Description
The ultimate goal of Business Intelligence (BI) initiative is to deliver data in the most appropriate way for making informed business decisions.
An Empirical Research into Metadata of Business Intelligence Systems
Yuriy Verbitskiy, University of South Australia, Adelaide, Australia, [email protected]
William Yeoh, University Tunku Abdul Rahman, Perak, Malaysia, [email protected]
Andy Koronios, University of South Australia, Adelaide, Australia, [email protected]
Abstract
The ultimate goal of Business Intelligence (BI)
initiative is to deliver data in the most appropriate
way for making informed business decisions. To
attain this goal, metadata for end users plays a
critical role in understanding the data provided by a
BI system. However, there has been little empirical
research about the metadata issues impacting the
implementation of BI systems. In particular, such
critical questions as „why does BI need metadata??
and „What metadata does BI need?? