced ?rm-related social media messages. Firms have been clustered based on
their customer relation type.
Figure 4-14 shows the daily volume of ?rm-related social media content, in which the ?rms are clustered on
their relation type and consequently ordered descending. This ?gure suggests that B2C ?rms – coloured in red
– are more likely to ?nd social media content that is related to their ?rm than ?rms performing B2B relations
(coloured in blue). Figure 4-14 shows one remarkable value; the average daily mentions of ArcelorMittal. When
analysing the content of the messages related to this ?rm, the explanation is discovered. ArcelorMittal has
constructed the belvedere for the Olympic Games, called the ArcelorMittal Orbit. During the measurement
period, the tower has been opened for the public, leading to discussions on social media.
In table 4-9, the average daily mentions of B2C ?rms have been consolidated, as are the B2B ?rms. Thus,
the ?nal column of table 4-9 presents an average of an average. Hence, the values are normalised and thereby
eliminating the fact that the number of respondents di?ers between the two groups. The ?rst hypothesis
formulated at the beginning of this chapter was:
H
1
: The volume of ?rm-related social media messages is higher for B2C ?rms than for B2B ?rms.
64 Content Analysis
Table 4-9: Average Daily Mentions
The table consolidates the messages of all ?rms operating the same customer relation type, i.e. B2C or B2B. The
?nal column illustrates the average daily mentions of an individual ?rm operating either a B2C or B2B relation.
Customer relation Collected posts Average daily mentions per ?rm
B2C 216.838 1.369
B2B 7.849 130
? 224.687
With the ?gures presented in table 4-9, a bar chart is created in order to draw conclusions with respect to the
?rst hypothesis. Figure 4-15 depicts the average daily volume of ?rm-related social media messages, consolidated
per customer relation type.
? = 1.369
? = 1.223
N = 13
600
800
1.000
1.200
1.400
1.600
g
e
D
a
i
l
y
M
e
n
t
i
o
n
s
[
x
1
/
d
a
y
]
? = 130
? = 186
N = 5
0
200
400
600
B2C B2B
A
v
e
r
a
g
Figure 4-15: Average Daily Mentions of Firms
Bar chart illustrating the average daily volume of ?rm-related social media messages. Firms have been consolidated
per customer relation type.
Figure 4-15 strongly suggests that B2C ?rms are far more often subject of discussion on social media sites than
B2B ?rms. Thus, the results of our content analysis strongly suggest that the ?rst hypothesis is to be accepted,
implying that the volume of ?rm-related social media messages di?ers for performing B2B or B2C relations,
with B2C ?rms being highly more mentioned on social media than B2B ?rms.
Volume per Industry Type
The second dimension on which the volume of ?rm-related social media content is investigated relates to
industries. Our sample consists of eighteen ?rms active in seven di?erent industries, see table 4-3 for an
overview. As a ?rst step to identify possible di?erences in the volume of daily messages between industries,
the ?rms have been clustered on industry type in ?gure 4-16, and have consequently been sorted in descending
order.
Figure 4-16 suggests that there exists a di?erence in the amount of user-generated content between di?erent
industries. Therefore, the di?erent volumes are consolidated per industry, and analysed in the following
paragraphs. Furthermore, ?gure 4-16 reveals that while an industry average may be lower than the average
of an other industry, an individual ?rm may still be mentioned higher than a ?rm of an other industry. For
example, ABN AMRO is mentioned more often than PostNL, while the industry ?nancial institutions is on
average less mentioned than the transport & storage industry. These insights suggest that there are company
speci?c aspects that also in?uence the amount of messages that are created in relation to the ?rms, i.e. the
industry type is not the only aspect in?uencing the volume of ?rm-related messages.
We examine the availability of user-generated content in the di?erent industries by comparing the average daily
mentions of the di?erent industries with each other. Table 4-10 presents the number of mentions of ?rms,
4-5 Descriptive Statistical Analysis 65
Heineken
Coca-Cola
Philips
3.000
3.500
TomTom
KLM
2.500
x
1
/d
a
y
]
1.500
2.000
A
v
e
r
a
g
e
D
a
ily
M
e
n
t
io
n
s
[
x
1
/d
a
y
Albert Heijn
C-1000
ABN AMRO
NS Bol.com
Arcelor
Mittal
500
1.000
A
v
e
r
NS Bol.com
Arcelor
Mittal
Blokker
Aegon PostNL
AkzoNobel
Unibail
Rodamco Arcadis Fugro
0
500
Information&
Communication
Industry Transport &Storage Wholesale &Retail Financial Institutions Mining&Quarrying
Consultancy, Research
&Other Specialised
Business Services Communication Business Services
Figure 4-16: Average Daily Mentions of Firms
Bar chart illustrating the average daily volume of ?rm-related messages. Firm have been clustered over the industries
and consequently ordered descending.
consolidated across di?erent industries. The second column of the table shows the total amount of social media
posts that have been collected in the corresponding industry. The third column shows the average daily mentions
of a ?rm in the corresponding industry. The values in the third column thus represent an average of an average,
thereby eliminating the fact that the number of ?rms – respondents – di?ers per industry type. Figure 4-17
presents these values. This ?gure suggests that the existence of ?rm-related UGC di?ers among industry type,
implying that di?ers per industry whether or not there exists user-generated content on social media.
Table 4-10: Average Daily Mentions, Consolidated per Industry
Industry Total collected posts Average daily mentions per ?rm
Mining and quarrying 44.904 269
Industry 72.378 3.080
Wholesale and retail 25.414 706
Transport and storage 33.550 929
Information and communication 38.530 1.500
Financial institutions 9.028 231
Consultancy, research and other
specialised business services
883 37
? 224.687
Figure 4-16, 4-17 and table 4-10 provided insight in the variations in the volume of ?rm-related social media
messages across di?erent industries. With this insight, we can examine the second hypothesis of this chapter,
which was formulated as:
H
2
: The volume of ?rm-related social media messages di?ers between industries.
The results of our analysis illustrate variations in the volume of ?rm-related social media messages, which
suggest – based on our sample – that there exists a variation in the daily volume of social media messages that
are created. When ordered descending, the industry ?rms are mentioned mostly, followed by information and
communication, transport and storage, wholesale and retail, mining and quarrying, ?nancial institutions and
consultancy, research and other specialised business services being the least mentioned on social media. Thus,
the results of this analysis indicate that it matters in which industry a ?rm is active whether or not the ?rm
will be subject on social media. It is therefore that – taken into account our sample – we accept the second
hypothesis, implying that the volume of ?rm-related social media messages di?ers between industries. However,
the categories on the two dimensions that are researched in this thesis are not fully independent. For example,
66 Content Analysis
? = 3.080
? = 179
N = 3
3.000
3.500
2.500
3.000
y
M
e
n
t
io
n
s
[
x
1
/d
a
y
]
? = 1.500
? = 1.441
N = 3
1.500
2.000
A
v
e
r
a
g
e
D
a
i
l
y
M
e
n
t
? = 929
? = 1.115
N = 3
? = 706
? = 412
N = 3
500
1.000
1.500
? = 269
? = 272
N = 2
? = 231
? = 273
N = 3
N = 37
? = 2
N = 2
0
500
Industry Information and Transport and Wholesale and Mining and Financial Consultancy,
Industry Information and
Communication
Transport and
Storage
Wholesale and
Retail
Mining and
Quarrying
Financial
Institutions
Consultancy,
Research and
Other Specialised
Business Services
Figure 4-17: Average Daily Mentions of Firms
Bar chart illustrating the volume of daily mentions for ?rms in di?erent industries. The social media messages
related to ?rms in the same industry have been consolidated.
the category consultancy, research and other specialised business services solely consists of B2B ?rms. This
issue is discussed later in this thesis.
4-5-3 Subjects of Social Media Posts
Next to an assessment of the amount of social media posts that are created on the web, this thesis examines
the subjects of the social media posts in order to link the messages to key-performance indicators. The social
media messages of the sample ?rms have been classi?ed into one of the categories that have been established
in section 4-3-1. These categories are based on ten categories of commonly applied key-performance indicators.
Consequently, the collected social media posts of the ?rms in the sample have been classi?ed into one of these
categories. The results of this activity are documented in appendix B from ?rm to ?rm. In this section, the
subjects of social media posts are analysed. First, the subjects of social media messages are discussed from ?rm
to ?rm. Next, the social media posts are analysed from the two dimensions which are the perspectives of this
thesis. Consequently, we analyse whether or not the customer relation type in?uences the types of subjects that
are contained in social media messages. Finally the same analysis is executed, only this time from an industry
perspective.
Subjects per Firm
Figure 4-18 shows the results of the classi?cation process, in which all ?rms are displayed. The percentages
in the ?gure indicate how many of the classi?ed posts are assigned to the corresponding KPI category. The
colours of the bars represent the main categories of subjects of social media posts. When purely looking at
the colours, it becomes clear that some ?rms’ social media posts contain much ?nancial result (‘orange’) posts,
while others contain a high portion of customer relations (‘red’) posts. Furthermore, we see the existence of
community (‘blue’) posts in each ?rm. For reasons of readability, under-represented categories of subjects have
been grouped under a category called other (‘grey’). Figure 4-18 shows two remarkable values of the ‘other’
category. For KLM, these messages mainly have been classi?ed as being spam. The three letters are used by
other people on social media as well, for instance because these are the initials of the person. Also C-1000 shows
a remarkably high percentage of ‘other’ posts. A closer look at C-1000’s messages reveals that many of these
posts can not be classi?ed into one of the ten categories and are hence classi?ed as ‘unde?ned’ posts. Mainly,
these posts contain expressions of people who use the C-1000 stores as a point of reference to meet each other.
A full overview is presented in appendix B.
Figure 4-18 illustrates that the subjects of social media posts related to ?rms di?ers from ?rm to ?rm. In
4-5 Descriptive Statistical Analysis 67
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
P
e
r
c
e
n
t
a
g
e
o
f
S
o
c
i
a
l
M
e
d
i
a
P
o
s
t
s
0%
10%
Short-termfinancial results Customer relations Community Other
Figure 4-18: Social Media Posts Subject Classi?cation
Stacked bar chart illustrating the percentages of social media posts assigned to di?erent categories based on the
posts’ subjects. Small percentages of categories have been merged under ‘other’.
the following paragraphs, we investigate whether or not it is likely that the factors customer relation type and
industry type a?ect the type of subjects of ?rm-related social media posts.
Subjects per Customer Relation Type
In this paragraph we assess whether the subjects of social media messages di?er for di?erent customer relation
types. Illustratively, we examine amongst other whether or not the percentage of product and service quality
related messages di?ers for ?rms performing a B2B or a B2C relation. As a ?rst step to analyse di?erences
in subjects across B2B and B2C ?rms, the individual ?rms have been grouped into their respective customer
relation type, and the percentages of the subjects have been plotted in ?gure 4-19. Again, for readability issues,
under-represented categories of social media subjects have been grouped under an other category.
The third hypothesis that is examined in this chapter was formulated as:
H
3
: The subjects of ?rm-related social media messages di?er between ?rms performing B2B and
B2C relations.
As can be concluded from ?gure 4-19, social media posts related to B2B ?rms contain a high percentage of
posts related to ?nancial results (‘orange’), while this percentage is under-represented for B2C ?rms. Such
information is not of any additional value for a ?rm, since these posts contain information that is yet available
at the ?rm. On the contrary, social media messages related to B2C ?rms contain a high portion of customer
relations (‘red’) related posts in comparison with B2B ?rms. For both type of ?rms it holds that a high portion
of the social media messages reveal the communities’ perceptions of the ?rm (‘blue’ bars). However, B2B ?rms’
community related social media posts are created by professionals, while in the community messages related
to B2C ?rms, these messages are created by consumers. For a detailed overview of the percentages of subjects
related to each ?rm, please see table B-1 in appendix B. These insights suggest that the subjects of social media
messages related to ?rms that pursue di?erent customer relation types vary, and that the third hypothesis is
to be accepted. Thus, ?rms performing di?erent customer relation types will ?nd di?erent subjects in their
?rm-related social media messages.
68 Content Analysis
40%
50%
60%
70%
80%
90%
100%
P
e
r
c
e
n
t
a
g
e
o
f
S
o
c
i
a
l
M
e
d
i
a
P
o
s
t
s
0%
10%
20%
30%
0%
B2B B2C
P
e
r
c
e
n
Short-termfinancial results Customer relations Community Other
Figure 4-19: Social Media Posts Subject Classi?cation, Consolidated per Customer Relation Type
Stacked bar chart illustrating the portion of social media messages related to di?erent subjects. Firms have been
consolidated per customer relation type. Categories that are under-represented are merged under ‘other’ posts.
Subjects per Industry Type
In the previous paragraph the availability of user-generated social media content related to di?erent social
media post categories has been investigated across the customer type dimension. This section performs the
same analysis, only this time the industry dimension serves as the distinguishing factor of the ?rm types. In the
appendix, ?gure D-1 (page 113) lists – in detail – for each industry the average amount of social media posts
related to the di?erent categories of social media posts.
The fourth hypothesis that is examined in this chapter was formulated as:
H
4
: The subjects of ?rm-related social media messages di?er between industries.
Figure 4-20 depicts the portions of social media messages of the di?erent subjects across the seven industries.
Under-represented subjects of social media messages have been merged in the other category. When looking at
the colours of the bars, di?erences are seen – again – in customer relations (‘red’) type of posts and ?nancial
results (‘orange’) posts. E.g. the posts related to wholesale and retail ?rms contain a higher portion of customer
relation posts than ?nancial results posts. The contrary is seen in mining and quarrying, and consultancy ?rms.
Thus, our results suggest that the subjects of social media messages di?er per industry type, implying that the
fourth hypothesis is to be accepted.
4-6 Interpretation of the Results
In the ?fth step of the content analysis, the results are interpreted into meaningful conclusions. In the beginning
of this chapter, the hypotheses that are to be examined by the content analysis have been formulated. These
hypotheses relate to two aspects: (i) volume of social media posts, and (ii) subjects of social media posts. The
content analysis has been designed and executed in a manner to examine these hypotheses, of which the results
are discussed in this section.
4-6-1 Volume of Social Media Posts related to Firms
The collecting process of social media messages related to the ?rms in our sample resulted in di?erent amounts
of messages for di?erent ?rms. E.g. for Heineken, 39.425 messages have been collected while only 428 posts have
4-6 Interpretation of the Results 69
20%
30%
40%
50%
60%
70%
80%
90%
100%
P
e
r
c
e
n
t
a
g
e
o
f
S
o
c
i
a
l
M
e
d
i
a
P
o
s
t
s
0%
10%
20%
Consultancy, research and
other specialised business
services
Mining and quarrying Financial institutions Transport and storage Information and communication Wholesale and retail Industry
Short-termfinancial results Customer relations Community Other
Figure 4-20: Social Media Posts Subject Classi?cation
The ?gure consolidates social media messages of ?rms in the same industries. The ?gure illustrates that the subjects
of social media messages related to ?rms di?ers per industry type.
been found that are related to Fugro. We constructed a variable labelled average daily mentions by dividing
the total amount of collected posts for a ?rm by the amount of days that the ?rm has been monitored. The
variable average daily mentions is deemed as the variable that re?ects the volume of social media posts. In our
sample, the mean average daily mentions amounts 1.025 messages per ?rm per day.
The average daily mentions di?ers from ?rm to ?rm. In order to draw generic conclusions, i.e. not ?rm speci?c,
the ?rms were ?rstly grouped into either business-to-business or busines-to-consumer ?rms. As the analysis
showed, ?rms conducting a B2C relation will ?nd more social media posts that are related to them than ?rms
active in the B2B sector. In our sample, we ?nd an average of 130 daily mentions of a B2B ?rm, while a B2C
?rm is daily – on average – 1.369 times mentioned. Secondly, the ?rms have been grouped into seven industries.
Again, the average daily mentions have been analysed and di?erences between the volumes are illustrated.
4-6-2 Subjects of Social Media Posts related to Firms
The second aspect of this thesis’ content analysis analyses the subjects of the social media posts. As discussed,
the subjects of the social media messages are to be linked to the ?rm’s key-performance indicators. The collected
social media messages have been classi?ed into one of the 28 categories of social media messages that have been
established. Figure 4-18 (page 67) illustrates that subjects of social media messages di?er from ?rm to ?rm.
In order to draw generic conclusions of the subjects of the social media messages related to di?erent ?rms, the
?rms have ?rstly been grouped based on their customer relation type. We can conclude that the subjects of
social media messages related to B2B ?rms contain a higher percentage of short term ?nancial results, news
and professionals related messages than messages related to B2C ?rms. Next, the analysis indicates that the
social media messages related to B2C ?rms contain a higher percentage of posts related to customer relations,
product and service quality and product and service innovation than messages related to B2B ?rms. Secondly,
?rms have been grouped into seven di?erent industries. In the same way as with the analysis of the volume
of social messages, we ?nd that the subjects of social media messages di?er among the ?rms participating in
the di?erent industries of our sample. The majority of social media messages related to ?rms (41%) express
how the external stakeholders of a ?rm perceive the company. In this thesis, such posts have been classi?ed as
community posts. 18% of the social media messages in our dataset contained the name of a ?rm, but did not
70 Content Analysis
contain any valuable information for the ?rm and have consequently been assigned as unde?ned posts. About
11% of the social media messages relate to ?nancial results, which consist of ?nancial performance discussions
(5%) and stock related discussions (6%). Table 4-11 lists the interpretation of the results of the content analysis
in a summarily manner.
Table 4-11: Conclusion of Content Analysis
Volume of Social Media Messages Subjects of Social Media Messages
Customer
Relation
B2C ?rms are more often subject of
discussion on social media than B2B
?rms.
B2C ?rms related social media messages are more
often subjected to customer relations, product and
service quality and product and service innovation
than B2B related ?rms. On the other hand, B2B
?rms’ related messages are more often subjected to
?nancial results, news and professionals discussing
the ?rm. However, the information contained in the
social media posts of B2B ?rms is often yet available
to the ?rm, hence not o?ering added value to the
?rm’s richness of management information.
Industry Our analysis shows a variation in
the volume of social media messages
across di?erent industries.
Our analysis indicates that there is a di?erence in
the subjects of social media posts related to ?rms in
di?erent industries.
4-7 Sub Conclusion: Social Media Posts that relate to KPI Categories
and the Performance Prism Perspectives
Firstly, our analysis indicates that there exists a di?erence in the volume of ?rm-related social media messages
that are daily generated. These di?erences are indicated when comparing ?rms with ?rms, but also when we
compare between B2B and B2C ?rms and when a comparison between di?erent industries is made. With respect
to the volume of ?rm-related social media content, we can state that especially B2C ?rms are able to collect
social media data for business intelligence purposes because it are these ?rms that are subject of interest on
social media sites.
Secondly, in our analysis social media messages have been assigned to di?erent categories of KPIs. The assigning
of messages to KPI categories was based on the subject of the messages. As the analysis indicated, it di?ers
from ?rm to ?rm which kind of KPIs are candidates to be measured using social media data. When taking a
customer relation type perspective, the analysis indicates that KPIs related to the community, i.e. the metrics
that re?ect the attitude of external stakeholders towards the ?rms, are the ones that are particularly suited
to be measured using social media data. Social media messages related to community metrics provide a ?rm
with insight that cannot be generated with internal systems, the information contained in these messages are
created by individuals discussing the ?rm and/or the ?rm’s products / services. Additionally, we see that a
substantial part of the social media messages related to B2C ?rms are related to customer relations metrics.
These messages contain questions and/or complaints of customers and are purposed to get in contact with the
?rm. Several ?rms embrace these messages by establishing a web care team that actively responds to customers
writing messages purposed to contact the ?rm. As regards to B2B ?rms, we see that a high percentage of
the social media posts relate to short-term ?nancial results. Regrettably, the information in these messages
are also available to the ?rm without the existence of social media messages. Most likely, the ?rm is aware of
the information in these messages before it is available on social media. Next, we see that the percentage of
professionals posts (a sub category of community) is higher for B2B ?rms than for B2C ?rms. These messages
contain valuable information for the ?rm, such as market analyses and the position of the ?rm in that situation,
or forecasts for macro economic developments and the e?ects on the ?rm and/or the ?rm’s ecosystem. However,
we have to bear in mind that the volume of social media messages related to B2B ?rms is much lower than for
B2C ?rms.
In chapter 2, a framework has been presented that illustrates the relation between the ?ve performance prism
perspectives and ten categories of key-performance indicators (page 23). As indicated, this thesis examines
possibilities to measure operational performance – i.e. the key-performance indicators of a ?rm – by means of
4-7 Sub Conclusion: Social Media Posts that relate to KPI Categories and the Performance Prism Perspectives 71
4%
Stakeholder
Satisfaction
Strategies
Processes
Capabilities
Stakeholder
Contribution
1. Short-termfinancial results
2. Customer relations 3. Employee relations
4. Operational performance
5. Product and service quality
6. Alliances
7. Supplier relations
8. Environmental performance
9. Product and service innovation
10. Community
KPI Category
Performance PrismPerspective
Social Media Data
9%
11%
0%
0%
0%
1%
41%
0%
Represented in social media data
1%
Not represented in social media data
Mediumrepresented in social media data
Figure 4-21: Social Media Data related to Key-Performance Indicators
Figure illustrates the link between social media data, key-performance indicators and performance prism perspectives.
The colours indicate for which KPI categories - and hence for which performance prism perspectives - social media
data can be found that relates to these categories.
social media data. With the knowledge we gained with a content analysis of a sample of social media messages
related to ?rms, we can conclude which categories of KPIs are candidates to be measured using social media data.
Figure 4-21 schematically shows for which type of KPIs there exists social media data that is related to these
KPIs. KPI categories that were not represented in the sample – i.e. smaller than 1% of the total messages that
have been classi?ed – have been coloured red, implying that the respective KPI category is under-represented in
social media data and hence not able to be measured by means of social intelligence. These under-represented
KPI categories are related to alliance metrics, supplier relations metrics, environmental performance metrics
and operational performance metrics. Next, categories of KPIs that were somehow represented in the analysed
messages – i.e. between 1% and 5% of all classi?ed messages – have been coloured orange. As can be concluded,
these KPI categories relate to employee relations, product and service quality and product and service innovation.
Finally, categories of KPIs that were – in comparison with the other categories – highly represented in the sample
data have been coloured in green. Formally, KPI categories of which more than 5% of the sample data could
be assigned to the respective category have been coloured green.
Consequently, since section 2-4 yet established links between the ten KPI categories and the ?ve performance
prism perspectives, we can draw conclusions on the applicability of social business intelligence for the di?erent
performance prism perspectives. Corresponding to the colours of the KPI categories, the ?ve performance
prism perspectives have been coloured red, yellow or green. The colours represent the applicability of social
intelligence for the di?erent perspectives. As can be concluded, performance metrics in the domains stakeholder
contribution and stakeholder satisfaction are especially suited to be measured using social media data.
The results of the content analysis showed that the applicability of social business intelligence di?ers from ?rm
to ?rm. Especially ?rms B2C ?rms are likely to ?nd ?rm-related messages. Whereas ?gure 4-21 shows the
overall percentages of social media messages related to the di?erent KPI categories, these ?gures vary from ?rm
to ?rm and are thus higher for B2C ?rms. For a detailed overview, see appendix D).
Chapter 5
Blueprint of a Social Business Intelligence
Procedure
Chapter 4 illustrated that ?rm-related social media messages contain information that can be linked to a ?rm’s
key-performance indicators. However, chapter 4 also showed that not all categories of KPIs can be linked to
the content created on social media, simply because the user-generated social media content does not relate
to all categories of key-performance indicators. For those KPIs that are related to the subjects of the social
media messages, a procedure is required that prescribes how a ?rm should acquire and process these social
media messages for business intelligence purposes. A blueprint for a procedure in which social media data are
collected and processed in a way that corresponds with companies’ general business intelligence processes is
developed in this chapter. We refer to such a procedure as a social business intelligence procedure. Given the
insights gained in chapter 4, we state that the following chapter is only relevant for certain ?rms; ?rms that
are mentioned on social media. Firms that are unable to ?nd related social media data should not invest in the
development of social business intelligence procedures.
Section 5-1 starts this chapter by formulating the requirements of a social business intelligence procedure.
Next, section 5-2 provides the blueprint of the procedure, and discusses the necessary steps that form the
procedure. In section 5-3 the procedure is veri?ed. In section 5-4, the real-time aspect of social business
intelligence is discussed. In section 5-5 traditional business intelligence procedures are compared with social
business intelligence. Finally, section 5-6 concludes the ?ndings of this chapter.
5-1 Requirements Formulation
Based on the business intelligence concepts that are discussed in chapter 2, the possibilities of social media
monitoring tools that are discussed in chapter 3, and the experience that we gained in chapter 4 by performing
a content analysis on the social media messages related to di?erent ?rms, nineteen requirements for a business
intelligence procedure have been formulated. Section 5-1-1 describes these requirements. In section 5-1-2 the
formulated requirements are veri?ed on consistency with general business intelligence procedures.
5-1-1 Description of Requirements
Table 5-1 lists the requirements for a social business intelligence procedure. These requirements are discussed
in the following sections. A social business intelligence procedure should:
1. Have access to social media platforms
Obviously, to acquire intelligence from social media messages, a ?rm should have access to the platforms
where these messages are produced.
2. Identify the social media platforms at which the ?rm is discussed
The fundamental purpose of social intelligence is to acquire insight in the the perception of the ?rm’s
5-1 Requirements Formulation 73
Table 5-1: Requirements
A social business intelligence procedure should . . .
1 Have access to social media platforms
2 Identify the social media platforms at which the ?rm is discussed
3 Identify the volume of social media messages related to the ?rm
4 Remove the spam from social media messages that initially seemed to relate to the ?rm
5 Anonymise personal data
6 Identify who the people are that discuss the ?rm on social media
7 Identify what the subjects of the social media messages related to the ?rm are
8 Determine whether the information contained in the social media messages related to the ?rm
o?ers additional value
9 (Automatically) Classify the social media messages related to a ?rm into categories
10 Relate the (categories of) subjects of the social media messages to the ?rm’s key-performance
indicators
11 Determine the ?rm’s social reputation
12 Determine the social reputation of the ?rm’s product(s)
13 Determine relations between social media metrics and the ?rm’s (social) key-performance indicators
14 Update the status of the social media metrics and the values of the KPIs constantly
15 Present the slope of the relations between social media metrics and KPIs on a time chart
16 Interpret the gained intelligence and position it into the ?rm’s developments
17 Assign the gained intelligence to the right persons in a ?rm
18 Allow a ?rm to engage on social media platforms
19 Regularly update the search terms to anticipate on changes
external stakeholders – including (potential) customers, competitors and parter ?rms – towards the ?rm
and/or the ?rm’s products/services. As illustrated in chapter 4, the majority of social media messages
that are related to ?rms, and publicly accessible, are written on Twitter. However, the distribution of
the platforms where the ?rm is discussed may vary from ?rm to ?rm. Therefore, before starting the
monitoring of social media messages, a ?rm should investigate where – i.e. on which platforms – the ?rm
is subject of discussion.
An overview of the platform distribution provides a ?rm insight into which social media platforms the
?rm should focus, engage or advertise. Though a ?rm may be subject of discussion on multiple platforms,
it does not imply that the ?rm is required to monitor these platforms individually. The social media
monitoring tools o?er the possibility to monitor and engage on multiple social media platforms through
one dashboard.
3. Identify the volume of social media messages related to the ?rm
The fact that a ?rm is subject of discussion on social media is of less value whenever there are little
messages available for a ?rm to analyse. Furthermore, the existence of more social media messages related
to a ?rm o?er opportunities to identify correlations between the amount of these messages and the ?rm’s
KPIs. Such analyses are of less value when there are little social media messages available. As chapter 4
revealed, the volume of social media messages di?ers from ?rm to ?rm. Especially business-to-consumer
?rms are discussed on social media, implying that these ?rms have the opportunity to acquire social
intelligence. A constant monitoring of the amount of social media messages related to the ?rms allows
for the detection of sudden deviations, illustrating that there is “something going on”, which may require
attention from the ?rm’s management.
4. Remove the spam from social media messages that initially seemed to relate to the ?rm
As experienced in chapter 4, many social media messages contain the name of the ?rm in the post, though
they do not relate to the ?rm. Especially ?rms carrying a commonly used name or an abbreviation (such
as KLM), are likely to receive many spam messages in their social media messages. Though it may help
to use speci?c user names for the ?rm’s web care team (e.g. @KLM_WebCare), the drawback of such a
name is that the ?rm will not detect all ?rm related messages since users will nevertheless use the generic
name in their posts. Spam related messages are to be removed from the dataset since they do not contain
any value for the ?rm.
5. Anonymise personal data
74 Blueprint of a Social Business Intelligence Procedure
As illustrated in section 3-4, the European Commission has drafted new Regulation on processing personal
data. As a consequence, ?rms are not naturally allowed to process data that allows one to retrace a natural
person from that data. In order to be in compliance with the expected new Regulation, ?rms intending
to collect and process social media data should anonymise the personal data.
6. Identify who the people are that discuss the ?rm on social media
Though A. M. Kaplan and Haenlein (2010) argue that the usage of social media is diversifying in terms
of the users’ age, it is wise to determine who the people are that discuss the ?rm on social media. It is
out of scope of this thesis to describe how di?erent customer groups (e.g. di?erent generations, men /
women, di?erent cultures) should be treated, but whenever a ?rm decides to engage into the social media
conversations, it should be aware of the people that make up their social media environment. Furthermore
a ?rm can decide that it does not consider the people that produce the social media messages as critical
customers, and therefore does not undertake any action.
7. Identify what the subjects of the social media messages related to the ?rm are
Whereas the second requirement of a social business intelligence procedure ensures that a ?rm has insight
in the amount of messages that are produced and containing the ?rm’s name, it is also valuable for a
?rm to have insight in what it is that social media users discuss in relation with the company. The
identi?cation of the subjects of social media posts forms the basis for the translation of social media posts
to key-performance indicators. Furthermore, the identi?cation of subjects – combined with the volume
of messages – provides a ?rm with insight in the topics that are “trending”, i.e. popular topics at the
moment. Trending topics related to ?rms can serve as a measure describing what people consider as
important, and which may be action points for the ?rm.
8. Determine whether the information contained in the social media messages related to the
?rm o?ers additional value
The content analysis of a set of social media messages related to ?rms revealed that there are also messages
that do contain the ?rm’s name, but neither do contain information that is of any value for the ?rm. We
have classi?ed posts of no value as unde?ned posts.
Furthermore chapter 4 illustrated that there are also messages that contain information that must be
available to the ?rm without analysing the social media. Especially for B2B ?rms, many social media posts
contain information about the ?nancial performance or share prices of the ?rm. Generally, a publication
or press article has been the source of these messages. These messages do not contain information that is
not available in the ?rm yet, and are therefore considered of less value for the ?rm.
9. (Automatically) Classify the social media messages related to a ?rm into categories
The unstructured character of social media posts makes it that these messages have to be preprocessed
before an analysis can commence. Classifying the messages into categories, e.g. into categories of subjects
(as we have done in chapter 4), categories of languages, categories of men and women, categories of
many or less followers, etc. allows a ?rm to structurally analyse the messages and derive that particular
information that the ?rm is interested in.
The unstructured nature and the large amount of messages that are generated in relation with some ?rms
makes it that social media data can be termed as “big data”. It is therefore desired that the classi?cation
process of the social media messages into categories runs automatically. Automatic classi?ers are existing
solutions to this problems, and are also available for text. These classi?ers require so-called “training sets”
in order to establish criteria at which a piece of text is either classi?ed in e.g. category A or in B. As we
have experienced, the subjects of social media messages di?er from ?rm to ?rm. Two social media posts
containing the word “Senseo” and “TomTom One XL” are both related to a product, but do not contain
the same words. Therefore, training sets should be established for speci?c ?rms. Our classi?cation can
be used to train classi?ers for the ?rms that participated in the sample of this thesis.
10. Relate the (categories of) subjects of the social media messages to the ?rm’s key-performance
indicators
As we have showed in chapter 4, it is possible to classify social media posts into categories that are related
to KPIs. The subjects of the social media posts serve as the basis to assign a certain social media message
to a certain key-performance indicator. For example, when a ?rm manages by a KPI representing the
customer satisfaction towards a certain product, it can use the social media messages related to that
product as a measure to determine the satisfaction level. As we have seen in the previous chapter, TV
commercials are also subject of discussion on social media. A ?rm may determine the success of such a
campaign by counting the messages that relate to the commercial.
5-1 Requirements Formulation 75
11. Determine the ?rm’s social reputation
A social business intelligence procedure should determine the ?rm’s reputation on social media. The
volume of messages related to the ?rm is not of any value whenever there is no insight in the nature
of these messages since it matters whether or not these messages are positive or negative. Social media
monitoring tools o?er the possibility to determine the sentiment of a social media post. Generally, posts
are classi?ed as either positive, neutral or negative. As section 3-3-1 illustrated, the sentiment analyses
may not always be as reliable. However, as we expect, sentiment analysis tools will be improved and able
to determine the sentiment of the most more accurate. The ?rm’s social reputation – e.g. measured by the
percentage of positive posts related to the ?rm – is an interesting indicator that may reveal correlations
with other KPIs, such as sales.
12. Determine the social reputation of the ?rm’s product(s) / service(s)
Whereas it is necessary to determine the ?rm’s social reputation, a ?rm may be interested in the reputation
of a particular product or service that it provides. Again, sentiment analysis is required for theses posts.
The social reputation of products – e.g. measured by the percentage of positive posts related to that
product or service – may reveal correlations with the sales or amount of returns of that product.
13. Determine relations between social media metrics and the ?rm’s (social) key-performance
indicators
One of the fundamental purposes of business intelligence is to identify which activities of a ?rm deliver
value. In order to determine which social media metrics actually relate to the ?rm’s, the social business
intelligence procedure should contain a step in which the relations between social media metrics and the
?rm’s KPIs are determined. An example of such a relation may be the amount of positive messages about
product x and the sales in a certain period of product x.
14. Update the status of the social media metrics and the values of the KPIs constantly
In order to develop real-time business intelligence, the system should automatically monitor the social
media metrics. “This will only be satis?ed whenever the right KPIs are de?ned before the metrics are
monitored” (Azvine et al., 2005). As such, the ?rm gets insight in the values of the social media metrics
and the values of the KPIs.
15. Present the slope of the relations between social media metrics and KPIs on a time chart
Whereas the previous requirement ensures that the information that is derived from social media is
presented in real-time, this requirement ensures that the slope of the values are presented in a way so that
deviations over time are easily recognised. Sudden events may trigger social media metrics to ?uctuate,
these events are able to be noti?ed when the values are presented in a time chart.
16. Interpret the gained intelligence and position it into the ?rm’s developments
Whereas the derived intelligence may reveal relations of social media metrics and KPIs and provide insight
in the external stakeholders’ perceptions of the ?rm, one should always position this intelligence in the
light of developments of the ?rm.
17. Assign the gained intelligence to the right persons in a ?rm
When it turns out that certain KPIs are in?uenced by social media metrics, and these KPIs are not
performing su?ciently, the acquired intelligence should be communicated to the responsible departments
in the ?rm. The departments can provide clarifying factors for the under performing KPIs, and can take
the acquired intelligence (e.g. related to the feature of a certain products) into their decision-making
process.
18. Allow a ?rm to engage on social media platforms
A social business intelligence procedure should allow ?rms to engage with the users on social media. As
the content analysis in chapter 4 illustrated, many ?rms engage in the social media discussions. Though
we cannot verify this statement, it is expected that there will be generated more user-generated content
whenever a ?rm actively participates on social media. We will elaborate about this statement in the
further research section (section 6-6).
19. Regularly update the search terms to anticipate on changes
An up-and-running social business intelligence procedure has been started by search terms that are related
to the ?rm. Since a ?rm is always in development, it will launch new products, services and employees
will come and go. Therefore, the search terms should be updated whenever there are events that in?uence
the required search terms. For example, whereas Microsoft’s search terms include “Windows 7”, it should
add “Windows 8” to these search terms by the time it launches – or pre-launches – this new product.
76 Blueprint of a Social Business Intelligence Procedure
5-1-2 Requirements Check on Business Intelligence Concepts
Chapter 2 described the business intelligence concept as it is applied within ?rms. Especially section 2-3
elaborated about the activities that make up the business intelligence process. Van Beek (2006) argues that a
BI process consists of three main tasks, being (i) registering, (ii) processing and (iii) reacting. Additionally, the
processing task consists of 15 sub tasks required to process the registered data. In total, 17 (1+15+1) tasks can
be distinguished that are required for a business intelligence process. We verify the requirements for a social
business intelligence procedure – that have been established in section 5-1-1 by controlling whether each of the
BI steps are represented by at least one of the requirements that we have established.
As can be seen concluded from table 5-2 (page 77), each activity is represented by at least one requirement.
This allows us to conclude that the requirements of the social business intelligence procedure are consistent
with existing BI procedures.
5-1 Requirements Formulation 77
T
a
b
l
e
5
-
2
:
B
u
s
i
n
e
s
s
I
n
t
e
l
l
i
g
e
n
c
e
A
c
t
i
v
i
t
i
e
s
a
n
d
S
o
c
i
a
l
B
u
s
i
n
e
s
s
I
n
t
e
l
l
i
g
e
n
c
e
R
e
q
u
i
r
e
m
e
n
t
s
S
o
c
i
a
l
B
u
s
i
n
e
s
s
I
n
t
e
l
l
i
g
e
n
c
e
R
e
q
u
i
r
e
m
e
n
t
s
R e g i s t e r
C o l l e c t
F i l t e r
C o m b i n e
A g g r e g a t e
V i s u a l i s e
I n t e r p r e t
I n t e r n a l i s e
R e v i s e
V e r i f y
E n r i c h
S h a r e
R e m e m b e r
D e c i d e
D i s t r i b u t e
A n t i c i p a t e
R e a c t
1
A
c
c
e
s
s
t
o
s
o
c
i
a
l
m
e
d
i
a
p
l
a
t
f
o
r
m
s
2
I
d
e
n
t
i
f
y
t
h
e
s
o
c
i
a
l
m
e
d
i
a
p
l
a
t
f
o
r
m
s
a
t
w
h
i
c
h
t
h
e
?
r
m
i
s
d
i
s
c
u
s
s
e
d
3
I
d
e
n
t
i
f
y
t
h
e
v
o
l
u
m
e
o
f
s
o
c
i
a
l
m
e
d
i
a
m
e
s
s
a
g
e
s
r
e
l
a
t
e
d
t
o
t
h
e
?
r
m
4
R
e
m
o
v
e
t
h
e
s
p
a
m
f
r
o
m
s
o
c
i
a
l
m
e
d
i
a
m
e
s
s
a
g
e
s
t
h
a
t
i
n
i
t
i
a
l
l
y
s
e
e
m
e
d
t
o
r
e
l
a
t
e
t
o
t
h
e
?
r
m
5
A
n
o
n
y
m
i
s
e
p
e
r
s
o
n
a
l
d
a
t
a
6
I
d
e
n
t
i
f
y
w
h
o
t
h
e
p
e
o
p
l
e
a
r
e
t
h
a
t
d
i
s
c
u
s
s
t
h
e
?
r
m
o
n
s
o
c
i
a
l
m
e
d
i
a
7
I
d
e
n
t
i
f
y
w
h
a
t
t
h
e
s
u
b
j
e
c
t
s
o
f
t
h
e
s
o
c
i
a
l
m
e
d
i
a
m
e
s
s
a
g
e
s
r
e
l
a
t
e
d
t
o
t
h
e
?
r
m
a
r
e
8
D
e
t
e
r
m
i
n
e
w
h
e
t
h
e
r
t
h
e
i
n
f
o
r
m
a
t
i
o
n
c
o
n
t
a
i
n
e
d
i
n
t
h
e
s
o
c
i
a
l
m
e
d
i
a
m
e
s
s
a
g
e
s
r
e
l
a
t
e
d
t
o
t
h
e
?
r
m
o
?
e
r
s
a
d
d
i
t
i
o
n
a
l
v
a
l
u
e
9
(
A
u
t
o
m
a
t
i
c
a
l
l
y
)
C
l
a
s
s
i
f
y
t
h
e
s
o
c
i
a
l
m
e
d
i
a
m
e
s
s
a
g
e
s
r
e
l
a
t
e
d
t
o
a
?
r
m
i
n
t
o
c
a
t
e
g
o
r
i
e
s
1
0
R
e
l
a
t
e
t
h
e
(
c
a
t
e
g
o
r
i
e
s
o
f
)
s
u
b
j
e
c
t
s
o
f
t
h
e
s
o
c
i
a
l
m
e
d
i
a
m
e
s
s
a
g
e
s
t
o
t
h
e
?
r
m
’
s
k
e
y
-
p
e
r
f
o
r
m
a
n
c
e
i
n
d
i
c
a
t
o
r
s
1
1
D
e
t
e
r
m
i
n
e
t
h
e
?
r
m
’
s
s
o
c
i
a
l
r
e
p
u
t
a
t
i
o
n
1
2
D
e
t
e
r
m
i
n
e
t
h
e
s
o
c
i
a
l
r
e
p
u
t
a
t
i
o
n
o
f
t
h
e
?
r
m
’
s
p
r
o
d
u
c
t
(
s
)
1
3
D
e
t
e
r
m
i
n
e
r
e
l
a
t
i
o
n
s
b
e
t
w
e
e
n
s
o
c
i
a
l
m
e
d
i
a
m
e
t
r
i
c
s
a
n
d
t
h
e
?
r
m
’
s
(
s
o
c
i
a
l
)
k
e
y
-
p
e
r
f
o
r
m
a
n
c
e
i
n
d
i
c
a
t
o
r
s
1
4
U
p
d
a
t
e
t
h
e
s
t
a
t
u
s
o
f
t
h
e
s
o
c
i
a
l
m
e
d
i
a
m
e
t
r
i
c
s
a
n
d
t
h
e
v
a
l
u
e
s
o
f
t
h
e
K
P
I
s
c
o
n
s
t
a
n
t
l
y
1
5
P
r
e
s
e
n
t
t
h
e
s
l
o
p
e
o
f
t
h
e
r
e
l
a
t
i
o
n
s
b
e
t
w
e
e
n
s
o
c
i
a
l
m
e
d
i
a
m
e
t
r
i
c
s
a
n
d
K
P
I
s
o
n
a
t
i
m
e
c
h
a
r
t
1
6
I
n
t
e
r
p
r
e
t
t
h
e
g
a
i
n
e
d
i
n
t
e
l
l
i
g
e
n
c
e
a
n
d
p
o
s
i
t
i
o
n
i
t
i
n
t
o
t
h
e
?
r
m
’
s
d
e
v
e
l
o
p
m
e
n
t
s
1
7
A
s
s
i
g
n
t
h
e
g
a
i
n
e
d
i
n
t
e
l
l
i
g
e
n
c
e
t
o
t
h
e
r
i
g
h
t
p
e
r
s
o
n
s
i
n
a
?
r
m
1
8
A
l
l
o
w
a
?
r
m
t
o
e
n
g
a
g
e
o
n
s
o
c
i
a
l
m
e
d
i
a
p
l
a
t
f
o
r
m
s
1
9
R
e
g
u
l
a
r
l
y
u
p
d
a
t
e
t
h
e
s
e
a
r
c
h
t
e
r
m
s
t
o
a
n
t
i
c
i
p
a
t
e
o
n
c
h
a
n
g
e
s
78 Blueprint of a Social Business Intelligence Procedure
5-2 Social Business Intelligence Procedure
A blueprint for a social business intelligence (“SBI”) procedure has been developed, in which all requirements
of section 5-1 are taken care of. An aggregate overview of the procedure is presented in ?gure 5-1.
Reaction based on
social intelligence
Strategic
mapping of
KPIs
Reacting
Search terms
Action plan(s) to
respond to gained
intelligence
g
Collecting
Mapping
insights to
business
units
Data pre
Unstructured
data
Information for
business units
Data pre-
processing
Categorising
Analysing
Structured, combined
(and anonymised) data
Categorised data
Figure 5-1: Blueprint: Social Business Intelligence Procedure
The SBI procedure consists of seven main activities that are related to each other, as ?gure 5-1 illustrates. Each
of the main activities are further exempli?ed in the following sections.
5-2-1 Strategic mapping of KPIs
Firms deduct key-performance indicators from their strategy. This process is further elaborated in section 2-2.
The KPIs that a ?rm eventually established are to be measured. As the content analysis of this thesis revealed,
some KPIs are not appropriate to be measured by social media because there does not exist any content that
related to these KPIs. Other KPIs are best measured by internal systems, and some KPIs are properly measured
by social media. From the list of KPIs that a ?rm uses, a selection can be made of indicators that are to be
measured by social media. Illustratively, ?gure 5-2 highlights the KPIs that are to be measured by social media.
These KPIs form the starting point of the social business intelligence procedure, since it are these KPIs for
which social media data is to be collected and analysed.
Strategy
Driver z Driver y Driver x
KPI 4 KPI 3 KPI 2 KPI 1 KPI 6 KPI 5
KPI 1 KPI 4
I. Strategic mapping of KPIs
KPI 3
S
o
c
ia
l M
e
d
ia
K
P
Is
II. Collecting
Social media
categories available
in the data set,
newaction points
Keywords
VII. Reacting
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Figure 5-2: Blueprint: Social Business Intelligence Procedure (Strategic Mapping of KPIs)
The KPIs selected to be measured by social data determine the categories that are to be analysed – i.e. the
subjects of social media messages – and hence the keywords that are to be used in the collecting process. On
the other hand, the available social media data determines whether or not it is possible to measure the KPI by
social media data. After all, a KPI for which no related social media data exist, can not be measured by social
data. Thus, there exists an interaction between on the one hand what a ?rm wants to measure by social media
5-2 Social Business Intelligence Procedure 79
data, and on the other hand what a ?rm is possible to measure using social media data. As we have seen in
chapter 4, not every KPI is subject of discussion on social media.
5-2-2 Collecting
After the ?rst step, in which the KPIs that are to be measured by social media data have been selected, the
data is to be collected. The step is schematically represented in ?gure 5-3
II. Collecting
Blogs
YouTube
News sites Etc.
Listening to social media channels
@Company_name
#service_x
#event_z
#product_y
@Competitor_na
me
Etc.
Select keywords
I. Strategic mapping of KPIs
II. Data Pre-Processing
Unstructured data
Search
terms
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Figure 5-3: Blueprint: Social Business Intelligence Procedure (Collecting)
Keywords related to the ?rm, the ?rm’s products/services and the selected KPIs are used to “listen” to multiple
social media channels at which the ?rm could be mentioned. The content analysis of this thesis revealed that it
di?ers per ?rm on which social media platform the ?rm is discussed. It is therefore that the ?rst step involving
social media platforms consists of the determination of the platforms at which the ?rm is discussed. As we have
experienced in chapter 4, search queries related to ?rms will result in unstructured data from multiple social
media platforms. These unstructured data are to be pre-processed, which is the next step in the SBI procedure.
5-2-3 Data Pre-Processing
The third step in the social business intelligence procedure consists of pre-processing the collected data. In
contrast to ‘regular’ BI data, social media data is unstructured, sourced from multiple platforms, containing
spam and personal data, and is therefore required to be pre-processed. Figure 5-4 illustrates this process.
III. Data Pre-Processing
Select attributes to
analyse
Combine different
data formats
IV. Categorising
II. Collecting
Unstructured data
Remove
duplicates
Remove spam
Data ready to be categorised
Anonymise
personal data
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Parse data in
table / database
Figure 5-4: Blueprint: Social Business Intelligence Procedure (Data Pre-Processing)
The collected data consists of social media messages that are sourced from multiple sources in di?erent formats,
such as CSV, JSON, XML, etc. Each data source may employ its own structure of social media messages,
80 Blueprint of a Social Business Intelligence Procedure
and not each platform may contain the same richness in attributes as the other. For instance the Twitter API
o?ers developers the opportunity to extract so called geotags – geographic coordinates of the origination of
the Tweet – while other social media platforms do not o?er this attribute to the messages. Each social media
post should be parsed – structured – into one and the same data format. Next, as we have experienced in
the scraping process of chapter 4, multiple search queries will lead to multiple messages yet available in the
database. Therefore, only social media messages that do not exist in the table should be added. The ?nal
step in the data pre-processing step consists of the removal of spam. After the data pre-processing has been
completed, the data is structured, clean and ready to be categorised.
5-2-4 Categorising
The third step in the SBI procedure consists of categorising the social media posts. The purpose of this step is
to divide the messages into clustered categories at which the ?rm is interested. The aspects at which the social
media posts are categorised may vary. Figure 5-5 schematically shows the third step of the SBI procedure.
IV. Categorising
Categorise the
data on specific
aspects
People
Competitive data
Feedback about competitors’
products
Feedback on people’s attitude to
competitors’ organisation
Feedback about latest advertising
campaign
Feedback about product features
Feedback about howpeople
perceive the brand / company
People’s opinions
Data about requests for customer
service
Data about the performance of
customer service
Feedback on the pricing of your
products / services
Example classifications
Less followers
Many followers
Positive speakers
Encouragers
Promotors
Negative speakers
Complainers
Saboteurs
Subjects e.g.
Example classifications
Public image
Customer relations
Recruitment
Product and service quality
Product and service innovation
Professionals’ opinions
Etc.
Trending Topics
III. Data Pre-Processing
Data ready to be categorised
V. Analysing
Categorised data
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Figure 5-5: Blueprint: Social Business Intelligence Procedure (Categorising)
One can decide to analyse the people that create the messages, and group these people in e.g. people with
many/less followers or friends, or into people that write/negative positive about the ?rm. We have labelled
the four categories of people. Encouragers are the people with less followers though speak positive about the
?rm or its products. Complainers are the people with less followers and write negative about the ?rm. People
with many followers who speak positive about the ?rm have been labelled as promoters, while people with
many followers writing negative have been termed saboteurs. An analysis of the people provides the ?rm with
intelligence about the power of the people that write about the ?rm, and may form the starting point of a social
media engagement strategy.
Another aspect at which social media messages may be classi?ed is based on their subjects. Our content analysis
of chapter 4 also categorised social media messages based on their subjects. The subjects that were represented
in our dataset related to public image, customer relations, recruitment, product and service quality, product and
5-2 Social Business Intelligence Procedure 81
service innovation, professionals’ opinions, etc. By classifying posts into categories based on subjects, it becomes
possible to link the volume of messages related to a certain subject to the companies’ corresponding KPIs. For
instance, public image posts – which may be additionally classi?ed as positive, neutral or negative – are related
to a customer satisfaction KPI. There are plenty of other categories that one can think of to categorise social
media messages, but to link the ?rm’s KPIs to social media data, one should classify the messages based on
their subjects.
Whereas the data on social media is generally publicly accessible, it is possible for a ?rm to perform the same
analysis based on search queries related to competitors and competitors’ products. As such, a competitive
analysis provides the company intelligence about their position with respect to the market average.
Furthermore, word counts can be used to determine so called trending topics; topics that are over-represented in
the social media messages related to the ?rm. Trending topics, or a top 10 of the words that are most frequently
used in the social media messages, provide a ?rm insight in the topics that are discussed on social media in
relation with their ?rm.
5-2-5 Analysing
Once the social media data has been structured and cleaned, the analysis of these data can commence in step 5
of the SBI procedure. It is in this step of the procedure where a translation is made from data to information.
This step is visualised in ?gure 5-6. Depending on the matter of interest, a ?rm can analyse a variety of data
and relations. It would be wise to at least plot the conversation volume – or amount of social media messages
related to the ?rm – against the di?erent social media channels to determine where the conversations related to
the ?rm take place. Next, whenever a category has been established in step 4 in which all messages related to a
certain product or product feature have been grouped, it is possible to determine the attitude of the public to
the product by applying sentiment analysis on these data. Such analysis provides the ?rm with insight in the
the products or product features that are to be improved. Furthermore, a comparable analysis on competitors’
social media data will show the ?rm’s position pertaining to the competitors and competitors’ products.
The most valuable intelligence will be gained when the ?rm combines the social media metrics – such as amount
of mentions, sentiment, messages originating from a certain region, etc. – with the companies’ KPIs, such as
sales volume, market share, customer satisfaction and the amount of customers. The slopes that will be gained
when these metrics are together plotted on a time chart may reveal relations. The right part of ?gure 5-6
illustrates such graphs. A correlation analysis may con?rm these relations. The intelligence that is gained in
the analysis phase may reveal that certain social media metrics are under performing, and that these social
media metrics in?uence key-performance indicators of the company. Consequently, a ?rm may undertake actions
to improve these metrics.
5-2-6 Mapping insights to Business Units
Key-performance indicators are related to di?erent departments in a ?rm, and the managers of these departments
may clarify the under-performance of the metrics and they may suggest actions to improve the KPIs. As ?gure
5-7 illustrates, the intelligence provided by step 5 should be communicated to the responsible business units.
Especially when under-performing KPIs are discovered.
For example, insights related to products should be communicated to the ?rm’s research & development
department, customer satisfaction intelligence to the ?rm’s customer relations management department, etc. It
are the employees of the responsible departments who posses the knowledge and experience to reason why a
KPI is under-performing, and – in collaboration with social media experts – are the ones who may develop an
action procedure to improve the indicator.
5-2-7 Reacting
The ?nal step in the SBI procedure comprises the execution of action plans required to improve under-performing
KPIs by means of social media. Illustratively, ?gure 5-8 shows two type of actions that may result from the
social business intelligence procedure. A ?rm can for instance decide to review its products (features) based
on complaints and suggestions that the SBI procedure provided. Or, a ?rm may decide to intervene in social
media discussions, for instance because customer satisfaction turned out to be low, and – at the same time –
the customer service of the ?rm turned out to be insu?cient.
82 Blueprint of a Social Business Intelligence Procedure
V. Analysing
C
o
m
p
e
t
it
o
r
x
C
o
m
p
e
t
it
o
r
y
F
ir
m
Competition
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
T
w
it
t
e
r
B
lo
g
s
F
a
c
e
b
o
o
k
Channel
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
Map KPIs on the graphs, e.g.:
jan apr jul oct
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
Sales volume
S
a
le
s
v
o
lu
m
e
20,000 1,000
Conversation volume
jan apr jul oct
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
Customer satisfaction
C
u
s
t
o
m
e
r
s
a
t
is
f
a
c
t
io
n
20,000 1
Conversation volume
jan apr jul oct
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
Market share
M
a
r
k
e
t
s
h
a
r
e
20,000 100%
Conversation volume
abc
Product feature
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
Sentiment
Determine
correlations to
identify
dependencies
Map social media reputation
Flag
underperforming
social media
metrics
jan apr jul oct
C
o
n
v
e
r
s
a
t
io
n
v
o
lu
m
e
Customer database
C
u
s
t
o
m
e
r
d
a
t
a
b
a
s
e
20,000
10,000
Conversation volume
VI. Mapping insights to
business units
IV. Categorising
Categorised data
Information for business units
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data
Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Figure 5-6: Blueprint: Social Business Intelligence Procedure (Analysing)
VII. React
VII. Reacting
VI. Mapping insights to Business Units
Customer
Relations
Management
Research &
Development
Marketing & Sales
Customer
satisfaction
information
Feedback related
to products
Sales related
information
Management
board
Competitive
intelligence
Action plan(s) to
respond to
gained intelligence
Etc.
V. Analysis
V. Analysing
Information for
business units
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data
Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Figure 5-7: Blueprint: Social Business Intelligence Procedure (Mapping Insights to Business Units)
5-3 Veri?cation of Procedure 83
VII. Reacting
Review product /
services
Intervene in social
media
conversations
Other activities
Reaction
based on
social intelligence
VI. Mapping insights to
Business Units
Action plan(s) to respond
to gained intelligence
Strategi c
mappi ng of
KPIs
Col l ecti ng
Data pre-
processi ng
Categori si ng
Anal ysi ng
Mappi ng
i nsi ghts to
busi ness
uni ts
Reacti ng
Search terms
Unstructured
data
Structured, combined
(and anonymised) data
Categorised data
Information for
business units
Action plan(s) to
respond to gained
intelligence
Reaction based on
social intelligence
Figure 5-8: Blueprint: Social Business Intelligence Procedure (Reacting)
5-3 Veri?cation of Procedure
The veri?cation of the developed social business intelligence procedure is veri?ed in this section. We will test
whether the requirements established in section 5-1 are ful?lled. For each individual requirement an activity is
searched for that ful?ls the requirement. If all requirements are ful?lled by at least one activity, we can conclude
that the social business intelligence procedure is veri?ed in accordance with the requirements. Table 5-3 (page
84) lists the seven main components of the SBI procedure in the columns, and the eighteen requirements in the
rows of the table. For each requirement, the activity that serves this requirement has been checked. As can be
concluded, each of the eighteen requirements are at least ful?lled by one of the main components. Therefore,
we can conclude that the procedure is in accordance with the requirements, which are in turn in accordance
with the activities required for general business intelligence.
84 Blueprint of a Social Business Intelligence Procedure
Table 5-3: Veri?cation Matrix
S
B
I
C
o
m
p
o
n
e
n
t
s
S
t
r
a
t
e
g
i
c
m
a
p
p
i
n
g
o
f
K
P
I
s
C
o
l
l
e
c
t
i
n
g
D
a
t
a
p
r
e
-
p
r
o
c
e
s
s
i
n
g
C
a
t
e
g
o
r
i
s
i
n
g
A
n
a
l
y
s
i
n
g
M
a
p
p
i
n
g
i
n
s
i
g
h
t
s
t
o
B
u
s
i
n
e
s
s
U
n
i
t
s
R
e
a
c
t
i
n
g
Requirements for a social business intelligence procedure I II III IV V VI VII
1 Access to social media platforms
2 Identify the social media platforms at which the ?rm is discussed
3 Identify the volume of social media messages related to the ?rm
4 Remove the spam from social media messages that initially seemed
to relate to the ?rm
5 Anonymise personal data
6 Identify who the people are that discuss the ?rm on social media
7 Identify what the subjects of the social media messages related to
the ?rm are
8 Determine whether the information contained in the social media
messages related to the ?rm o?ers additional value
9 (Automatically) Classify the social media messages related to a
?rm into categories
10 Relate the (categories of) subjects of the social media messages to
the ?rm’s key-performance indicators
11 Determine the ?rm’s social reputation
12 Determine the social reputation of the ?rm’s product(s)
13 Determine relations between social media metrics and the ?rm’s
(social) key-performance indicators
14 Update the status of the social media metrics and the values of
the KPIs constantly
15 Present the slope of the relations between social media metrics
and KPIs on a time chart
16 Interpret the gained intelligence and position it into the ?rm’s
developments
17 Assign the gained intelligence to the right persons in a ?rm
18 Allow a ?rm to engage on social media platforms
19 Regularly update the search terms to anticipate on changes
5-4 Real-Time Social Business Intelligence 85
5-4 Real-Time Social Business Intelligence
In the introduction of this thesis, the concept of real-time business intelligence has been introduced. One of
the aspects that makes social media data valuable is the fact that it is created real-time and that these data
are directly available. The real-time aspect of social media data is one of the main reasons why this thesis has
been executed in the ?rst place. In this section we pay attention to the speed of the social business intelligence
procedure as it is presented in section 5-2.
In our analysis, a scan for new social media messages has been executed on a daily basis. However, it would be
valuable for a ?rm to be informed directly whenever social media activity related to the ?rm deviates from its
steady state. For example, an increase in the average amount of hourly ?rm-related messages may indicate an
event of which the ?rm should be aware from a risk management perspective. Deviations in the volume of social
media messages are relatively easy to detect, since detection systems simply count the amount of messages that
has been generated in the past period and compare this amount with the average amount. A scan to detect
variations in the volume of ?rm-related messages should be executed periodically, the results are then almost
immediately available. Figure 5-9 shows an illustrative example of a comparison between the average volume
and the actual volume of today. Such a graph would announce a ?rm that it is suddenly more frequent subject
of discussion on social media than in normally. The commercial tool that has been used in this thesis to analyse
social media automatically refreshes the ?rm-related messages, comparable to the streams o?ered by Twitter.
0
10
20
30
40
50
60
70
80
0
-
1
1
-
2
2
-
3
3
-
4
4
-
5
5
-
6
6
-
7
7
-
8
8
-
9
9
-
1
0
1
0
-
1
1
1
1
-
1
2
1
2
-
1
3
1
3
-
1
4
1
4
-
1
5
1
5
-
1
6
1
6
-
1
7
1
8
-
1
9
1
9
-
2
0
2
0
-
2
1
2
1
-
2
2
2
2
-
2
3
2
3
-
0
V
o
l
u
m
e
[
m
e
s
s
a
g
e
/
h
o
u
r
]
Average Today
Figure 5-9: Illustrative example: Variations in volume of ?rm-related social media messages
However, insight in deviations in the volume of ?rm-related social media messages is not su?cient to speak of
social business intelligence. As the procedure in the previous section illustrated, a process of pre-processing,
categorising, analysing and mapping insights to business units is required after a ?rm has determined which
key-performance indicators are to be measured by social media data is required. All these steps are to be
automated by one critical mechanism; text classi?cation. A tool that is able to automatically classify social
media posts into pre-determined categories – e.g. subjects – is a prerequisite for real-time social business
intelligence. As we have seen in the content analysis of this thesis, manually classifying social media posts is
a time-consuming process and therefore not advised. The further development of automatic text-classi?cation
tools and the incorporation of such tools in social media monitoring packages is advised. The programming of
each automatic text classi?er requires a training set of texts that have manually been classi?ed. Our dataset
can be used for such purposes. Furthermore, the automated classi?cation is to be linked with internal data.
So far, business intelligence tools like SAP, Oracle, QlikView, etc. manage and process internal data, whereas
dedicated tools like uberVU, NetBase, Radian6, etc. are used to process external – social – data. Thus, data
from di?erent systems is required to measure the in?uence of social media metrics on organisational performance.
Such systems are to be developed.
In section 3-3-1 we concluded that the current state of social business intelligence can be termed ‘early adoption’.
With the experiences gained in chapter 4 and the required features mentioned in the previous paragraph, the
next (research and development) steps required for social business intelligence systems are determined. These
steps are further elaborated in section 6-6 ‘further research’.
5-5 Social Business Intelligence versus Business Intelligence
The newly developed social business intelligence di?ers from ‘traditional’ business intelligence methods. In this
section, the most important di?erences concerned with social business intelligence are discussed. Table 5-4 lists
86 Blueprint of a Social Business Intelligence Procedure
the main di?erences between the two concepts, which are consequently elaborated.
Table 5-4: ‘Traditional’ and Social Business Intelligence compared
‘Traditional’ Business Intelligence Social Business Intelligence
Data is structured Data is unstructured
Only the values of the data automatically ?uctuate
in the course of time
The nature of the data may alter in the course of
time
Data are mainly numerical Data are mainly textual
Relations between data and KPIs are obvious Relations between data and KPIs are fuzzy
Few sources of data Multiple sources of data
Data sources are internal Data sources are external
One data format Multiple data formats
Origins of the data are known Origins of the data are unknown
Data represents known subjects (products in stock,
sales per month, etc.)
It is beforehand not clear what is contained in the
data (complaints, suggestions, etc.)
Spam
Data may contain personal data for which no explicit
consent of data processing is provided
The far most important aspect of social media data is the fact that it is unstructured. When collecting social
media data, one big mishmash of textual data is found containing di?erent subjects in di?erent languages. On
the contrary, ‘normal’ BI systems source data from structured sources in which pre-determined variables are
stored. As a consequence, the collecting (also referred to as extracting) process will deliver structured data in
any case. In other words, in normal BI systems it is known beforehand what will be measured, whereas the
subjects of social media messages di?er from ?rm to ?rm.
Additionally, the nature of the social media data may alter in the course of time. Due to the fact that anyone
who has access to social media is able to formulate new subjects, it is not unlikely that the subjects contained
in social media messages may vary. This aspect of social business intelligence is not found in normal BI
processes. In normal BI the to be measured variables are pre-determined, the values of these variables will
alter in the course of time. However, the nature of the data will not change. E.g. a metric measuring the
amount of products that are currently in stock will not suddenly start measuring the number of employees
or any other variable. However, subjects contained in social media messages may change. Therefore, social
business intelligence systems are required to cope with ?uctuations in the nature of the data.
Apart from (simple) word counting mechanisms, whether purposed to show the volume of ?rm-related messages
or to show trending topics, social media data is textual in nature. Fundamental data analysis methods and
algorithms underlying traditional business intelligence are developed to process numerical data. Therefore, a
translation step is necessary from raw textual data to numerical data before such analyses can be executed on
social media data. In our social business intelligence procedure this step is provided by categorising the textual
data. The categorising step ensures that equivalent social media posts are grouped – i.e. structured – so that
the number of messages in each group serve as the basis for numerical analyses. E.g. if all messages containing
the subject ‘product x’ have been grouped in one category, the number of the messages in this group are ready
to serve as the input for further (numerical) analyses.
In this thesis we have assigned social media posts to key-performance indicators based on the subjects contained
in these messages. Whereas in traditional business intelligence relations between data and KPIs are obvious
and linear, these relations are less evident in social business intelligence. For instance, the volume of messages
related to a certain product may in?uence the sales of that product. However, this relation is not self-evident
since a causal relation between the two variables is not guaranteed. Other factors – such as lotteries – may
in?uence the chatter volume, but this does not necessarily underwrite the intention of people to buy the product.
The fact remains that these relations may still exist, and it is therefore that monitoring and analysing social
media data in relation with the ?rm’s KPIs may reveal valuable intelligence. As indicated in chapter 4, KPIs
related to customer relations and the perceptions of stakeholders can be measured using social media data.
Whereas in traditional BI data is stored in relatively few (structured) data sources, there are many social media
platforms from which data is to be sourced. These sources di?er from traditional BI sourcing systems since
they are in the external environment of a ?rm, implying that the data formats and other institutional aspects
are determined by the social media platforms. Furthermore, the platforms can change the format in which the
5-6 Sub Conclusion 87
data is delivered. Sourcing data from multiple social media platforms means combining di?erent formats. This
step requires more e?ort than in traditional BI systems. In addition, the attributes that are passed to a ?rm
when it crawls social media data di?ers per platform.
The creators of the data – social media users – are unknown to the ?rm in social business intelligence. As
such, it can be hard to determine the trustworthiness of the data. A user can post whatever he or she wants
on the web, without ensuring that the message actually re?ects his or her opinion or intention. However, there
are ample examples of self-regulating platforms on the web, of which Wikipedia is probably the most famous.
Contributions of users to Wikipedia that are incorrect are automatically corrected by other users with good
intentions. Moreover, since messages are grouped in our procedure, it is rather easy to target the popular topics
(which require attention) and determine the trustworthiness. Next, we expect that natural language processing
tools will be improved, allowing the detection of cynicism and other di?culties concerned with social media
data.
In traditional BI systems it is beforehand crystal clear what will be measured, e.g. the time to assemble a
product from ?ve components, which clearly a?ects the operating expenditures of a company through the cost
of workers. In social business intelligence, the contents in the data are not clear beforehand. As we have seen in
chapter 4, subjects of social media messages di?er from ?rm to ?rm. Thus, not each KPI that a ?rm is willing
to be measured by social media data can actually be measured by these data. It is therefore that the contents
of the social media data determine what can be measured. Each company can be willing to measure KPIs
by social media data, however without existence of any data, this will not be possible. In traditional business
intelligence, a ?rm is much less dependent on external stakeholders for the possibilities of BI.
Reaction based on
social intelligence
Strategic
mapping of
KPIs
Reacting
Search terms
Action plan(s) to
respond to gained
intelligence
g
Collecting
Mapping
insights to
business
units
Data pre
Unstructured
data
Information for
business units
Data pre-
processing
Categorising
Analysing
Structured, combined
(and anonymised) data
Categorised data
Figure 5-10: Speci?c Steps in Social Business Intelligence
When recalling the cycle visualising the social business intelligence procedure developed in section 5-2, the
speci?c steps required in social business intelligence can be highlighted. Figure 5-10 shows the social business
intelligence cycle, in which the speci?c social BI steps are highlighted in orange. In the orange steps, a di?erent
method is required as compared to ‘traditional’ BI. Since our procedure is based on existing business intelligence
methods, the procedure also shows overlap. The steps that are relatively equal to standard BI are coloured
in blue. As can be concluded from ?gure 5-10, di?erent activities are mainly required in the collecting and
processing steps of business intelligence. It is in these steps where the data is converted from unstructured to
structured data that is ready for analysis.
5-6 Sub Conclusion
A social business intelligence procedure (“SBI”) should ?t within the general ‘way of executing’ business
intelligence, since it is not possible that social media metrics measure all key-performance indicators as well
as ?rm’s internal BI systems will do. Therefore, SBI is considered as an additional component to business
88 Blueprint of a Social Business Intelligence Procedure
intelligence, rather than a replacing procedure. However, as the content analysis of chapter 4 illustrated, there
are certain categories of KPIs that are in?uenced by – or at least related to – a substantial amount of social
media messages that are related to the ?rm. For these type of KPIs, which di?er from ?rm to ?rm, a procedure
has been developed that prescribes the necessary steps to acquire, process and ?nally gain intelligence from
?rm related social media messages. The procedure is based on general BI concepts, existing technologic social
media analysis solutions and the experience gained in the execution of a content analysis into the social media
messages related to eighteen di?erent ?rms.
A SBI procedure consists of seven main components, being (i) strategic mapping of KPIs, (ii) collecting, (iii)
data pre-processing, (iv) categorising, (v) analysing, (vi) mapping insights to the business units, and (vii)
reacting. The seven steps can be interpreted as a cycle, i.e. the output of the last step in?uences the ?rst step.
(i) Strategic mapping of KPIs The very ?rst step of social business intelligence sets the scene for the objects
that are to be collected and analysed. Namely, in the ?rst step the key-performance indicators that are to
be measured by social media data are selected. As we have seen in chapter 4, not each type of KPI is to be
measured by social media data since there does simply not exist any related social media data to these types
of KPIs. Firms should mainly focus on KPIs related to customer relations, public image and – to a less extent
– on product and service innovation when selecting KPIs that are to be measured using social media data.
Whenever a ?rm has selected the social KPIs, it can start collecting the appropriate data.
(ii) Collecting The second step of the SBI procedure related to data collection. In contradiction to regular BI
systems, the data is to be sourced from external parties in social business intelligence. People create ?rm-related
messages on di?erent platforms, of which the vast majority of publicly accessible messages are created on Twitter.
The search terms that are used to ?lter out the content at which the ?rm is interested should be based on the
social KPIs selected in the previous step.
(iii) Data pre-processing The social media data has been collected from multiple platforms which adhere to
their own data format. The di?erent format are to be combined into one uniform database, so that – in a
later step – data analysis can be applied on the complete dataset. Furthermore, the ?rm should select those
attributes that are necessary for the analysis, not each platform o?ers the same richness of attributes to a social
media post. In addition, the data should be anonymised to be in compliance with new Regulations regarding
data privacy. Finally, spam – i.e. social media posts that do not relate to the ?rm – should be removed from
the collected data.
(iv) Categorising The data pre-processing step resulted in a structured database in which the social media
messages from multiple platforms are combined. In the categorising step, the messages are clustered on di?erent
issues of interest, depending on the ?rm’s subject of interest. E.g., messages related to certain products can be
categorised, or one can cluster the messages that are created by people with many followers, etc. Again, the
criteria at which the messages are categorised are determined by the selection of the social KPIs in the ?rst
step.
(v) Analysing So far, the collected data has not provided any insights. It is in this step of the social business
intelligence procedure where data is transformed into information. The categories that were established in the
previous step are analysed in this step. For instance, sentiment analysis can be applied on the categories related
to the ?rm’s products in order to acquire intelligence related to customer experiences of the products. However,
the most valuable intelligence is gained when social media data is related to internal data. For instance, the
volume of social media messages related to a certain product may be correlated with the sales volume of that
product. It is in this phase of the SBI procedure where such relations are explored.
(vi) Mapping insights to business units In the ?rst step of the procedure, KPIs have been selected. These
KPIs typically relate to a certain function of the ?rm, and hence have an ‘owner’. The intelligence gained in the
previous step relates to KPIs, and should feed back to the owner of the KPI. Generally, it are the people in the
?rm that are responsible for the KPI who are the ones that can reason how the KPI is in?uenced. Therefore,
these people are the ones that can draft an action plan in case the KPI needs improvement.
5-6 Sub Conclusion 89
(vii) Reacting The ?nal step of the social intelligence procedures consists of the execution of the action plans
that are developed in collaboration with people from the business lines that are responsible for the respective
KPIs. Actions on the gained intelligence may involve revisions of internal processes or strategies, or external
interventions such as social media engagement.
The developed social business intelligence procedure is based on general business intelligence processes, and the
requirements of the social BI procedure have been derived from general BI processes. For the veri?cation of the
SBI procedure, all requirements have been checked on ful?lment by systematically tracking which requirement
is ful?lled by which activity.
Chapter 6
Conclusions & Discussion
First of all, this chapter presents the conclusions of the research in section 6-1. In section 6-2 the contributions
of this work to existing and future research are discussed. Section 6-3 proceeds by discussing the implications of
the ?ndings in this thesis for practice. Section 6-4 consequently re?ects on the thesis and the research process.
In section 6-5 the research is critically reviewed, and limitations are discussed. Finally, section 6-6 provides
suggestions for future research related to the subject of this thesis; social business intelligence.
6-1 Conclusions
Firms are increasingly using social media, while at the same time business intelligence systems are increasingly
applied for performance measurement of business activities. Though these two concepts o?er room for synthesis,
it also raises questions related to the applicability and opportunities o?ered by combining social media and
business intelligence. So far, it is not clear which ?rms are able to ?nd ?rm-related social media data and if
they are able, how these data should be incorporated in the business intelligence processes of ?rms. As one of
the ?rst researches into the opportunities of leveraging social media data for BI purposes, this this was purposed
to draw generic conclusions on the applicability of social business intelligence by distinguishing ?rms on generic
aspects. Firms were distinguished on customer relation type – either B2C or B2B – and on industry type.
Therefore, the main research question of this thesis has been formulated as:
How can ?rms use social media data for business intelligence, taking into account the ?rm’s speci?c
industry and relationship with end-users?
The main research question has been divided into three sub questions, which are answered in the following
sections. The ?rst sub question was de?ned as:
1. What is the current state of social media in relation with business intelligence?
Social media is a natural consequence of Web 2.0, and can be de?ned as Web 2.0 based applications allowing
users to create and share user-generated content with pre-selected users and/or communities. The applications
through which users are active are known as social media platforms. In 2012, there are many social media
platforms available, which di?er in scope and functionality. Each platform adheres to its own policy regarding
data crawling and data format.
Social media is a topic on the agenda of many ?rms in 2012. Though many ?rms acknowledge the opportunities
of social media, there also exists a degree of reluctance from managers towards social media. Research indicates
that executives who avoid social media do not understand what social media is, how to engage with it and learn
from it. On the other hand, ?rms that do embrace the world of social media particularly perform activities in
the ?eld of marketing, customer relations management, reputation management and co-creation / prosuming
activities through the various social media platforms.
Business intelligence – irrespective of the variables to be measured – can be perceived as a cycle consisting
of three main steps; (i) register, (ii) process and (iii) react. Before a BI cycle can commence, it has to be
6-1 Conclusions 91
determined ‘what to measure’. The variables that are to be registered are generally aligned with a ?rm’s
strategy and corresponding business model, termed key-performance indicators (“KPIs”). The three BI steps
are required when a ?rm intends to apply business intelligence on the ?rm’s social media activities. However,
social media data di?ers from “regular” business information. Unlike internal business data, social media data
is created by non-professionals and stored into a variety of databases that are owned by external parties who
employ their own database structure and access limitations. Therefore, a di?erent BI approach is required for
social media data.
Firms employ di?erent key-performance indicators. Especially lower level KPIs are ?rm speci?c, while top level
KPIs are generic and employed by many ?rms. Based on Adam & Neely’s (2001) generic performance prism
perspectives, ten categories of key-performance indicators have been established. The ten categories are de?ned
as (short-term) ?nancial results, customer relations, employee relations, operational performance, product and
service quality, alliances, supplier relations, environmental performance, product and service innovation and
community. It are these categories of KPIs for which related social media data has been searched for.
In social business intelligence, a ?rm analyses the activities on social media related to the ?rm and determines
the e?ect of these activities on the ?rm’s performance. Existing social media monitoring tools – which are
becoming increasingly available on the market – mainly reveal the performance of the ?rm on social media as
a separate component of the ?rm. The intelligence that such monitoring tools provide relate to the volume of
posts, engagement of users, sentiment, geography, topics and themes in the social media messages, in?uencer
ranking, channel distribution, etc. However, the purpose of business intelligence is to reveal the underlying
parameters that determine the ?rm’s performance, that is, not limited to solely social media performance.
In order to understand the in?uence of social media activities on the ?rm’s performance, a link between the
company’s key-performance indicators and social media parameters is required. In social business intelligence,
such links are required.
The second sub question was formulated as follows:
2. In which contexts are ?rms able to acquire social media data for business intelligence?
In this research, the context of a ?rm has been described based on two generic dimensions. Firstly, ?rms
were distinguished from each other based on the industry in which they operate. Secondly, ?rms’ contexts
were described by distinguishing di?erent customer relations types; i.e. B2B or B2C relations. The volume of
messages that contain the name of a ?rm di?ers from ?rm to ?rm. E.g. in our sample 39.425 messages related
to Heineken have been collected, while during the same period only 428 messages related to Fugro have been
found. Our analysis indicates that there exists variation in the volume of ?rm-related social media posts across
di?erent industries. Firms classi?ed as industrials, information & communication were more frequently subject
of discussion on social media than consulting or mining & quarrying ?rms. Our analysis also illustrates that
there exists variation in the volume of social media posts across B2B and B2C ?rms. B2C ?rms are far more
often subject of discussion on social media than B2B ?rms.
Apart from an assessment of the volume of social media content related to ?rms, we analysed the subjects of the
messages in order to gain an understanding of the type of information contained in the social media messages.
Our analysis shows that the subjects of social media messages di?er from ?rm to ?rm. The majority of social
media messages related to ?rms (41%) express how the external stakeholders of a ?rm perceive the company. In
this thesis, such posts have been classi?ed as community posts. 18% of the social media messages in our dataset
contained the name of a ?rm, but did not contain any valuable information for the ?rm and have consequently
been assigned as unde?ned posts. About 11% of the social media messages relate to ?nancial results, which
consist of ?nancial performance discussions (5%) and stock related discussions (6%).
The content analysis of this research suggests that the subjects of social media messages related to B2B
?rms contain a higher percentage of short term ?nancial results, news and professionals related messages than
messages related to B2C ?rms. Unfortunately for B2B ?rms, such type of information is yet available internally.
Acquiring social media data to gain additional management information is therefore of less value for B2B ?rms.
Next, the analysis indicates that the social media messages related to B2C ?rms contain a higher percentage
of posts related to customer relations, product and service quality and product and service innovation than
messages related to B2B ?rms. It are these types of information that deliver additional value to the ?rm, since
this information is not available at ?rms internally.
In addition, the content analysis of this research suggests that the subjects of social media posts di?er between
industries, but that the majority of the subjects in each industry relates to community, i.e. social media posts
92 Conclusions & Discussion
revealing how the community perceives the company. The results indicate that ?rms active in the information
& communication, ?nancial institutions and transport & storage industries are more subjected to social media
messages related to customer relations, while ?rms active in the mining and quarrying and consulting industries
will ?nd messages related to ?nancial performance.
As indicated in chapter 1, this thesis describes a ?rm’s context based on two dimensions; customer relation type
and industry type. By distinguishing ?rms based on customer relation type, we can state that B2C ?rms are
able to acquire (i) a high volume of social media messages related to their ?rm and (ii) social media messages
that contain information that is not yet available in the internal information systems of the ?rm, hence enriching
the business intelligence. Next, when distinguishing ?rms based on their industry we conclude that the volume
of ?rm related messages di?ers between ?rms. Additionally, the analyses suggest that the subjects of the social
media messages di?er from industry to industry. However, in our sample, there exists interaction between
the customer relation type and the industries. The di?erences in volume and subjects are more visible when
distinguishing between B2B and B2C ?rms rather than distinguishing between industries. Table 6-1 summarises
these conclusions.
Table 6-1: Conclusion of Content Analysis
Volume of Social Media Messages Subjects of Social Media Messages
Customer
Relation
B2C ?rms are more often subject of
discussion on social media than B2B
?rms.
B2C ?rms related social media messages are more
often subjected to customer relations, product and
service quality and product and service innovation
than B2B related ?rms. On the other hand, B2B
?rms’ related messages are more often subjected to
?nancial results, news and professionals discussing
the ?rm. However, the information contained in the
social media posts of B2B ?rms is often yet available
to the ?rm, hence not o?ering added value to the
?rm’s richness of management information.
Industry Our analysis shows a variation in
the volume of social media messages
across di?erent industries.
Our analysis indicates that there is a di?erence in
the subjects of social media posts related to ?rms in
di?erent industries.
The third sub question of this thesis has been formulated as:
3. Which processes are required to incorporate social media data into general business intelligence
frameworks?
‘Social’ data di?ers from internally generated and collected data on the following aspects. Firstly, social data
has not been veri?ed before it is published. Anyone can create a social media message, it will not be veri?ed
before it is available to the world. Social media messages may contain jokes or cynicism, making it hard for
?rms to interpret what the writer of the message actually means. Secondly, social media data is unstructured.
The data are sourced from multiple sources in di?erent formats and languages. Each source may employ its
own structure of social media messages, and not each platform may contain the same richness in attributes as
the other. Thirdly, the unstructured nature of the data and the huge amount of data that is generated makes
it that social media data can be labelled as ‘big data’, implying that the problems of big data may also be
applicable on social media data. Fourth, as with many data on the web social media messages contain spam,
which is to be removed before commencing an analysis of the messages. Fifth, while internal data may represent
evident relations (e.g. between the number of employees and the ?rm’s revenues), relations between social media
metrics and key-performance indicators are less evident. When a ?rm intends to add a social component to its
existing business intelligence, these aspects should be considered.
Structuring social media data is an important activity required to make an analysis on such data. Firstly, the
sourced data has to adopt one and the same data structure. Whereas each social media platform will deliver data
to its own favour – e.g. by CSV, JSON, XML, or other formats – the data should be parsed into one common
data format. Secondly, dividing the messages into categories makes the dataset ready for interpretation. Many
types of categories are possible. A classi?cation based on the people reveals which users are actively engaged
with the ?rm, which users have much power (in terms of followers and friends), which users speak positive
6-2 Contributions to Research 93
about the ?rm, etc. A classi?cation based on the subjects of the social media messages reveals which topics are
considered important by the social media users, and, more important a classi?cation based on subjects allows
a ?rm to link social media messages to the ?rms’ key-performance indicators. For instance, public image posts
– which may be additionally classi?ed as positive, neutral or negative – are related to a customer satisfaction
KPI. Though there are plenty of other categories to classify social media messages, to link the ?rm’s KPIs to
social media data the messages should be categorised based on their subjects.
Reaction based on
social intelligence
Strategic
mapping of
KPIs
Reacting
Search terms
Action plan(s) to
respond to gained
intelligence
g
Collecting
Mapping
insights to
business
units
Data pre
Unstructured
data
Information for
business units
Data pre-
processing
Categorising
Analysing
Structured, combined
(and anonymised) data
Categorised data
Figure 6-1: Blueprint: Social Business Intelligence Procedure
A social business intelligence (“SBI”) procedure prescribes how a ?rm should collect and process social media
data to gain intelligence, at which the ?rm can consequently base their decision-making. A SBI procedure
consists of seven main components, being (i) strategic mapping of KPIs, (ii) collecting, (iii) data pre-processing,
(iv) categorising, (v) analysing, (vi) mapping insights to the business units, and (vii) reacting. The seven steps
can be interpreted as a cycle, i.e. the output of the last step in?uences the ?rst step. Figure 6-1 schematically
shows the social business intelligence procedure.
At the start of this thesis, the objective has been formulated as:
The objective of this research is to develop a procedure to utilise social media data for business
intelligence, for which the applicability is investigated for ?rms in di?erent industries and for
di?erent relations with end-users.
Taking into account the answers on the research questions, we state that the objective of this thesis has
been achieved. A social business procedure has been developed, veri?ed on consistency with general business
intelligence processes and tailored to the challenges arising from processing social media data. In addition, the
applicability of this procedure is investigated. The results of our study indicate that especially ?rms performing
B2C relations are able to execute social business intelligence, because (i) these ?rms are subject of discussion
on social media, hence ?rm-related social media exists for these ?rms and (ii) the information contained in
B2C related messages o?er additional information for the ?rm. Furthermore, the results of this study also
indicate that there exists a di?erence in the volume of ?rm-related content between di?erent industries, in
which industrials and information & communication ?rms are more frequent subject of discussion on social
media than consulting and mining & quarrying ?rms.
6-2 Contributions to Research
Social media is a hot topic in the academic world. Existing research in the ?eld of social media is often aimed at
marketing e?orts or other activities in which the ?rm expresses or should express itself to the outside world. This
94 Conclusions & Discussion
thesis focused on the incoming information from a ?rm point of view, i.e. the extraction of information from
social media to support decision-making. The fact that this thesis investigated the applicability of social media
data for organisational decision-making, makes it that this thesis touches the world of business intelligence. In
our opinion, this aspect distinguishes this thesis from other research.
Future research aimed at deriving information – in whatever form – from social media for business purposes,
should be aware that the type of ?rm a?ects the applicability of such activities, and that one should not draw
generic conclusions applicable to all ?rms. As this research indicates, the existence of and subjects contained
in ?rm-related social media messages di?ers between ?rms. Furthermore, the method of this research is partly
based on traditional content analysis, in which a new type of data has been analysed. In the following section,
our experiences of applying a content analysing on social media data are shared.
6-2-1 Methodological Innovation
The content analysis methodology is not new. Krippendor? (2004) refers to propaganda analysis during World
War I as one of the earlies structured approaches of analysing texts. However, the fact that this thesis performed
a content analysis on social media messages makes it that part of the research in this thesis is innovative. The
structured approach that was executed was based on earlier work from Bos and Tarnai (1999), whom created
a framework to execute a content analysis. Though their framework did not speak of social media platforms
or data, we found that their research framework – with adjustments – is also applicable on social media data.
The adjustments of – or additions to – the framework relate to the data collection and preparation steps of
the content analysis procedure. Whereas textual data is considered structured in the framework of Bos and
Tarnai (1999), a content analysis on social media data requires a data structuring process before commencing
the categorisation process. The strength of the framework lies in the fact that it is generic, hence applicable
in many domains. With the experience of the execution of the content analysis on social media data in this
thesis, we recommend using the framework of Bos and Tarnai (1999) for future social media content analyses.
Figure 6-2 shows the additional steps required when applying a content analysis on social media. These steps
are highlighted in blue, and are established based on our experiences in the execution of the content analysis
on social media data.
Research outline, research questions,
formulation of hypotheses, material to
investigate
Operationalising the categories,
determining the sample, determining the
unit of analysis
Establishment of categories
Theoretical level
Determining reliability and validating the
categories
Appropriate statistical analyses
Data collection and evaluation
Immanent interpretation of the results,
discussion of the results on the basis of
the problem
Interpretation of the results
Select platforms, select keywords,
determine measurement period, determine
attributes to scrape
Social media domain
Create uniformdatabase, create uniform
data structure, merge data from multiple
platforms, remove spam
Pretest
Data preparation
Figure 6-2: Adjusted (Social Media) Content Analysis Procedure
The additional steps required when applying a content analysis take place after the second step. After the
research questions, the formulation of hypotheses, the determination of the categories that are to be analysed
and the sample establishment, the social media domain comes in. In this step, the researcher needs to determine
from which platforms the data is to be sourced. Each platform adheres to its own data format, and not each
platform provides access to all posts. Furthermore, each platform has its own focus, implying that di?erent
people are active on di?erent platform. Next, the keywords are to be determined. Comparable to search engines,
social media scrapers scan for keywords in the many posts created on the web. A researcher may decide to
6-3 Implications for Practice 95
base its keywords on user names, hashtags (subject of message, assigned by the creator of the message), the
researcher may decide to scrape all posts in a certain area, or during a time. Other elements to select the
messages that are to be analysed are also possible. Next, the attributes to scrape are to be determined. Each
social media post exists of various attributes, e.g. user name, time, location, content, hyperlink, etc. It di?ers
per platform which attributes are shared.
Next, the data is to be prepared. In case that the research exists of data sourced from multiple platforms,
the data is to be merged and structured into a uniform database. In addition, the data is likely to contain
spam. These messages are to be removed before commencing the analysis. The framework then proceeds in the
original steps of Bos and Tarnai (1999).
6-3 Implications for Practice
The ?ndings in this research have implications for (consulting) ?rms willing to use social media data for business
intelligence. First, the ?ndings of this research indicate that B2B ?rms are less likely to ?nd social media data.
Furthermore, if a B2B ?rm will ?nd messages related to the ?rm, these messages are likely to contain information
that is yet available to the ?rm. Hence, the applicability of social business intelligence for B2B ?rms is limited.
On the contrary, B2C ?rms are often subject of discussion, and the messages related to B2C ?rms contain
information that is not yet available to the ?rm internally. Taken into account the results of this research, the
opportunities and promises of social media found in many reports and white papers are mainly applicable to
B2C ?rms. Therefore, ?rms and ?rms o?ering consultancy on the domain of social business intelligence, should
be aware that the opportunities of social business intelligence are limited to B2C ?rms.
In addition, a stepwise procedure for social business intelligence has been developed. Such a procedure was
necessary to be developed since the new data source – social media platforms – di?ers from the systems at
which normally data is collected and stored. This procedure is applicable on ?rms for which social media data
is available.
6-4 Re?ection
In this section, a re?ection on the research process is presented. First, the developed Twitter scraper is discussed.
Next, additional research steps that were executed whenever there was more time available are presented.
Finally, a detailed stepwise approach of our data collection process is presented.
6-4-1 Twitter Scraper
During the early stages of this Master thesis project, a software tool has been developed that scrapes messages
created on Twitter. The tool has been written in PHP language and is designed to work with MySQL databases,
which we managed using phpMyAdmin. At the same time that the tool was up and running, access to one of
the commercial social media monitoring tools (uberVU) was granted to the author of this thesis, for which we
are grateful. Clients of uberVU pay a monthly fee of at least $1.000 to access the software. The tool allowed us
to scan a variety of social media platforms, whereas our own tool solely scraped Twitter. Therefore, the decision
was made to use the commercial o?-the-shelf software rather than our own to collect the data. In addition,
uberVU o?ers additional features like sentiment analysis, in?uencer ranking, location of the message, etc. that
are not available in our tool. However, our own scraper – though it solely scrapes tweets – may serve research
projects that are aimed at tweets.
6-4-2 If I had More Time
This research has been executed during a period of six months, i.e. from July 2012 to December 2012. The
limited time available for this research has implications for both the depth and the breadth of the research, hence
on the conclusions and the applicability of the conclusions. In case that the research time would be longer, we
would have surveyed more ?rms so that the analysed categories (B2B/B2C and industries) would have consisted
of more respondents, allowing for statistical analyses. Whereas the conclusions of this research are exploratory,
the statistical testing of di?erences between groups of ?rms would allow for the generalisation of the statements.
Furthermore, adding more respondents to the sample leads to more industries being represented in the sample,
96 Conclusions & Discussion
so that the conclusions of the research are also applicable on other industries. Next, we would have analysed
more messages in the content analysis. The social media messages in this research have been collected during
a period of two weeks. In case the messages would have been collected during a longer period of time, e.g. six
months, the sample would have existed of more messages. As such, it would be possible to gain insight in the
‘steady volume’ of daily messages, and, more interestingly, deviations in the steady volume. Deviations might
be due to the announcement of ?nancial ?gures, marketing events, etc. However, to perform a content analysis
on many social media messages, an automatic classi?er is required. Manually classifying the posts, as we did
in this research, would then take too much time. To train such an automatic classi?er, the manually classi?ed
posts in this research are suited. Furthermore, if we had more time, we would have investigated whether or not
other factors, such as the number of employees, revenues, market capitalisation, etc. also a?ect the number of
social media messages that are related to a ?rm. With respect to the social business intelligence procedure that
has been developed, we would have validated the framework by pilot projects and the involvement of business
intelligence experts.
6-4-3 Stepwise Description of Data Collection Process
In this section we share our method of the data collection process. These steps are also incorporated in the
adapted content analysis framework in section 6-2-1.
• Determine keywords / search terms
After establishing the sample, the search terms required to ?lter out the related social media messages
have been determined. In this thesis we searched for the ?rm names. However, it is also possible to search
more speci?c, e.g. on the name of a product or a speci?c event.
• Use search terms in social media monitoring tool
The search terms were consequently used in the social media monitoring tool.
• Export the search results into a database
This functionality is not o?ered by each social media monitoring tool, but vital for the data collection
process. Tools that do not o?er the exportation of search results in whatever format are not suited for
further analyses on the data because most analysis software requires the data to be stored on a local
machine.
• Structure the database
Depending on the output of the export process, the database is to be structured. The social media
monitoring tools used in this thesis exported the messages into comma-separated values, which could
easily be loaded into MS Excel. The richness of attributes o?ered by the social media monitoring tools
determines the complexity of the structuring of the database.
• Daily search for new results
In order to collect a large dataset, daily runs for new messages were executed. The social media monitoring
tools used in this thesis o?ered to possibility to export up to 10,000 messages per search run. Pre-testing the
collection process illustrated that this constraint was su?cient to get a complete picture of the ?rm-related
messages created on the social media platforms by daily searching for new results. Whenever a tool has
been used that o?ered a lower exporting capacity, e.g. 1,000 messages, the frequency of searching for new
results would have been higher.
• Verify that the new search results do not yet exist in the database
A daily run for ?rm-related messages resulted in the collection of duplicates, i.e. messages that were
already collected yesterday. These messages have been identi?ed based on their unique URL that was
contained as an attribute to each message using LOOKUP functions in MS Excel. More speci?cally, the
value of each URL was LOOKED UP in the existing spreadsheet. The textual format of these lookup
values required some computer power, but our 4GB RAM / i5 machine turned out to have su?cient power
for these calculations. Whenever a message did found a match, this meant that the message did yet exist.
The messages that did yet exist were not added to the database.
• Start analyses
The data collection process resulted in a structured database in a format so that MS Excel could handle
the data for analyses.
Whereas the above steps are described in detail, the underlying ideas will be applicable on each social media
data collection process.
6-5 Limitations 97
6-5 Limitations
This research and consequently its outcomes have limitations that should be taken into consideration when
adopting the conclusions of this thesis. The limitation are discussed in this section.
The ?rst aspect of the research limitations, or aspects to consider when interpreting the conclusions related to
the population from which the social media messages are drawn. This aspect has been assigned by Krippendor?
(2004) as an important issue to consider when performing a content analysis. Not everyone uses social media,
and even less people actually create content on the platforms. Therefore, ?rms that analyse social media data
should be aware that these data do not represent the full (potential) client base. It is very likely that social
media users have other preferences than non-social media users. A ?rm should always place the conclusions
from social business intelligence in the light of their complete client base before it makes a decision to undertake
actions, because the actions may only serve those needs of the ones that engaged on social media. Nevertheless,
anno 2012 social media is relatively young. The user groups – e.g. age groups or countries – that use social
media may increase in the coming years. We expect that social media will be further embedded in the lives of
people that grow up in the social media era.
Secondly, each social media platform has its own privacy policy. This implies that a user either has the possibility
to determine whether or not it shares its messages to the public, or that the platform determines the publicly
available messages. As a consequence, only messages that were publicly available have been analysed in our
dataset. It is likely that people who have not publicised their social media messages also discuss ?rms or ?rms’
products / services, these messages are not available in our dataset. Still, the dataset is representative for ?rms
conducting social business intelligence, since they will not get access to private messages either.
Third, not all social media platforms have been part of our analysis. Platforms such as Sina Weibo – the Chinese
counterpart of Twitter –, Qzone, Renren, Habbo are not part of our analysis. However, the platforms that did
exist in our sample are the ones used in the Western world. Therefore we state that the conclusions of our
report are valid for Western world ?rms.
Fourth, the volume of messages, likes, shares, retweets, etc. can easily be in?uenced by a ?rm, though this does
not necessarily mean that the user is actually engaged with the ?rm. For instance, a ?rm may decide to ra?e an
iPad or organise other lotteries. People can participate in the lottery by e.g. sharing a promotion message or by
‘liking’ the ?rm’s page. Whereas such activities certainly lead to an increase in the number of likes, shares, etc.,
the underlying reason why people pay attention to the ?rm is for the price, and not necessarily the engagement
in the company or its products. We refer to content that is generated according to such mechanisms as biased
chatter, and doubt if such activities actually lead to an improvement of the ?rm’s KPIs, e.g. the number of
sales.
Fifth, one of the conclusions of this research is that the volume and the subjects of social media messages di?er
among industries. Due to time restrictions, we have not been able to analyse the social media messages of more
than eighteen ?rms. As a result, each industry consisted of two or three ?rms, which we deem a small sample.
It is therefore that the conclusions of this research are to be perceived as exploratory rather than con?rming
hypotheses.
Sixth, our research grouped ?rms based on the industry type in which they are active. With respect to the
volume of ?rm-related messages, intra group di?erence have been spotted. These notable ?ndings indicate that
the industry aspect is not the only determining factor in?uencing the volume of ?rm-related messages. Other
factors, such as (world-wide) brand awareness or the size of the company are likely to in?uence the amount
of ?rm-related messages that are daily generated. These company speci?c aspects have deliberately not been
taken into account since the purpose of this thesis is to draw generic conclusions on the applicability of social
business intelligence.
Next, this thesis analysed the applicability of social business intelligence on two dimensions; industry type and
customer relation type. We did not correct for interaction e?ects between these two groups. It is e.g. likely
that there exist more B2B ?rms in the consulting industry. However, the results of this study are exploratory
and – from the insights we gained – future research containing larger samples should correct for such interaction
e?ects.
Next, this thesis mainly focused on the business perspective of social business intelligence. Less attention has
been paid to the technical perspective. For example, the question “What kind of database is best to store
the unstructured data that is captured in text form?” is unanswered. Though the developed social business
procedure prescribes which components are required to collect and analyse social media data in relation with
the ?rm’s performance, the technical requirements related to these components are underexposed.
98 Conclusions & Discussion
Finally, the social business intelligence procedure that has been developed in this thesis is compliant with general
business intelligence concepts that are adhered to in ?rms. However, the procedure has – due to time restrictions
– not been validated, that is, tested on a real case. Nevertheless, the individual components of the procedure
are tested. The general BI steps are yet used by ?rms, and the collection and classi?cation processes of social
media messages have been performed in this thesis.
6-6 Future Research
During the execution of this research, ideas for future research related to this thesis have been devised. These
ideas are presented in this section. The suggested researches build further on the conclusions of this thesis.
6-6-1 Classi?er
We have manually classi?ed social media posts in categories. With these manually classi?ed posts, it is possible
to create an automatic classi?cation process. In automatic classifying, a classi?er will be “trained” so that
it recognises which words and phrases relate to a certain category. The classi?ed messages in this research
can serve as the training set for an automatic classi?er. As we have experienced, and which is also argued
by Gianfortoni, Adamson, and Rosé (2011), classi?cation of social media posts, e.g. by gender, age, political
a?liation and sentiment analysis is di?cult, and even more problems arise when models trained in one domain
are applied in another domain. Therefore, a social media classi?er should not be used generally on each domain.
We even argue that each ?rm requires its own classi?er, only because the product names of ?rms di?er.
In the development of the social business intelligence procedure we argued that the categorising process is one
of the important steps in structurally analysing social media messages. This process is even more challenged
by the increase of user-generated social media content showing big data characteristics. From a social business
intelligence view, it is desired that research in the ?eld of automatic text classifying – tailored to ?rms – proceeds.
6-6-2 Social Media Posts Categories
A part of this research required the establishment of social media posts categories. Whereas the starting point of
the establishment of the categories in this thesis was based on former research, the social media messages in the
dataset have driven the establishment of additional categories. Future research in the domain of social media,
and more speci?cally the classi?cation of social media messages can use the categories that were established in
this thesis.
6-6-3 The Real Source
Many messages are forwarded – retweetet and shared – from users to others. Thereby, messages do not stay
within one social media platform. It is interesting to investigate which platforms contains the most initial
creations of information. As such, ?rms can manage their reputation by actively following those platforms that
create the most initial messages, before the message goes viral and may harm the ?rm’s reputation.
6-6-4 Case Study: Relations of Social Media Metrics and Key-Performance Indicators
The social business intelligence procedure that has been developed in this thesis contains a component in which
social media messages are assigned to key-performance indicators based on the subject of the messages. A
research in which the relations between the social media messages and the actual values of various KPIs of a
?rm are investigated would reveal the strength of these relations. Thereby, the KPIs that have been assigned
in this thesis as being able to be measured using social media data could be used for such an analysis.
Appendix A
Performance Prism Perspectives and
Key-Performance Indicators Categories
Table A-1 on the next page assigns each KPI category de?ned by Ittner et al. (2003) to a performance prism
perspective de?ned by Neely et al. (2001). Based on their subjects, social media posts will be assigned to KPI
categories in this thesis. With the assignment of KPI categories to performance prism perspectives, we can
derive conclusions of the existence of social media data related to di?erent performance prism perspectives.
100 Performance Prism Perspectives and Key-Performance Indicators Categories
Table A-1: Assigning Key-Performance Indicator Categories to Performance Prism Perspectives
KPI Category Performance Prism
Perspective
Elucidation
Short-term ?nancial results Stakeholder satisfaction Shareholders are the actors that are interested in the
return on their investment.
Customer relations Stakeholder satisfaction
/ contribution
On the one hand, KPIs related to customer relations
can represent the customer satisfaction. On the
other hand, customers may also contribute to the
?rm, e.g. by payments and/or co-creation activities.
Employee relations Stakeholder satisfaction
/ contribution
Employees are the actors interested in getting
awarded for their contributing value to the ?rm.
Therefore, KPIs related to customer relations can
involve both perspectives.
Operational performance Processes KPIs related to operational performance re?ect the
performance of business processes, generally in time
of volume, speed, reliability, etc.
Product & service quality Capabilities KPIs related to product and service quality represent
how capable a ?rm is in performing its activities.
Alliances Stakeholder satisfaction
/ contribution
Metrics related to the ?rm’s alliances are on the
one hand purposed to satisfy the participating
parties and on the other purposed to measure the
contributing value of the alliance to the ?rm.
Supplier relations Stakeholder satisfaction
/ contribution
KPIs related to supplier relations are purposed to
measure the contributing value of the suppliers’
products/services to the ?rm, or the metrics can
specify the satisfaction of the customers (e.g. in
terms of the price paid for the product).
Environmental performance Stakeholder satisfaction A ?rm’s activities may a?ect the environment.
Groups representing the environment may not
be satis?ed whenever the ?rm’s activities a?ect
the environment. KPIs re?ecting environmental
performance hence relate to stakeholder satisfaction.
Product & service innovation Processes One of the key business process relates to the
development of new products and services.
Community Stakeholder satisfaction
/ contribution
Metrics related to the ?rm’s community relate to the
public image of the company. As such, these metrics
involve stakeholders.
Appendix B
Classi?cation of Social Media Posts
In section 4-3-1, categories for social media posts have been established. Consequently, the collected social media
posts of the ?rms in the sample have been classi?ed into one of these categories. The results are discussed in
this chapter. The results will be discussed from ?rm to ?rm. However, we will refer to ?gure B-1 – which is
depicted below – when discussion the individual ?rms. This ?gure contains a heat-map of the KPI categories
per ?rm, indicating which category is presented the most (green) and which the least (red) in the sample. The
percentages that are contained in the heat-map are visualised in a stacked bar chart in ?gure B-2.
KPI Category
A
B
N
A
M
R
O
A
e
g
o
n
A
k
z
o
N
o
b
e
l
A
l
b
e
r
t
H
e
i
j
n
A
r
c
a
d
i
s
A
r
c
e
l
o
r
M
i
t
t
a
l
B
l
o
k
k
e
r
B
o
l
.
c
o
m
C
-
1
0
0
0
C
o
c
a
-
C
o
l
a
F
u
g
r
o
H
e
i
n
e
k
e
n
K
L
M
N
S
P
h
i
l
i
p
s
P
o
s
t
N
L
T
o
m
T
o
m
U
n
i
b
a
i
l
-
R
o
d
a
m
c
o
A
v
e
r
a
g
e
1. Short-term financial results 1% 5% 21% 0% 37% 16% 0% 0% 0% 0% 36% 3% 1% 0% 1% 2% 3% 75% 11%
1.1 Financial performance discussions 0% 0% 1% 0% 32% 8% 0% 0% 0% 0% 0% 0% 1% 0% 1% 0% 0% 47% 5%
1.2 Stock related discussions 1% 5% 20% 0% 5% 8% 0% 0% 0% 0% 36% 3% 0% 0% 0% 2% 3% 28% 6%
2. Customer relations 17% 4% 0% 20% 2% 0% 7% 17% 1% 4% 0% 2% 14% 33% 0% 39% 6% 0% 9%
2.1 Explaining firm 4% 0% 0% 5% 0% 0% 0% 3% 0% 1% 0% 1% 4% 4% 0% 2% 2% 0% 1%
2.2 Understanding firm 2% 0% 0% 4% 0% 0% 0% 4% 0% 1% 0% 0% 7% 4% 0% 1% 2% 0% 1%
2.3 Thanking firm 1% 0% 0% 1% 1% 0% 0% 1% 0% 1% 0% 0% 1% 1% 0% 0% 1% 0% 0%
2.4 Informing firm 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 11% 0% 0% 0% 0% 1%
2.5 Questioning customer 3% 1% 0% 3% 0% 0% 5% 4% 0% 1% 0% 1% 1% 4% 0% 5% 1% 0% 2%
2.6 Complaining customer 4% 2% 0% 6% 0% 0% 3% 4% 1% 0% 0% 1% 1% 9% 0% 30% 0% 0% 3%
2.7 Thanking customer 2% 1% 0% 2% 0% 0% 0% 1% 0% 0% 0% 0% 1% 1% 0% 1% 0% 0% 0%
3. Employee relations 1% 4% 10% 5% 5% 2% 8% 0% 8% 1% 10% 0% 0% 3% 2% 21% 0% 1% 4%
3.1 Recruitment 1% 2% 9% 1% 4% 2% 1% 0% 0% 1% 8% 0% 0% 3% 2% 7% 0% 1% 2%
3.2 Employee posts 1% 2% 1% 4% 1% 0% 7% 0% 8% 0% 1% 0% 0% 0% 0% 14% 0% 0% 2%
4. Operational performance 1% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 1% 0% 0% 0% 0% 0%
5. Product and service quality 1% 0% 0% 1% 0% 0% 1% 0% 0% 2% 0% 1% 0% 1% 7% 1% 7% 0% 1%
6. Alliances 0% 2% 0% 0% 0% 0% 0% 1% 0% 0% 0% 0% 0% 0% 0% 3% 0% 0% 0%
7. Supplier relations 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 1% 0% 0% 0% 0% 0% 0% 0%
8. Environmental performance 0% 0% 0% 2% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
9 Product and service innovation 2% 4% 1% 1% 0% 0% 0% 0% 0% 1% 0% 1% 0% 1% 1% 0% 8% 0% 1%
10 Community 71% 43% 55% 28% 44% 60% 27% 72% 25% 57% 49% 44% 7% 30% 61% 23% 26% 15% 41%
10.1 Promotion 56% 2% 1% 1% 8% 2% 0% 20% 2% 2% 0% 3% 1% 2% 1% 0% 2% 1% 6%
10.2 News 3% 4% 21% 2% 5% 6% 0% 0% 0% 1% 1% 4% 0% 1% 0% 1% 0% 1% 3%
10.3 Public image 10% 36% 11% 26% 18% 32% 26% 50% 23% 50% 3% 33% 4% 27% 6% 20% 22% 3% 22%
10.4 Professionals 3% 1% 3% 0% 12% 20% 0% 2% 0% 2% 45% 3% 2% 0% 0% 2% 0% 11% 6%
10.5 Distributors 0% 0% 18% 0% 0% 0% 1% 0% 0% 1% 0% 2% 0% 0% 54% 0% 2% 0% 4%
Undefined 4% 16% 10% 43% 11% 18% 37% 4% 50% 23% 4% 28% 2% 2% 1% 11% 49% 9% 18%
Spam 0% 22% 2% 0% 1% 3% 19% 6% 15% 13% 1% 18% 75% 28% 27% 0% 2% 0% 13%
Classified Posts 7.067 1.449 922 2.848 455 1.097 1.788 2.574 1.651 1.078 428 2.050 2.498 1.441 1.623 1.013 1.100 512
Figure B-1: Social Media Posts Classi?cation
ABN AMRO
ABN AMRO Group N.V. is a Dutch bank with 6.8 million clients and around 25.000 employees. The ?rm
organises multiple marketing events each year, of which the pictures were uploaded to the ?rm’s Picasa pro?le
during our sample period. This declares the high percentage of social media posts made on this platform, as
can be seen in table C-1. If we would neglect the Picasa posts, which are somehow irregular posts and moreover
made by the ?rm on its own, we would ?nd that 90% of the social media posts have been sourced from Twitter,
which is in line with the other ?rms.
102 Classi?cation of Social Media Posts
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Unibail-Rodamco Arcadis Fugro Akzo Nobel ArcelorMittal Aegon PostNL NS Albert Heijn ABN AMRO Bol.com KLM Blokker TomTom Coca-Cola Heineken C-1000 Philips
P
e
r
c
e
n
t
a
g
e
o
f
S
o
c
ia
l
M
e
d
ia
P
o
s
t
s
Short-termfinancial results Customer relations Employee relations Operational performance Product and service quality Alliances Supplier relations Environmental performance Product and service innovation Community Undefined Spam Short-termfinancial results Customer relations Employee relations Operational performance Product and service quality Alliances Supplier relations Environmental performance Product and service innovation Community Undefined Spam
Figure B-2: Social Media Posts Classi?cation
We have classi?ed all 7.067 collected social media posts in the ABN AMRO data set. The results are shown in
?gure B-1. The columns show the social media post categories as a percentage of the total classi?ed posts of
that ?rm. For ABN AMRO, we can conclude that the majority of the social media posts (71%) are related to
the community category, which is due to the large share of promotion type of posts. As explained, this ?gure
is high because of the Picasa posts made by the ?rm in the sample period. Secondly, we see that many posts
in the ABN AMRO data set are related to customer relations. ABN AMRO operates a web care team that
actively monitors the social media sites for customers that complain or ask questions. Here, we see that the
web care team of ABN AMRO partly replaces the traditional telephone help desk. The table also shows that
a substantial part of the posts are not able to be classi?ed into one of the categories. These posts particularly
relate to people using ABN AMRO o?ces as a point of recognition to meet or illustrate where they are. For
example, posts contain appointments like “let’s meet in front of the ABN AMRO o?ce before we go into town”.
Whereas these messages de?nitely contain the name of the ?rm, they do not contain any relevant information
for the ?rm that can be marked as “social intelligence”.
Aegon
Aegon N.V. provides insurances, pensions and asset management products to over 47 million clients in the
world. As can be concluded from table C-1, the majority (81%) of the collected posts have Twitter as a source.
The channel distribution of Aegon is in line with the distribution of other ?rms.
Over the sample period, 1.449 posts have been collected containing the word “Aegon”. All these posts have been
classi?ed in the categories that were established in section 4-3-1. Figure B-1 shows the results of the classi?cation
of Aegon’s social media posts. As can be concluded, the majority (43%) of the social media posts are related
to the KPI category involving community related indicators. More speci?cally, the majority of the community
related posts are classi?ed as public image posts. Public image posts contain the ?rm’s name, indicating that
people are talking about the ?rm, but the posts are not purposed to get in contact with the ?rm. When
analysing the public image posts more detailed, we ?nd that many of these posts relate to Aegon’s sponsoring
activities. The ?rm is for example sponsor of the Dutch soccer club Ajax, it sponsors a tennis centre called
Aegon Arena and it sponsors the Dutch rowing team. The public image social media posts related to sponsoring
activities may serve as a measure to determine the exposure of the sponsoring activities. A substantial part of
Aegon’s social media messages have been classi?ed as spam. Spam messages contain the name of the ?rm but
do not relate the ?rm. The dataset of Aegon revealed that a person named “Aegon The Conqueror” showed up
often in posts. Aegon The Conqueror is a character in a popular TV series called Game of Thrones. These kind
of spam messages, that show up in the dataset because the name of the ?rm is commonly used for the naming
of other entities or people, do not contain any information that may be valuable for the company.
103
Akzo Nobel
Akzo Nobel N.V. is a Dutch ?rm active in paint, lacquer, coatings and other specialised chemical products. The
company operates in 80 countries and employs around 55.000 people. As can be concluded from table C-1, the
922 posts that have been collected are for the majority sourced from Twitter.
Figure B-1 shows the results of the classi?cation process of the 922 social media messages that have been
collected. As can be concluded, around 21% of the social media messages that are related to Akzo Nobel refer
to news articles. Around 18% of the social media posts are made by distributors of Akzo Nobel’s products
who promote their products on retail websites like Amazon. Another substantial part – 20% – of the messages
relates to stock related discussions. A remarkable phenomenon in the classi?cation of Akzo Nobel’s social media
messages is the fact that there are almost no messages related to customer relations, whereas we have seen
customer related messages in the social media posts of other companies.
Albert Heijn
Albert Heijn is a Dutch supermarket, which is a subsidiary of Royal Ahold N.V. Albert Heijn operates around
850 stores in the Netherlands, and is with 34% market share the market leader in the Netherlands. The company
also operates stores in Belgium, Germany and Curacao. Table C-1 illustrates that almost all of the collected
social media messages containing Albert Heijn have been sourced from Twitter.
As ?gure B-1 shows, a substantial part (20%) of the social media messages of Albert Heijn involve customer
relations management. On the one hand customers are responsible for many complaints (6% of the classi?ed
posts), while Albert Heijn’s web care team actively responses to these messages by either showing an
understanding (4% of the classi?ed posts) of the customer’s complaint, or even explaining (5% of the classi?ed
posts) the customer something that they asked. Here, we clearly see that the ?rm’s help desk moves to social
media. Another substantial part (26%) of social media messages involving Albert Heijn relate to the ?rm’s
public image. These messages contain customer’s opinions about e.g. the latest Albert Heijn commercial or
about the products that they bought in the store.
Arcadis
Arcadis N.V. is a Dutch engineering consultancy, o?ering solutions in the ?eld of infrastructure, civilised areas
and environmental projects. The company is active in more than 70 countries and employing around 18.000
people. During the period of monitoring the companies, we collected 455 posts related to Arcadis. This ?gure
is substantially lower than the number of posts that we collected from other ?rms. Of the 455 social media
posts that have been collected and are related to Arcadis, around 93% has been derived from Twitter.
Shown by ?gure B-1 on page 101, 32% of the collected messages involve Arcadis’ ?nancial performance
discussions. These posts particularly involve professionals discussing the ?nancial performance of the company
and talking about future projections of the company’s ?nancial position. Next, 18% of the Arcadis’ posts have
been classi?ed as public image posts. These posts particularly refer to publicised articles or news messages that
have been spread by the company. Presumably, the company has spread these articles / news messages to gain
exposure. The ?rm could use the amount of public image messages that refer to the articles as a measure to
determine the exposure as a result of the publicised articles.
ArcelorMittal
ArcelorMittal is the world’s largest steel producer, active in 27 countries and employing 320.000 people. During
the period of monitoring, we collected 5.532 social media posts related to ArcelorMittal. Of these posts, around
83% have been derived from Twitter. Other sources involved blogs (5%), news sites (2%) and other platforms.
As can be seen in ?gure B-1, social media messages related to ArcelorMittal do not involve customer relations.
Rather, the social media messages relate to the ?rm’s public image (32% of the posts) and professionals (20%
of the posts) talking about the company. The posts classi?ed as public image involve marketing activities of the
?rm. Especially the “ArcelorMittal Orbit”, a steel tower constructed by the ?rm on the 2012 London Olympics,
was subject of discussion. This category of social media messages is the only one that involves non-professional
104 Classi?cation of Social Media Posts
people, because all other social media messages related to ArcelorMittal involve professionals. 20% of the
sample posts have been classi?ed as professionals, in which professionals discuss joint-ventures or the industries’
outlook. Finally, a substantial amount of the posts (16%) relate to the companies ?nancial results.
Blokker
Blokker is a Dutch store selling products related to household. The Blokker stores are subsidiaries of Blokker
Holding B.V., which operates over 2.900 stores in 11 countries, thereby employing 25.000 people. In total, 2.769
social media messages related to Blokker have been collected, of which the majority (91%) has been derived
from Twitter. Please see table C-1 on page 108 for a distribution of the sources of the social media posts.
Figure B-1 (page 101) shows that 27% of the classi?ed social media posts are related to the community category.
The community category comprises messages that reveal the community’s perception of the ?rm. Of these 27%
community related posts, the vast majority consists of social media posts that have been classi?ed as public
image posts. Public image posts are messages that are made by individuals and contain the ?rm’s name, without
explicitly seeking contact with the ?rm. In the case of Blokker, many messages involve statements of people
announcing to their followers that they are planning to visit one of the stores, or people referring to articles that
are sold by the ?rm. As can be concluded, customer also ask questions to the ?rm and also complain about the
?rm. However, no messages of the company have been found in the sample that respond to these messages.
Bol.com
Bol.com is an online web-shop selling a variety of products, such as books, DVDs, games, blu-rays, electronics,
computers, etcetera. Since the foundation of the company in 1999, it has shown solely growth percentages of
18% y-o-y and above in terms of revenues. As of 2012, Bol.com is a subsidiary of Royal Ahold N.V. We have
collected 5.782 social media posts that are related to Bol.com, of which 89% has been derived from Twitter.
Of the classi?ed posts related to Bol.com, around 50% are related to the ?rm’s public image. For Bol.com, these
public image posts particularly involve people who share to their followers their recent purchase of a product
through Bol.com or people illustrating to other people that a certain product can be bought at Bol.com. Next,
a substantial amount of the posts consists of promotion activities. These posts are made by the ?rm or by
?rm’s selling products through Bol.com’s website for marketing purposes. Finally, 17% of the Bol.com classi?ed
social media posts relate to customer relations. We clearly see that people ask questions or complain, and that
Bol.com’s web care team is consequently responsible for the social media posts that have been classi?ed as
either explaining ?rm (3%) or understanding ?rm (4%).
C1000
C1000 is a Dutch supermarket organisation with a market share of 11,5%, employing around 7.000 people and
operating 425 stores in the Netherlands. In the future, many C1000 stores will be turned into Jumbo stores
as Jumbo Supermarkten acquired C1000 in 2012. Allmost all of the 5.782 social media posts that have been
collected in relation with C1000 have been derived from Twitter.
The majority of the social media posts (50%) of C1000 have been classi?ed as unde?ned, implying that these
messages cannot be classi?ed into one of the categories that have been established. When taking a closer look
at the unde?ned messages, we see that many users refer to C1000 as a location to meet each other, or users
share that they are heading for or just returned from C1000. These messages do not contain any management
information, and are therefore classi?ed as unde?ned. The other portion of the classi?ed social media messages
related to C1000 are classi?ed as public image posts. These messages contain statements of customers sharing
their followers what they have bought or seen at a C1000 store. Next, marketing campaigns are discussed by
people.
Coca-Cola
Coca-Cola is one of the many drinks o?ered by The Coca-Cola Company, selling Coca-Cola all over the world
(except for North-Korea and Cuba). As expected, Coca-Cola is one of the company’s in our sample that
105
delivered the most messages because it is one of the most famous brands of the world. During the monitoring
period, 32.953 messages containing the word Coca-Cola have been collected. In line with other companies in
our sample, 89% of the collected Coca-Cola social media messages have been derived from Twitter.
As ?gure B-1 illustrates, half of the classi?ed posts of Coca-Cola can be positioned under the category labelled
public image. These posts contain perceptions of customers to the company, to marketing campaigns of the ?rm
or are made by people talking about the ?rm’s sponsorships. Next, a substantial part of the classi?ed social
media posts of coca-cola are classi?ed as unde?ned. These messages to not contain any valuable management
information, and contain for instance statements of people that they are drinking Coca-Cola right now, or that
they wish that they were drinking one now.
Fugro
Fugro N.V. is a Dutch company that collects and interprets data related to the earth’s surface. The company
provides advice to ?rms active in the oil- and gas industry, the mining industry and the construction industry.
Fugro is active in over 50 countries, operating 275 o?ces and employing around 14.000 employees. During the
monitoring period, merely 428 social media posts related to Fugro have been collected. Of all ?rms in our
sample, there is no ?rm with less search results. Though the sample of Fugro is small, it shows a channel
distribution that is comparable to the other ?rms in our sample; around 90% of the derived post are sourced
from Twitter.
As indicated by ?gure B-1 on page 101, the vast majority (45%) of the social media posts related to Fugro
are classi?ed as posts made by professionals. These posts consists of professionals talking about new vessels
that Fugro either ordered or received, or how macro trends are e?ecting the market in which Fugro operates.
Furthermore, automated messages creating tools post a message each time that a Fugro vessel leaves or arrives
at a harbour. These posts are also classi?ed as professionals. Another substantial part of the social media posts
related to Fugro have been assigned to the category labelled stock related discussions. Unfortunately the two
categories that are responsible for the majority of the social media post categories do not o?er any information
to the company that is not available at the company internally.
Heineken
Heineken N.V. is a Dutch multinational providing beer and other drinks. The company is active in 178 countries,
employing 70.000 people. During our period of monitoring 39.425 social media posts have been collected by the
search terms related to Heineken, making Heineken the ?rm with the most mentions of our sample. As can be
concluded from table C-1, 82% of these social media posts have been sourced from Twitter, while Facebook is
responsible for 15% of the messages.
Of the 2.050 posts that have been classi?ed, 33% have been classi?ed as public image posts (see table B-1). Many
of these posts relate to Heineken commercials seen by people on the television, or other ways that Heineken
pursues to expose itself such as the Holland Heineken House at the 2012 London Olympics. The number of
public image posts can serve as a measure to determine the success of the desired exposure by these kind of
marketing events. Another substantial part of Heineken’s posts are considered as spam, because they refer to
other entities, people naming themselves Heineken on the web or people that are actually named Heineken.
KLM
KLM – Koninklijke Luchtvaart Maatschappij – N.V. is a Dutch airliner that operates 116 airplanes across the
globe. The ?rm has three subsidiaries – KLM Cityhopper, Martinair and Transavia.com – while the parent
company is Air France-KLM. The search terms used to ?lter out the social media posts related to KLM (see
table 4-5 on page 58) resulted in 26.364 messages which have been scraped. 86% of these messages have been
sourced from Twitter, while Facebook is responsible for 9% of the posts. These ?gures are in line with the
channel distribution of other ?rms in the sample.
10% of the collected posts – i.e. 2.498 posts – have been classi?ed into one of the social media categories that
have been established in section 4-3-1. Figure B-1 shows that an astonishing 75% of these posts have been
classi?ed as being spam. A closer look at the spam classi?ed posts shows that the letters K, L and M are
106 Classi?cation of Social Media Posts
used by many people in their username, e.g. @Klm_babe, @DaOne_KLM, @klm_klm_klm, @klm_nico, and
@miyu_klm. Probably, the initials of these people correspond with the name of the ?rm. However, the dataset
of KLM also shows a substantial amount of posts classi?ed as customer relations. KLM operates a webcare team
that actively monitors the messages directed to KLM, at which the employees of the webcare team consequently
respond. Again – as we have seen with ABN AMRO and Albert Heijn – we see that the traditional customer
help-desk is (partly) moving to the social media.
NS
Nederlandse Spoorwegen N.V. (“NS”) is a Dutch railway company operating the main rail network in the
Netherlands. During the monitoring period, 5.863 social media posts related to NS have been collected. 85%
of the NS posts have been sourced from Twitter, while Facebook is responsible for 12% of these posts.
Figure B-1 shows that 33% of the classi?ed posts are related to customer relations. Especially the category
informing ?rm is over-represented in the dataset, this is due to the ?rm that uses social media to inform
customers that certain tracks of the network are subject to delays. These posts do not contain any information
that is not available internally, because the nature of the direction of these messages is outgoing; from ?rm to
customers. Next, 9% of the classi?ed posts are complaining customers, while 4% of the posts are questioning
customers. These posts contain information that may not be available to the ?rm. The ?rm operates a web
care team that answers questions and shows understanding for the experienced problems (4% and 4% of the
posts respectively).
Philips
Koninklijke Philips Electronics N.V. is a Dutch electronics ?rm active in more than 60 countries and employing
122.000 people. The ?rm is organised into three main divisions: Philips Consumer Lifestyle, Philips Healthcare
and Philips Lighting. Philips is the largest manufacturer of lighting in the world. During the monitoring period
32.748 posts have been collected using the Philips search queries of table 4-5, corresponding to almost 3.000
daily posts. Philips is the second largest ?rm in our sample in terms of collected social media messages. 68%
of the messages have been sourced from Twitter, 12% from Facebook, 9% from Blogs and 10% from other
platforms including Friendfeed and YouTube.
As indicated by ?gure B-1, the vast majority (54%) of the social media messages related to Philips are classi?ed
as distributors posts. The distributors posts are made by professionals that are selling Philips products to
consumers. Often, distributors use Amazon.com as a site to sell the products, while they use social media to
announce the public their o?ers. Philips is a common surname. As a result, many posts in the Philips sample
have been classi?ed as spam as they do contain the name Philips, but do not relate to the ?rm. 7% of the
classi?ed posts are labelled as product and service quality posts. These posts comprise product- reviews and
experiences of users, containing valuable information for R&D related activities.
PostNL
PostNL N.V. is a mail and parcel company operating in the Netherlands, Germany, Italy and the United
Kingdom. In total, 1.323 social media messages have been collected using the search terms related to PostNL.
91% of these posts have been collected from Twitter, which is in line with other ?rms in the sample.
More than in any other dataset of the ?rms, 30% of PostNL’s social media messages have been classi?ed as
complaining customers. The messages contain statements of customers who are complaining about the service
of the ?rm, about broken parcels, late deliveries, etc. The ?rm operates a web care team, though it does
only respond to a limited amount of complaining and questioning customers. 20% of the PostNL posts have
been classi?ed as public image posts, posts made by individuals talking about the ?rm. A surprising ?gure in
PostNL’s social media classi?cation overview is the high percentage – 14% – of employee posts. Apparently,
PostNL’s employees – and especially postmen – share that they are working at the ?rm.
107
TomTom
TomTom N.V. is a Dutch producer of automotive navigation systems. TomTom is Europe’s leading manufacturer
of navigation systems. The ?rm employs around 3.500 people. During the monitoring period, 32.748 social media
messages have been collected using the search queries related to TomTom. In line with other ?rms, 91% of
these messages have been derived from Twitter.
Of all classi?ed TomTom messages, 49% has been labelled unde?ned, implying that these messages could not
be assigned to any of the other categories. These messages relate to the ?rm, although people use TomTom in
their message, though these messages do not contain any valuable information for the ?rm. The high percentage
of unde?ned posts is due to the fact that people use the word TomTom as a term for navigation systems in
generally, or to refer to anyone who is navigating. Apparently, TomTom has become a word in the general
vocabulary used by the society, though the relation to the ?rm TomTom is not always present. 7% of the
posts are related to product and service quality, in which users share experiences of the usage of TomTom’s
products. Another 8% of the posts are classi?ed as product and service innovation posts, in which users either
make innovative suggestions for future products, or share their opinion towards new products / services. These
two categories contain valuable information for R&D departments, on the one hand to measure the success of
existing products and on the other hand to develop new products.
Unibail-Rodamco
Unibail-Rodamco is a ?rm specialised in commercial property investments. It is the largest commercial real
estate company in Europe, managing three types of assets; shopping centers, convention centers and o?ce
properties. Unibail-Rodamco employs around 1.500 people. Only 512 social media messages related to
Unibail-Rodamco have been collected, i.e. 39 daily posts on average. 95% of these posts have been sourced
from Twitter.
More than in any other ?rm in our sample, as illustrated by ?gure B-1 75% of the posts related to
Unibail-Rodamco relate to ?nancial results. These posts are either related to ?nancial performance discussions
of the ?rm or to stock related discussions. Furthermore, 11% of the classi?ed posts are messages classi?ed as
professionals; people writing about the ?rm from a professional point of view.
Appendix C
Social Media Platform Distribution
Table C-1: Social Media Channel Distribution
Platform Facebook Twitter Blogs News Other Total
Abs % Abs % Abs % Abs % Abs % Abs
ABN AMRO 124 2% 3.000 42% 70 1% 15 0% 3.858 55% 7.067
Aegon 110 8% 1.173 81% 79 5% 20 1% 67 5% 1.449
Akzo Nobel 30 3% 806 87% 43 5% 25 3% 18 2% 922
Albert Heijn 328 3% 11.116 96% 77 1% 1 0% 59 1% 11.581
Arcadis 8 2% 422 93% 9 2% 10 2% 6 1% 455
ArcelorMittal 439 8% 4.569 83% 296 5% 89 2% 139 3% 5.532
Blokker 155 6% 2.526 91% 71 3% 3 0% 14 1% 2.769
Bol.com 472 8% 5.124 89% 115 2% - 0% 71 1% 5.782
C1000 362 3% 10.583 96% 81 1% 5 0% 33 0% 11.064
Coca-Cola 1.653 5% 29.347 89% 999 3% 69 0% 885 3% 32.953
Fugro 6 1% 385 90% 20 5% 15 4% 2 0% 428
Heineken 5.726 15% 32.332 82% 494 1% 122 0% 751 2% 39.425
KLM 2.316 9% 22.601 86% 617 2% 90 0% 740 3% 26.364
NS 703 12% 4.970 85% 103 2% - 0% 87 1% 5.863
Philips 4.641 12% 26.260 68% 3.404 9% 138 0% 4.007 10% 38.450
PostNL 77 6% 1.207 91% 27 2% - 0% 12 1% 1.323
TomTom 1.308 4% 29.787 91% 630 2% 67 0% 956 3% 32.748
Unibail-Rodamco 3 1% 487 95% 9 2% 12 2% 1 0% 512
Total 18.461 8% 186.695 83% 7.144 3% 681 0% 11.706 5% 224.687
109
2%
42%
55%
ABN AMRO
8%
81%
5%
1%
5%
Aegon
n=7.067 n=1.449
1%
0%
Facebook Twitter Blogs News Other Platforms
81%
Facebook Twitter Blogs News Other Platforms
2%
2%
2%
1%
Arcadis
8%
5%
2%
2%
ArcelorMittal
n=455 n=5.532
93%
Facebook Twitter Blogs News Other Platforms
83%
Facebook Twitter Blogs News Other Platforms
3%
1%
0%
0% 5%
3%
0%
3%
n=11.064 n=32.953
96%
C-1000
Facebook Twitter Blogs News Other Platforms
89%
Coca-Cola
Facebook Twitter Blogs News Other Platforms
110 Social Media Platform Distribution
3%
87%
5%
3%
2%
Akzo Nobel
3%
1%
0%
0%
Albert Heijn
n=922 n=11.581
87%
Facebook Twitter Blogs News Other Platforms
96%
Facebook Twitter Blogs News Other Platforms
6%
3%
0%
0%
Blokker
8%
2%
0%
1%
Bol.com
n=2.769 n=5.782
91%
Facebook Twitter Blogs News Other Platforms
89%
Facebook Twitter Blogs News Other Platforms
1%
5%
4%
0%
15%
1%
0%
2%
n=428 n=39.425
90%
Fugro
Facebook Twitter Blogs News Other Platforms
82%
Heineken
Facebook Twitter Blogs News Other Platforms
111
4%
2%
0%
3%
TomTom
1%
2%
2%
0%
Unibail-Rodamco
n=32.748 n=512
9%
2%
0%
3%
KLM
12%
2%
0%
1%
NS
91%
Facebook Twitter Blogs News Other Platforms
95%
Facebook Twitter Blogs News Other Platforms
n=26.364 n=5.863
86%
Facebook Twitter Blogs News Other Platforms
85%
Facebook Twitter Blogs News Other Platforms
12%
9% 0%
11%
6%
2%
0%
1%
n=38.450 n=1.323
68%
Philips
Facebook Twitter Blogs News Other Platforms
91%
PostNL
Facebook Twitter Blogs News Other Platforms
Appendix D
Descriptive Statistics of Social Media
Post Categories
Social Media Categories across Di?erent Customer Relation Types
Table D-1: Social Media Post Categories: Across Customer Relation Type (Descriptives)
Customer Relation Type B2C B2B Total
µ N ? µ N ? µ N ?
short_term_?nancial_results 1,27% 13 1,54% 37,00% 5 23,27% 11,20% 18 20,01%
?nancial_performance_discussions 0,20% 13 0,44% 17,70% 5 21,04% 5,06% 18 13,01%
stock_related_discussions 1,07% 13 1,56% 19,31% 5 13,10% 6,13% 18 10,62%
customer_relations 12,76% 13 12,34% 0,56% 5 0,93% 9,37% 18 11,80%
explaining_?rm 1,92% 13 1,87% 0,07% 5 0,10% 1,41% 18 1,79%
understanding_?rm 1,95% 13 2,19% 0,07% 5 0,10% 1,43% 18 2,04%
thanking_?rm 0,43% 13 0,41% 0,26% 5 0,59% 0,39% 18 0,45%
informing_?rm 0,83% 13 2,98% 0,00% 5 0,00% 0,60% 18 2,54%
questioning_?rm 2,15% 13 1,79% 0,08% 5 0,09% 1,57% 18 1,78%
complaining_customer 4,73% 13 8,16% 0,02% 5 0,04% 3,42% 18 7,19%
thanking_customer 0,60% 13 0,64% 0,07% 5 0,10% 0,45% 18 0,59%
employee_relations 4,08% 13 5,91% 5,59% 5 4,17% 4,50% 18 5,41%
recruitment 1,34% 13 1,91% 4,85% 5 3,78% 2,31% 18 2,93%
employee_posts 2,74% 13 4,43% 0,66% 5 0,51% 2,16% 18 3,86%
operational_performance 0,22% 13 0,50% 0,00% 5 0,00% 0,16% 18 0,43%
product_and_service_quality 1,77% 13 2,51% 0,05% 5 0,12% 1,29% 18 2,26%
alliances 0,44% 13 0,88% 0,13% 5 0,20% 0,35% 18 0,76%
supplier_relations 0,11% 13 0,28% 0,00% 5 0,00% 0,08% 18 0,24%
environmental_performance 0,15% 13 0,43% 0,00% 5 0,00% 0,11% 18 0,37%
product_and_service_innovation 1,52% 13 2,21% 0,28% 5 0,63% 1,17% 18 1,97%
community 39,81% 13 20,50% 44,61% 5 17,79% 41,14% 18 19,39%
promotion 7,08% 13 15,50% 2,46% 5 3,15% 5,79% 18 13,28%
news 1,22% 13 1,46% 6,82% 5 8,21% 2,78% 18 4,90%
public_image 25,50% 13 14,61% 13,61% 5 12,12% 22,20% 18 14,67%
professionals 1,27% 13 1,21% 18,14% 5 15,98% 5,96% 18 11,02%
distributors 4,73% 13 15,36% 3,58% 5 8,00% 4,41% 18 13,48%
unde?ned 20,74% 13 18,88% 10,19% 5 4,90% 17,81% 18 16,77%
spam 17,13% 13 20,00% 1,58% 5 1,24% 12,81% 18 18,28%
113
S
o
c
i
a
l
M
e
d
i
a
C
a
t
e
g
o
r
i
e
s
a
c
r
o
s
s
D
i
?
e
r
e
n
t
I
n
d
u
s
t
r
i
e
s
I n d u s t r y
S
o
c
i
a
l
M
e
d
i
a
P
o
s
t
C
a
t
e
g
o
r
y
?
N
?
?
N
?
?
N
?
?
N
?
?
N
?
?
N
?
?
N
?
?
N
?
s
h
o
r
t
_
t
e
r
m
_
f
i
n
a
n
c
i
a
l
_
r
e
s
u
l
t
s
1
8
,
5
9
%
2
3
,
4
7
%
1
,
3
3
%
3
1
,
5
1
%
0
,
0
0
%
3
0
,
0
0
%
1
,
1
4
%
3
0
,
9
5
%
1
,
5
4
%
2
2
,
1
8
%
2
7
,
1
5
%
3
4
1
,
8
2
%
3
6
,
2
3
%
2
0
,
6
8
%
1
1
,
2
0
%
1
8
2
0
,
0
1
%
f
i
n
a
n
c
i
a
l
_
p
e
r
f
o
r
m
a
n
c
e
_
d
i
s
c
u
s
s
i
o
n
s
4
,
5
8
%
2
5
,
2
5
%
0
,
2
5
%
3
0
,
4
3
%
0
,
0
0
%
3
0
,
0
0
%
0
,
5
2
%
3
0
,
8
4
%
0
,
1
8
%
2
0
,
2
6
%
1
5
,
8
2
%
3
2
7
,
4
0
%
1
5
,
9
3
%
2
2
2
,
5
3
%
5
,
0
6
%
1
8
1
3
,
0
1
%
s
t
o
c
k
_
r
e
l
a
t
e
d
_
d
i
s
c
u
s
s
i
o
n
s
1
4
,
0
1
%
2
8
,
7
2
%
1
,
0
8
%
3
1
,
6
5
%
0
,
0
0
%
3
0
,
0
0
%
0
,
6
3
%
3
1
,
0
8
%
1
,
3
6
%
2
1
,
9
3
%
1
1
,
3
3
%
3
1
4
,
4
8
%
2
0
,
2
9
%
2
2
1
,
8
6
%
6
,
1
3
%
1
8
1
0
,
6
2
%
c
u
s
t
o
m
e
r
_
r
e
l
a
t
i
o
n
s
0
,
3
1
%
2
0
,
1
8
%
2
,
2
7
%
3
1
,
8
3
%
9
,
6
0
%
3
9
,
7
4
%
2
8
,
7
6
%
3
1
2
,
9
9
%
1
1
,
2
5
%
2
7
,
9
4
%
7
,
1
6
%
3
9
,
1
6
%
1
,
1
0
%
2
1
,
5
5
%
9
,
3
7
%
1
8
1
1
,
8
0
%
e
x
p
l
a
i
n
i
n
g
_
f
i
r
m
0
,
0
5
%
2
0
,
0
8
%
0
,
4
1
%
3
0
,
3
6
%
1
,
5
6
%
3
2
,
7
0
%
3
,
4
2
%
3
1
,
4
4
%
2
,
1
2
%
2
0
,
6
8
%
1
,
5
4
%
3
2
,
4
2
%
0
,
1
1
%
2
0
,
1
6
%
1
,
4
1
%
1
8
1
,
7
9
%
u
n
d
e
r
s
t
a
n
d
i
n
g
_
f
i
r
m
0
,
0
5
%
2
0
,
0
8
%
0
,
2
9
%
3
0
,
4
7
%
1
,
3
2
%
3
2
,
2
9
%
3
,
9
0
%
3
3
,
1
7
%
2
,
9
7
%
2
1
,
2
4
%
0
,
9
6
%
3
1
,
2
7
%
0
,
1
1
%
2
0
,
1
6
%
1
,
4
3
%
1
8
2
,
0
4
%
t
h
a
n
k
i
n
g
_
f
i
r
m
0
,
0
0
%
2
0
,
0
0
%
0
,
3
1
%
3
0
,
5
4
%
0
,
2
9
%
3
0
,
5
1
%
0
,
5
9
%
3
0
,
5
3
%
0
,
5
6
%
2
0
,
0
3
%
0
,
3
1
%
3
0
,
3
7
%
0
,
6
6
%
2
0
,
9
3
%
0
,
3
9
%
1
8
0
,
4
5
%
i
n
f
o
r
m
i
n
g
_
f
i
r
m
0
,
0
0
%
2
0
,
0
0
%
0
,
0
0
%
3
0
,
0
0
%
0
,
0
0
%
3
0
,
0
0
%
3
,
5
9
%
3
6
,
2
1
%
0
,
0
0
%
2
0
,
0
0
%
0
,
0
0
%
3
0
,
0
0
%
0
,
0
0
%
2
0
,
0
0
%
0
,
6
0
%
1
8
2
,
5
4
%
q
u
e
s
t
i
o
n
i
n
g
_
f
i
r
m
0
,
1
0
%
2
0
,
0
1
%
0
,
7
4
%
3
0
,
5
8
%
2
,
7
3
%
3
2
,
1
4
%
3
,
0
9
%
3
2
,
3
0
%
2
,
5
5
%
2
2
,
3
2
%
1
,
0
5
%
3
1
,
3
6
%
0
,
1
1
%
2
0
,
1
6
%
1
,
5
7
%
1
8
1
,
7
8
%
c
o
m
p
l
a
i
n
i
n
g
_
c
u
s
t
o
m
e
r
0
,
0
5
%
2
0
,
0
6
%
0
,
3
9
%
3
0
,
2
1
%
3
,
1
2
%
3
2
,
5
4
%
1
3
,
5
5
%
3
1
5
,
1
7
%
2
,
3
1
%
2
2
,
6
2
%
1
,
8
9
%
3
1
,
9
0
%
0
,
0
0
%
2
0
,
0
0
%
3
,
4
2
%
1
8
7
,
1
9
%
t
h
a
n
k
i
n
g
_
c
u
s
t
o
m
e
r
0
,
0
5
%
2
0
,
0
8
%
0
,
1
3
%
3
0
,
1
4
%
0
,
5
8
%
3
0
,
9
0
%
0
,
6
3
%
3
0
,
1
2
%
0
,
7
4
%
2
1
,
0
4
%
0
,
7
7
%
3
0
,
9
0
%
0
,
1
1
%
2
0
,
1
6
%
0
,
4
5
%
1
8
0
,
5
9
%
e
m
p
l
o
y
e
e
_
r
e
l
a
t
i
o
n
s
6
,
1
8
%
2
5
,
5
2
%
0
,
8
6
%
3
0
,
7
8
%
6
,
9
8
%
3
2
,
0
5
%
8
,
0
6
%
3
1
1
,
5
7
%
0
,
1
7
%
2
0
,
2
5
%
2
,
0
7
%
3
1
,
4
4
%
7
,
2
1
%
2
3
,
6
8
%
4
,
5
0
%
1
8
5
,
4
1
%
r
e
c
r
u
i
t
m
e
n
t
5
,
5
8
%
2
5
,
3
1
%
0
,
8
1
%
3
0
,
8
5
%
0
,
7
4
%
3
0
,
4
9
%
3
,
2
7
%
3
3
,
5
6
%
0
,
1
7
%
2
0
,
2
5
%
1
,
2
3
%
3
0
,
6
0
%
5
,
9
6
%
2
3
,
4
6
%
2
,
3
1
%
1
8
2
,
9
3
%
e
m
p
l
o
y
e
e
_
p
o
s
t
s
0
,
6
1
%
2
0
,
2
1
%
0
,
0
5
%
3
0
,
0
8
%
6
,
2
3
%
3
2
,
3
7
%
4
,
7
8
%
3
8
,
1
7
%
0
,
0
0
%
2
0
,
0
0
%
0
,
8
0
%
3
0
,
9
6
%
1
,
0
3
%
2
0
,
5
3
%
2
,
1
6
%
1
8
3
,
8
6
%
o
p
e
r
a
t
i
o
n
a
l
_
p
e
r
f
o
r
m
a
n
c
e
0
,
0
0
%
2
0
,
0
0
%
0
,
0
0
%
3
0
,
0
0
%
0
,
0
0
%
3
0
,
0
0
%
0
,
4
7
%
3
0
,
7
3
%
0
,
0
2
%
2
0
,
0
3
%
0
,
4
5
%
3
0
,
7
8
%
0
,
0
0
%
2
0
,
0
0
%
0
,
1
6
%
1
8
0
,
4
3
%
p
r
o
d
u
c
t
_
a
n
d
_
s
e
r
v
i
c
e
_
q
u
a
l
i
t
y
0
,
1
4
%
2
0
,
1
9
%
3
,
6
7
%
3
3
,
7
3
%
0
,
9
5
%
3
0
,
5
9
%
0
,
5
3
%
3
0
,
4
1
%
3
,
3
1
%
2
4
,
5
7
%
0
,
3
1
%
3
0
,
5
3
%
0
,
0
0
%
2
0
,
0
0
%
1
,
2
9
%
1
8
2
,
2
6
%
a
l
l
i
a
n
c
e
s
0
,
2
2
%
2
0
,
3
1
%
0
,
0
3
%
3
0
,
0
5
%
0
,
0
0
%
3
0
,
0
0
%
0
,
8
6
%
3
1
,
4
8
%
0
,
3
8
%
2
0
,
2
8
%
0
,
7
6
%
3
1
,
2
5
%
0
,
1
2
%
2
0
,
1
7
%
0
,
3
5
%
1
8
0
,
7
6
%
s
u
p
p
l
i
e
r
_
r
e
l
a
t
i
o
n
s
0
,
0
0
%
2
0
,
0
0
%
0
,
3
4
%
3
0
,
5
9
%
0
,
0
5
%
3
0
,
0
8
%
0
,
0
2
%
3
0
,
0
4
%
0
,
0
0
%
2
0
,
0
0
%
0
,
0
6
%
3
0
,
1
0
%
0
,
0
0
%
2
0
,
0
0
%
0
,
0
8
%
1
8
0
,
2
4
%
e
n
v
i
r
o
n
m
e
n
t
a
l
_
p
e
r
f
o
r
m
a
n
c
e
0
,
0
0
%
2
0
,
0
0
%
0
,
0
5
%
3
0
,
0
5
%
0
,
5
5
%
3
0
,
9
0
%
0
,
0
0
%
3
0
,
0
0
%
0
,
0
0
%
2
0
,
0
0
%
0
,
0
5
%
3
0
,
0
9
%
0
,
0
0
%
2
0
,
0
0
%
0
,
1
1
%
1
8
0
,
3
7
%
p
r
o
d
u
c
t
_
a
n
d
_
s
e
r
v
i
c
e
_
i
n
n
o
v
a
t
i
o
n
0
,
7
1
%
2
1
,
0
0
%
1
,
2
1
%
3
0
,
4
0
%
0
,
3
5
%
3
0
,
5
5
%
0
,
3
7
%
3
0
,
3
7
%
3
,
8
8
%
2
5
,
4
3
%
2
,
0
7
%
3
2
,
2
2
%
0
,
0
0
%
2
0
,
0
0
%
1
,
1
7
%
1
8
1
,
9
7
%
c
o
m
m
u
n
i
t
y
5
7
,
4
5
%
2
4
,
0
9
%
5
4
,
6
4
%
3
9
,
2
9
%
2
6
,
8
6
%
3
1
,
4
7
%
2
0
,
1
5
%
3
1
1
,
8
0
%
4
9
,
2
8
%
2
3
2
,
8
3
%
4
2
,
9
0
%
3
2
8
,
3
8
%
4
6
,
7
4
%
2
3
,
6
2
%
4
1
,
1
4
%
1
8
1
9
,
3
9
%
p
r
o
m
o
t
i
o
n
1
,
7
9
%
2
0
,
6
9
%
1
,
8
2
%
3
0
,
9
6
%
0
,
8
7
%
3
0
,
8
4
%
1
,
3
3
%
3
0
,
9
1
%
1
1
,
3
9
%
2
1
2
,
6
3
%
1
9
,
2
6
%
3
3
1
,
4
8
%
4
,
0
7
%
2
5
,
4
3
%
5
,
7
9
%
1
8
1
3
,
2
8
%
n
e
w
s
1
3
,
4
2
%
2
1
0
,
4
7
%
1
,
7
0
%
3
2
,
0
6
%
0
,
6
5
%
3
1
,
0
2
%
0
,
7
5
%
3
0
,
3
9
%
0
,
0
0
%
2
0
,
0
0
%
2
,
4
0
%
3
1
,
6
5
%
3
,
3
3
%
2
3
,
0
6
%
2
,
7
8
%
1
8
4
,
9
0
%
p
u
b
l
i
c
_
i
m
a
g
e
2
1
,
6
9
%
2
1
4
,
5
7
%
2
9
,
6
5
%
3
2
2
,
5
6
%
2
4
,
7
9
%
3
1
,
4
8
%
1
6
,
8
2
%
3
1
1
,
7
7
%
3
5
,
8
7
%
2
2
0
,
1
5
%
1
6
,
3
2
%
3
1
7
,
5
8
%
1
0
,
8
7
%
2
1
0
,
7
4
%
2
2
,
2
0
%
1
8
1
4
,
6
7
%
p
r
o
f
e
s
s
i
o
n
a
l
s
1
1
,
6
0
%
2
1
1
,
9
6
%
2
,
0
7
%
3
1
,
4
5
%
0
,
1
1
%
3
0
,
1
1
%
1
,
2
5
%
3
0
,
8
5
%
1
,
0
3
%
2
1
,
4
6
%
4
,
9
2
%
3
4
,
9
9
%
2
8
,
4
7
%
2
2
2
,
8
5
%
5
,
9
6
%
1
8
1
1
,
0
2
%
d
i
s
t
r
i
b
u
t
o
r
s
8
,
9
5
%
2
1
2
,
6
5
%
1
9
,
3
9
%
3
3
1
,
5
2
%
0
,
4
5
%
3
0
,
7
7
%
0
,
0
0
%
3
0
,
0
0
%
1
,
0
0
%
2
1
,
4
1
%
0
,
0
0
%
3
0
,
0
0
%
0
,
0
0
%
2
0
,
0
0
%
4
,
4
1
%
1
8
1
3
,
4
8
%
u
n
d
e
f
i
n
e
d
1
3
,
5
8
%
2
5
,
5
5
%
1
7
,
0
9
%
3
1
4
,
6
5
%
4
3
,
5
2
%
3
6
,
5
9
%
5
,
0
6
%
3
4
,
8
6
%
2
6
,
2
5
%
2
3
1
,
8
5
%
9
,
5
9
%
3
6
,
3
9
%
7
,
5
9
%
2
5
,
1
2
%
1
7
,
8
1
%
1
8
1
6
,
7
7
%
s
p
a
m
2
,
8
3
%
2
0
,
6
3
%
1
8
,
5
1
%
3
5
,
8
4
%
1
1
,
1
5
%
3
9
,
8
4
%
3
4
,
5
9
%
3
3
7
,
7
4
%
3
,
9
1
%
2
2
,
7
1
%
7
,
4
3
%
3
1
2
,
4
5
%
1
,
0
2
%
2
0
,
1
2
%
1
2
,
8
1
%
1
8
1
8
,
2
8
%
W h o l e s a l e a n d R e t a i l
T r a n s p o r t a n d S t o r a g e
I n f o r m a t i o n a n d
C o m m u n i c a t i o n
F i n a n c i a l I n s t i t u t i o n s
C o n s u l t a n c y , R e s e a r c h
a n d O t h e r S p e c i a l i s e d
B u s i n e s s S e r v i c e s
T o t a l
M i n i n g a n d Q u a r r y i n g
I n d u s t r y
F
i
g
u
r
e
D
-
1
:
S
o
c
i
a
l
M
e
d
i
a
P
o
s
t
C
a
t
e
g
o
r
i
e
s
:
A
c
r
o
s
s
I
n
d
u
s
t
r
i
e
s
(
D
e
s
c
r
i
p
t
i
v
e
s
)
114 Descriptive Statistics of Social Media Post Categories
Boxplots of Social Media Categories across Firms
Case Processing Summary
Cases
Valid Missing Total
N Percent N Percent N Percent
alliances
supplier_relations
environmental_performan
ce
product_and_service_inno
vation
community
promotion
news
public_image
professionals
distributors
undefined
spam
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
18 100,0% 0 0,0% 18 100,0%
short_term_financial_results
short_term_financial_results
0,8
0,6
0,4
0,2
0,0
Unibail-Rodamco
Page 2
customer_relations
0,4
0,3
0,2
0,1
0,0
explaining_firm
Page 5
employee_relations
0,25
0,20
0,15
0,10
0,05
0,00
PostNL
recruitment
Page 13
operational_performance
1,25E-2
1,0E-2
7,5E-3
5,0E-3
2,5E-3
0,0E0
ABN AMRO
NS
PostNL
Bol.com
product_and_service_quality
Page 16
product_and_service_quality
0,08
0,06
0,04
0,02
0,00
Philips
TomTom
alliances
Page 17
alliances
0,030
0,025
0,020
0,015
0,010
0,005
0,000
PostNL
Aegon
supplier_relations
Page 18
115
supplier_relations
1,2E-2
1,0E-2
8,0E-3
6,0E-3
4,0E-3
2,0E-3
0,0E0
Heineken
ABN AMRO
Albert Heijn
NS
environmental_performance
Page 19
environmental_performance
0,020
0,015
0,010
0,005
0,000
Albert Heijn
ABN AMRO
product_and_service_innovation
Page 20
product_and_service_innovation
0,08
0,06
0,04
0,02
0,00
TomTom
Aegon
community
Page 21
community
0,8
0,6
0,4
0,2
0,0
promotion
Page 22
undefined
0,6
0,5
0,4
0,3
0,2
0,1
0,0
spam
Page 28
spam
0,8
0,6
0,4
0,2
0,0
KLM
Page 29
Appendix E
Corporate Engagement
The next page lists the user names of ?rms that have been found in our dataset. These user names have been
used to assign social media messages in the category ‘?rm-to-customer’.
117
I
D
A
B
N
A
M
R
O
#
P
o
s
t
s
A
e
g
o
n
#
P
o
s
t
s
A
k
z
o
N
o
b
e
l
#
P
o
s
t
s
A
l
b
e
r
t
H
e
i
j
n
#
P
o
s
t
s
A
r
c
a
d
i
s
#
P
o
s
t
s
A
r
c
e
l
o
r
M
i
t
t
a
l
#
P
o
s
t
s
B
l
o
k
k
e
r
#
P
o
s
t
s
B
o
l
.
c
o
m
#
P
o
s
t
s
C
-
1
0
0
0
#
P
o
s
t
s
1
a
b
n
a
m
r
o
_
t
e
g
e
l
e
n
2
A
e
g
o
n
R
e
l
i
g
a
r
e
L
i
f
e
I
n
s
u
r
a
n
c
e
1
A
k
z
o
N
o
b
e
l
_
A
&
A
C
2
A
l
b
e
r
t
H
e
i
j
n
N
u
l
a
n
d
3
A
R
C
A
D
I
S
G
S
1
A
r
c
e
l
o
r
M
i
t
t
a
l
U
S
A
1
B
l
o
k
k
e
r
2
b
o
l
.
c
o
m
b
a
b
y
4
9
C
1
0
0
0
K
o
e
t
s
i
e
r
1
2
A
B
N
A
M
R
O
C
o
m
m
e
r
c
i
a
l
1
A
E
G
O
N
S
e
g
u
r
o
s
1
A
k
z
o
N
o
b
e
l
G
r
a
d
u
a
t
e
s
2
A
l
b
e
r
t
H
e
i
j
n
S
t
i
e
n
s
3
A
R
C
A
D
I
S
L
F
N
1
a
r
c
e
l
o
r
m
i
t
t
a
l
_
d
0
B
l
o
k
k
e
r
2
B
o
l
.
c
o
m
F
i
l
m
T
V
4
7
C
1
0
0
0
N
o
o
r
d
m
a
n
1
3
A
B
N
A
M
R
O
E
c
h
t
2
A
E
G
O
N
U
K
1
a
k
z
o
n
o
b
e
l
a
e
r
o
1
A
l
b
e
r
t
H
e
i
j
n
S
t
r
i
j
p
1
A
R
C
A
D
I
S
S
h
e
l
t
e
r
4
A
r
c
e
l
o
r
M
i
t
t
a
l
D
o
f
a
s
c
o
0
b
l
o
k
k
e
r
h
o
l
t
e
n
1
b
o
l
.
c
o
m
g
a
m
e
s
9
C
1
0
0
0
P
u
t
s
1
4
a
b
n
a
m
r
o
e
c
h
t
2
a
e
g
o
n
_
n
l
3
A
k
z
o
N
o
b
e
l
A
e
r
o
0
a
l
b
e
r
t
h
e
i
j
n
1
a
r
c
a
d
i
s
_
l
f
n
0
A
r
c
e
l
o
r
M
i
t
t
a
l
2
B
o
l
.
c
o
m
M
u
z
i
e
k
4
4
C
1
0
0
0
_
F
r
a
n
e
k
e
r
1
5
A
B
N
A
M
R
O
M
E
P
P
E
L
5
4
7
A
E
G
O
N
N
e
d
e
r
l
a
n
d
9
G
r
a
d
s
@
A
k
z
o
N
o
b
e
l
1
A
l
b
e
r
t
H
e
i
j
n
X
L
T
i
l
b
u
r
g
2
6
5
A
R
C
A
D
I
S
B
e
l
g
i
q
u
e
8
A
r
c
e
l
o
r
M
i
t
t
a
l
O
r
b
i
t
1
B
o
l
.
c
o
m
P
a
r
t
n
e
r
1
1
1
c
1
0
0
0
b
e
e
k
m
a
n
1
6
a
b
n
a
m
r
o
n
i
j
m
e
g
e
n
3
.
8
5
1
A
E
G
O
N
R
e
l
i
g
a
r
e
1
5
A
k
z
o
N
o
b
e
l
1
A
l
b
e
r
t
B
e
z
o
r
g
s
e
r
v
i
c
e
1
A
R
C
A
D
I
S
2
7
B
o
l
.
c
o
m
s
e
r
v
i
c
e
2
7
5
c
1
0
0
0
d
a
l
f
s
e
n
2
7
A
B
N
A
M
R
O
1
A
E
G
O
N
1
A
k
z
o
N
o
b
e
l
A
e
r
o
s
p
a
c
e
1
A
l
b
e
r
t
H
e
i
j
n
1
A
R
C
A
D
I
S
C
o
r
p
C
o
m
m
1
B
o
l
.
c
o
m
a
c
t
i
e
s
2
6
C
1
0
0
0
B
e
e
k
m
a
n
1
8
A
B
N
A
M
R
O
N
V
1
A
E
G
O
N
N
e
d
e
r
l
a
n
d
4
4
A
l
b
e
r
t
H
e
i
j
n
1
6
1
8
1
b
o
l
.
c
o
m
5
4
C
1
0
0
0
K
e
e
s
E
l
z
i
n
g
a
6
9
A
B
N
A
M
R
O
M
a
r
k
e
t
s
7
a
e
g
o
n
s
e
g
u
r
o
s
1
A
H
R
a
p
p
o
r
t
s
t
r
a
a
t
1
b
o
l
_
c
o
m
1
2
c
1
0
0
0
d
a
y
s
2
1
0
a
b
n
a
m
r
o
t
u
r
b
o
1
A
E
G
O
N
C
l
a
s
s
i
c
4
b
o
l
_
c
o
m
_
a
c
t
i
e
s
6
c
1
0
0
0
m
u
l
d
e
r
1
1
1
A
B
N
A
M
R
O
Y
E
R
S
E
K
E
O
P
E
N
1
b
o
l
_
c
o
m
_
a
p
i
1
c
1
0
0
0
n
i
j
e
n
e
r
i
n
g
1
1
2
A
B
N
A
M
R
O
E
c
o
n
B
u
r
e
a
u
2
1
b
o
l
_
c
o
m
_
b
a
b
y
2
c
1
0
0
0
o
f
f
e
r
e
i
n
2
1
3
b
o
l
_
c
o
m
_
b
a
n
e
n
2
c
1
0
0
0
p
a
n
n
i
n
g
e
n
1
1
4
b
o
l
_
c
o
m
_
g
a
m
e
s
2
c
1
0
0
0
p
u
t
s
2
1
5
b
o
l
_
c
o
m
_
k
o
k
e
n
4
8
c
1
0
0
0
w
a
h
l
e
1
1
6
b
o
l
_
c
o
m
_
m
u
z
i
e
k
1
C
1
0
0
0
W
i
e
g
m
a
n
s
6
1
7
b
o
l
_
c
o
m
_
p
a
r
t
n
e
r
2
1
8
w
e
r
k
e
n
b
i
j
b
o
l
.
c
o
m
8
O
w
n
p
o
s
t
s
4
.
4
3
7
8
0
8
2
7
7
4
2
4
5
6
9
9
3
0
%
T
o
t
a
l
s
a
m
p
l
e
6
3
%
6
%
1
%
2
%
9
%
0
%
0
%
9
%
0
%
I
D
C
o
c
a
-
C
o
l
a
#
P
o
s
t
s
F
u
g
r
o
#
P
o
s
t
s
H
e
i
n
e
k
e
n
#
P
o
s
t
s
K
L
M
#
P
o
s
t
s
N
S
#
P
o
s
t
s
P
h
i
l
i
p
s
#
P
o
s
t
s
P
o
s
t
N
L
#
P
o
s
t
s
T
o
m
T
o
m
#
P
o
s
t
s
U
n
i
b
a
i
l
R
o
d
a
m
c
o
#
P
o
s
t
s
1
C
o
c
a
-
C
o
l
a
G
B
1
F
u
g
r
o
S
e
a
c
o
r
e
1
H
e
i
n
e
k
e
n
M
u
s
i
c
H
a
l
l
1
9
K
L
M
A
i
r
l
i
n
e
s
2
N
S
o
n
l
i
n
e
6
P
h
i
l
i
p
s
A
r
e
n
a
2
O
R
P
o
s
t
N
L
1
3
t
o
m
t
o
m
b
e
n
e
l
u
x
3
U
n
i
b
a
i
l
R
o
d
a
m
c
o
3
2
C
o
c
a
-
C
o
l
a
2
4
x
7
1
H
e
i
n
e
k
e
n
N
L
4
K
L
M
B
e
l
g
i
u
m
1
N
S
S
t
a
t
i
o
n
Z
w
o
l
l
e
2
P
h
i
l
i
p
s
A
r
g
e
n
t
i
n
a
1
P
o
s
t
N
L
1
T
o
m
T
o
m
7
0
3
C
o
c
a
-
C
o
l
a
A
r
g
e
n
t
i
n
a
3
H
e
i
n
e
k
e
n
t
w
e
e
t
s
8
K
L
M
C
a
n
a
d
a
3
N
S
S
t
a
t
i
o
n
s
6
1
P
h
i
l
i
p
s
A
V
E
N
T
U
K
3
2
P
o
s
t
N
L
P
r
i
d
e
1
T
o
m
T
o
m
B
r
a
s
i
l
4
4
C
o
c
a
-
C
o
l
a
B
o
i
D
a
D
o
n
4
h
e
i
n
e
k
e
n
_
h
o
u
s
e
0
K
L
M
C
o
u
p
o
n
C
o
d
e
s
1
N
S
U
n
o
f
f
i
c
i
a
l
7
7
P
h
i
l
i
p
s
B
r
a
s
i
l
1
P
o
s
t
N
L
V
a
n
d
a
a
g
2
T
o
m
T
o
m
D
e
u
t
s
c
h
l
a
n
d
2
5
C
o
c
a
-
C
o
l
a
B
r
a
s
i
l
3
h
e
i
n
e
k
e
n
_
t
w
e
e
t
s
0
K
L
M
D
e
u
t
s
c
h
l
a
n
d
9
N
S
V
e
r
t
r
a
g
i
n
g
3
P
h
i
l
i
p
s
C
o
r
p
C
o
m
m
s
1
P
o
s
t
N
L
W
e
b
c
a
r
e
2
7
T
o
m
t
o
m
E
n
g
l
i
s
h
1
6
C
o
c
a
-
C
o
l
a
C
a
n
a
d
a
4
7
H
e
i
n
e
k
e
n
E
e
t
c
a
f
e
2
K
L
M
E
C
U
A
D
O
R
1
7
4
4
N
S
v
e
r
t
r
a
g
i
n
g
1
P
h
i
l
i
p
s
E
n
t
L
i
g
h
t
i
n
g
8
P
o
s
t
N
L
R
e
c
r
u
i
t
m
e
n
t
2
T
o
m
T
o
m
F
r
a
n
c
e
7
7
C
o
c
a
C
o
l
a
B
e
l
g
i
u
m
2
H
e
i
n
e
k
e
n
H
o
u
s
e
6
A
i
r
F
r
a
n
c
e
-
K
L
M
2
N
S
1
3
2
P
h
i
l
i
p
s
L
i
g
h
t
i
n
g
3
T
o
m
T
o
m
I
t
a
l
i
a
3
8
C
o
c
a
-
C
o
l
a
1
2
R
o
y
a
l
D
u
t
c
h
A
i
r
l
i
n
e
s
1
N
S
H
i
s
p
e
e
d
2
2
P
h
i
l
i
p
s
1
T
o
m
T
o
m
O
f
f
i
c
i
a
l
1
5
9
C
o
c
a
-
C
o
l
a
C
l
o
t
h
i
n
g
2
K
L
M
f
a
r
e
s
2
6
n
s
v
e
r
t
r
a
g
i
n
g
n
l
5
P
h
i
l
i
p
s
F
a
s
t
4
T
o
m
T
o
m
P
o
r
t
u
g
a
l
3
1
0
C
o
c
a
-
C
o
l
a
C
o
l
o
m
b
i
a
4
K
L
M
I
r
e
l
a
n
d
2
N
S
_
W
i
j
c
h
e
n
2
P
h
i
l
i
p
s
G
r
o
o
m
i
n
g
2
T
o
m
T
o
m
S
o
u
t
h
A
f
r
i
c
a
6
1
1
C
o
c
a
-
C
o
l
a
C
R
1
K
L
M
I
T
A
L
I
A
5
n
s
_
z
w
o
l
l
e
1
P
h
i
l
i
p
s
H
e
a
l
t
h
c
a
r
e
2
7
T
o
m
T
o
m
S
p
a
i
n
7
1
2
C
o
c
a
-
C
o
l
a
E
c
u
a
d
o
r
1
K
L
M
K
a
z
a
k
h
s
t
a
n
6
P
h
i
l
i
p
s
H
o
m
e
L
i
g
h
t
i
n
g
3
T
o
m
T
o
m
S
w
i
t
z
e
r
l
a
n
d
4
1
3
C
o
c
a
-
C
o
l
a
E
l
S
a
l
v
a
d
o
r
1
K
L
M
M
e
x
i
c
o
9
P
h
i
l
i
p
s
I
n
L
i
g
h
t
i
n
g
3
1
4
C
o
c
a
-
C
o
l
a
E
s
p
a
ñ
a
2
K
L
M
R
U
S
S
I
A
1
P
h
i
l
i
p
s
I
n
d
i
a
2
1
5
C
o
c
a
C
o
l
a
F
M
C
O
L
O
M
B
I
A
3
K
L
M
S
p
a
i
n
3
4
P
h
i
l
i
p
s
I
n
d
o
n
e
s
i
a
1
0
1
1
6
C
o
c
a
-
C
o
l
a
F
M
E
C
U
A
D
O
R
2
K
L
M
U
K
1
6
P
h
i
l
i
p
s
J
a
p
a
n
9
1
7
C
o
c
a
-
C
o
l
a
F
r
e
e
s
t
y
l
e
2
K
L
M
z
a
k
e
l
i
j
k
1
P
h
i
l
i
p
s
J
o
b
s
N
A
9
1
8
C
o
c
a
-
C
o
l
a
G
B
2
k
l
m
_
e
s
1
P
h
i
l
i
p
s
J
o
b
s
N
A
9
1
9
C
o
c
a
-
C
o
l
a
G
u
a
t
e
m
a
l
a
1
K
L
M
_
J
P
1
0
P
h
i
l
i
p
s
L
i
g
h
t
i
n
g
1
2
2
0
C
o
c
a
-
C
o
l
a
H
B
C
I
t
a
l
i
a
1
k
l
m
_
p
t
1
P
h
i
l
i
p
s
S
p
a
i
n
4
2
1
C
o
c
a
-
C
o
l
a
H
o
n
d
u
r
a
s
1
k
l
m
_
u
k
1
P
h
i
l
i
p
s
S
u
p
p
o
r
t
U
K
1
2
2
C
o
c
a
-
C
o
l
a
I
n
d
o
n
e
s
i
a
3
k
l
m
_
u
s
1
P
h
i
l
i
p
s
T
h
a
i
l
a
n
d
1
0
2
3
C
o
c
a
-
C
o
l
a
K
o
r
e
a
1
0
5
k
l
m
f
a
r
e
s
3
P
h
i
l
i
p
s
T
u
r
k
i
y
e
2
6
2
4
C
o
c
a
-
C
o
l
a
N
i
c
a
r
a
g
u
a
1
k
l
m
i
t
a
l
i
a
1
P
h
i
l
i
p
s
U
K
G
r
a
d
u
a
t
e
s
4
6
2
5
C
o
c
a
-
C
o
l
a
P
a
n
a
m
á
1
k
l
m
m
a
l
a
y
s
i
a
1
P
h
i
l
i
p
s
_
N
e
d
e
r
l
a
n
d
6
2
6
C
o
c
a
-
C
o
l
a
V
e
n
e
z
u
e
l
a
1
1
k
l
m
n
o
r
w
a
y
1
P
h
i
l
i
p
s
C
a
r
e
N
L
4
2
6
C
o
c
a
-
C
o
l
a
V
e
n
e
z
u
e
l
a
1
1
k
l
m
n
o
r
w
a
y
1
P
h
i
l
i
p
s
C
a
r
e
N
L
4
2
7
C
o
c
a
-
C
o
l
a
W
e
b
3
K
L
M
R
o
y
a
l
D
u
t
c
h
A
i
r
l
i
n
e
1
p
h
i
l
i
p
s
j
a
p
a
n
2
2
8
C
o
c
a
C
o
l
a
B
u
f
f
a
l
o
1
p
h
i
l
i
p
s
j
o
b
s
n
a
3
2
9
C
o
c
a
C
o
l
a
I
s
r
a
e
l
3
9
p
h
i
l
i
p
s
l
i
g
h
t
2
3
0
C
o
c
a
C
o
l
a
T
v
1
P
h
i
l
i
p
s
L
i
g
h
t
i
n
g
1
3
1
p
h
i
l
i
p
s
n
l
j
o
b
s
1
3
2
P
h
i
l
i
p
s
S
o
u
n
d
P
r
o
j
e
c
t
3
4
3
3
p
h
i
l
i
p
s
s
p
a
i
n
2
3
4
P
h
i
l
i
p
s
T
u
r
k
i
y
e
0
3
5
p
h
i
l
i
p
s
u
k
g
r
a
d
s
1
8
O
w
n
p
o
s
t
s
2
6
1
0
3
9
1
8
8
4
3
1
2
3
9
0
4
6
1
2
5
3
%
T
o
t
a
l
s
a
m
p
l
e
1
%
0
%
0
%
7
%
5
%
1
%
3
%
0
%
1
%
F
i
g
u
r
e
E
-
1
:
S
o
c
i
a
l
M
e
d
i
a
P
o
s
t
s
C
l
a
s
s
i
?
c
a
t
i
o
n
Bibliography
Agichtein, E., Castillo, C., Donato, D., Gionis, A., & Mishne, G. (2008). Finding high-quality content in social
media. In Proceedings of the international conference on web search and web data mining (pp. 183–194).
Association for Computing Machinery.
Araujo, T., & Neijens, P. (2012). Friend me: Which factors in?uence top global brands participation in social
network sites. Internet Research, 22, 626 – 640.
Auditore, P. J. (2012a). The market for social media and business intelligence tools. Asteria Research.
Retrieved July 31, 2012, from http://sandhill.com/article/the-market-for-social-media-and
-business-intelligence-tools/
Auditore, P. J. (2012b). Share-unisphere research study: Adoption of social media bi on the rise.
SHARE-Unisphere. Retrieved July 31, 2012, from http://www.dbta.com/Articles/Editorial/
Trends-and-Applications/SHARE-Unisphere-Research-Study-Adoption-of-Social-Media-BI-on
-the-Rise-83119.aspx
Azvine, B., Cui, Z., & Nauck, D. D. (2005). Towards real-time business intelligence. BT Technology Journal ,
23, 214–225.
Berthon, P. R., Pitt, L. F., McCarthy, I., & Kates, S. M. (2007). When customers get clever: Managerial
approaches to dealing with creative consumers. Business Horizons, 50(1), 39–47.
Bonneau, J., & Preibusch, S. (2010). The privacy jungle: On the market for data protection in social networks.
In T. Moore, D. Pym, & C. Ioannidis (Eds.), Economics of information security and privacy (pp. 121–167).
Springer US.
Bos, W., & Tarnai, C. (1999). Content analysis in empirical social research. International Journal of Educational
Research, 31(8), 659–671.
Bouwman, H., Faber, E., Haaker, T., Kijl, B., & de Reuver, M. (2008). Conceptualizing the stof model. In
H. Bouwman, H. de Vos, & T. Haaker (Eds.), Mobile service innovation and business models (pp. 31 –
69). Berlin, Heidelberg: Springer.
Boyd, D. M., & Ellison, N. B. (2007). Social network sites: De?nition, history, and scholarship. Journal of
Computer-Mediated Communication, 13(1), 210–230.
Bryant, L. (2011). Moving from social media monitoring to social business intelligence. Headshift, Dachis Group.
Retrieved July 25, 2012, from http://www.headshift.com/our-blog/2011/06/15/from-social-media
-monitoring-to-social-business-intelligence/
Bughin, J., Chui, M., & Manyika, J. (2010). Clouds, big data, and smart assets: Ten tech-enabled business
trends to watch. McKinsey & Co. Quarterly. (White Paper).
CBS. (2012). Standaard bedrijfsindeling 2008. versie 2012. Centraal Bureau voor de Statistiek.
Cha?ey, D. (2009). E-business and e-commerce management (4th edition). Upper Saddle River, NJ, USA:
Prentice-Hall, Inc.
Chen, D.-N., Jeng, B., Lee, W.-P., & Chuang, C.-H. (2008). An agent-based model for consumer-to-business
electronic commerce. Expert Systems with Applications, 34(1), 469–481.
DesAutels, P. (2011). Ugis: Understanding the nature of user-generated information systems. Business Horizons,
54(3), 185 - 192.
Dey, L., & Haque, S. K. M. (2008). Opinion mining from noisy text data. In Proceedings of the second workshop
on analytics for noisy unstructured text data (pp. 83–90). New York, NY, USA: ACM.
Dhar, V., & Chang, E. A. (2009). Does chatter matter? the impact of user-generated content on music sales.
Journal of Interactive Marketing, 23(4), 300–307.
Bibliography 119
Doeland, D. (2012). Social media & internet key performance indicators. Retrieved August 10,
2012, from http://ddmca.nl/diensten/social-media-and-internet-monitoring/social-media-key
-performance-indicators/
Dong-Hun, L. (2010). Growing popularity of social media and business strategy. Korean Consumer & Society.,
112–117.
Dugan, L. (2012). Twitter to surpass 500 million registered users on wednesday. Retrieved August 15, 2012,
from http://www.mediabistro.com/alltwitter/500-million-registered-users_b18842
Eccles, R. (1991). The performance measurement manifesto. Harvard Business School , 69, 131–137.
Eckerson, W. W. (2009). Performance management strategies. how to create and deploy e?ective metrics. The
Data Warehousing Institute. (White Paper).
Elbashir, M. Z., Collier, P. A., & Davern, M. J. (2008). Measuring the e?ects of business intelligence systems:
The relationship between business process and organizational performance. International Journal of
Accounting Information Systems, 9(3), 135–153.
evolve24. (2012). Website. Retrieved September 26, 2012, from http://www.evolve24.com/
Fortuin, L. (1988). Performance indicators – why, where and how? European Journal of Operational Research,
34, 1–9.
Gartner. (2009). Gartner reveals ?ve business intelligence predictions for 2009 and beyond. Retrieved June 17,
2012, from http://www.gartner.com/it/page.jsp?id=856714
Gast, A., & Zanini, M. (2012). The social side of strategy. McKinsey Quarterly.
Gates, S. (2001). Aligning strategic performance measures and results. The Conference Board. Balanced
Scorecard Interest Group.
Gessner, G. H., & Volonio, L. (2005). Quick response improves returns on business intelligence investments.
Information Systems Management, 22(3), 66–74.
Ghose, A., & Panagiotos, I. (2010). The economining project at nyu: Studying the economic value of
user-generated content on the internet. Journal of Revenue and Pricing Management, 8, 241–246.
Gianfortoni, P., Adamson, D., & Rosé, C. P. (2011, July). Modeling of stylistic variation in social media with
stretchy patterns. In Proceedings of the ?rst workshop on algorithms and resources for modelling of dialects
and language varieties (pp. 49–59). Edinburgh, Scotland: Association for Computational Linguistics.
Guo, J., Zhang, P., Tan, J., & Guo, L. (2012). Mining hot topics from twitter streams. Procedia Computer
Science, 9, 2008–2011.
Hanna, R., Rohm, A., & Crittenden, V. L. (2011). We’re all connected: The power of the social media
ecosystem. Business Horizons, 54(3), 265–273.
Huang, Q., Li, Q., Li, H., & Yan, Z. (2012). An approach to incremental deep web crawling based on incremental
harvest model. Procedia Engineering, 29(0), 1081–1087.
IBM. (2011). From social media to social crm. reinventing the customer relationship. IBM Institute for Business
Value analysis. (White Paper).
Ittner, C., Larcker, D., & Randall, T. (2003). Performance implications of strategic performance measurement
in ?nancial services ?rms. Accounting, Organizations and Society, 28, 715–741.
Jansen, B. J., Zhang, M., Sobel, K., & Chowdury, A. (2009). Twitter power: Tweets as electronic word of
mouth. Journal of the American Society for Information Science, 60(11), 2169–2188.
Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite! the challenges and opportunities of social
media. Business Horizons, 53(1), 59–68.
Kaplan, A. M., & Haenlein, M. (2012). The britney spears universe: Social media and viral marketing at its
best. Business Horizons, 55(1), 27–31.
Kaplan, R. S., & Norton, D. P. (1992). The balanced scorecard – measures that drive performance. Harvard
Business Review(5), 71–79.
Kaplan, R. S., & Norton, D. P. (1993). Putting the balanced scorecard to work. Harvard Business Review,
71(5), 134–147.
Kaplan, R. S., & Norton, D. P. (1996). Using the balanced scorecard as a strategic management system.
Harvard Business Review(5), 1–13.
Kapow. (2009). Five key lessons for converting social media data into business intelligence. Kapow Software
(White Paper).
Kennerley, M., & Neely, A. (2003). Measuring performance in a changing business environment. International
Journal of Operations and Production Management, 23(2), 213–229.
Kietzmann, J. H., Hermkens, K., McCarthy, I. P., & Silvestre, B. S. (2011). Social media? get serious!
understanding the functional building blocks of social media. Business Horizons, 54(3), 241–251.
Kirtis, A. K., & Karahanb, F. (2011). To be or not to be in social media arena as the most cost-e?cient
marketing strategy after the global recession. Procedia Social and Behavioral Sciences, 24, 260–268.
120 Bibliography
Klassen, A. (2009). Forget twitter; your best online marketing tool is the humble product review. Ad Age, 1.
Kozinets, R. V., de Valck, K., Wojnicki, A. C., & Wilner, S. J. (2010). Networked narratives: Understanding
word-of-mouth marketing in online communities. Journal of Marketing, 74(2), 71–89.
KPMG. (2009). Does your business intelligence tell you the whole story?
Krippendor?, K. (2004). Content analysis: An introduction to its methodology. (2nd, Ed.). Thousand Oaks,
CA: Sage.
Laroche, M., Habibi, M. R., & Richard, M.-O. (2012). To be or not to be in social media: How brand loyalty
is a?ected by social media? International Journal of Information Management(0), -.
Lee, C.-H. (2012). Unsupervised and supervised learning to evaluate event relatedness based on content mining
from social-media streams. Expert Systems with Applications, 39, 13338–13356.
Lonnqvist, A., & Pirttimaki, V. (2006). The measurement of business intelligence. Information Systems
Management, 23(1), 32–40.
Lovejoy, K., Waters, R., & Saxton, G. D. (2012). Engaging stakeholders through twitter: How nonpro?t
organizations are getting more out of 140 characters or less. Public Relations Review, 38, 313–318.
Marwick, A. E., & Boyd, D. (2011). I tweet honestly, i tweet passionately: Twitter users, context collapse, and
the imagined audience. New Media & Society, 13(1), 114–133.
McCorkindale, T. (2010). Can you see the writing on my wall? a content analysis of the fortune 50’s facebook
social networking sites. Public Relations, 4, 1–13.
Molloy, J. C. (2011). The open knowledge foundation: Open data means better science. Public Libary of
Science, 9(12).
Murdough, C. (2009). Social media measurement: It’s not impossible. Journal of Interactive Advertising, 10,
94–99.
Murugesan, S. (2007). Understanding web 2.0. IT Professional , 8, 34–41.
Neely, A., & Adams, C. (2005). Performance prism. In E. in Chief: Kimberly Kempf-Leonard (Ed.), Encyclopedia
of social measurement (pp. 41–48). New York: Elsevier.
Neely, A., Adams, C., & Crowe, P. (2001). The performance prism in practice. In Measuring business excellence
(Vol. 5, pp. 6–12). MCB University Press.
Neely, A., Bourne, B., & Kennerley, M. (2000). Performance measurement system design: Developing and
testing a process-based approach. International Journal of Operations and Production Management, 20,
1119–1145.
Neely, A., Gregory, M., & Platts, K. (1995). Performance measurement system dsign: A literature review and
research agenda. International Journal of Operations and Production Management, 15, 80 – 116.
Netbase. (2012). Website. Retrieved September 26, 2012, from http://www.netbase.com/our-network/
partners/sap-social-media-analytics/
Neuendorf, K. A. (2002). The content analysis guidebook. Thousand Oaks, CA: Sage.
O’reilly, T. (2007). What is web 2.0: Design patterns and business models for the next generation of software.
Communications & Strategies, 65, 17–37.
Osterwalder, A., & Pigneur, Y. (2010). Business model generation. Hoboken, New Jersey: John Wiley & Sons,
Ltd.
Patterson, A. (2012). Social-networkers of the world, unite and take over: A meta-introspective perspective on
the facebook brand. Journal of Business Research, 65(4), 527–534.
Peterson, R. A., & Wilson, W. R. (1992). Measuring customer satisfaction: Fact and artifact. Journal of the
Academy of Marketing Science, 20, 61–71.
Pirttimäki, V., & Hannula, M. (2003). Process models of business intelligence. Frontiers of E-Business Research,
250–260.
PWC. (2007). Guide to key performance indicators. communicating the measures that matter. Price Waterhouse
Coopers. (White Paper).
Radian6. (2012). Website. Retrieved September 26, 2012, from http://www.radian6.com/
Ramakrishnan, T., Jones, M. C., & Sidorova, A. (2012). Factors in?uencing business intelligence (bi) data
collection strategies: An empirical investigation. Decision Support Systems, 52(2), 486–496.
Ratner, B. (2003). Statistical modeling and analysis for database marketing: e?ective techniques for mining big
data. Chapman & Hall / CRC Press.
Reinhold, O., & Alt, R. (2011). Analytical social crm: Concept and tool support. BLED 2011 Proceedings,
Paper 50.
Rouibah, K., & Ould-ali, S. (2002). Puzzle: a concept and prototype for linking business intelligence to business
strategy. The Journal of Strategic Information Systems, 11(2), 133–152.
Shahin, A., & Ali Mahbod, M. (2007). Prioritization of key performance indicators: An integration of analytical
hierarchy process and goal setting. International Journal of Productivity and Performance Management,
Bibliography 121
56, 226 – 240.
Sloan, P. (2012). Facebook earnings match estimates; 955 million users. Retrieved August 15,
2012, from http://news.cnet.com/8301-1023_3-57480865-93/facebook-earnings-match-estimates
-955-million-users/
Smith, A. N., Fischer, E., & Yongjian, C. (2012). How does brand-related user-generated content di?er across
youtube, facebook, and twitter? Journal of Interactive Marketing, 26(2), 102–113.
Smith, R. (2006). Key performance indicators. the basics of maintenance and reliability. Chapter 6.
Stephens, W. (2012). Research and social media: Why & how to do a content analysis. Retrieved
July 20, 2012, from http://www.accountplanningblog.com/2012/05/research-and-social-media
-why-how-to-do.html
Takac, C., Hinz, O., & Spann, M. (2011). The social embeddedness of decision making: opportunities and
challenges. Electronic Markets, 21, 185-195.
Tirunillai, S., & Tellis, G. (2012). Does chatter really matter? dynamics of user-generated content and stock
performance. Marketing Science, 31, 198–215.
Tsai, Y., & Cheng, Y. (2012). Analyzing key performance indicators (kpis) for e-commerce and internet
marketing of elderly products: A review. Archives of Gerontology and Geriatrics, 55(1), 126 - 132.
Turban, E., Lee, J., King, D., & Chung, H. (1999). Electronic commerce: A managerial perspective.
Prentice-Hall, Inc., New Jersey, USA.
Twitter. (2012). Twitter turns six. Retrieved June 17, 2012, from http://blog.twitter.com/2012/03/
twitter-turns-six.html
Van Beek, D. (2006). De intelligente organisatie. prestatieverbetering en organisatie-ontwikkeling met business
intelligence. Tutein Nolthenius: ’s Hertogenbosch.
Watson, H., & Wixom, B. (2007). The current state of business intelligence. Computer, 40(9), 96–99.
Weinberg, B. D., & Pehlivan, E. (2011). Social spending: Managing the social media mix. Business Horizons,
54(3), 275–282.
Wikipedia. (2012). Wikipedia. Retrieved September 13, 2012, from http://en.wikipedia.org/wiki/
Wikipedia
Wisner, J., & Fawcett, S. (1991). Linking ?rm strategy to operating decisions through performance
measurement. Production and Inventory Management Journal , Third Quarter, 5 – 11.
You, W., Xia, M., Liu, L., & Liu, D. (2012). Customer knowledge discovery from online reviews. Electronic
Markets, 1–12.
Zeng, D., Chen, H., Lusch, R., & Li, S.-H. (2010). Social media analytics and intelligence. Intelligent Systems,
IEEE, 25(6), 13–16.
Zhang, M., Jansen, B. J., & Chowdhury, A. (2011). Business engagement on twitter: a path analysis. Electronic
Markets, 21(3), 161–175.
122 Bibliography
doc_586858168.pdf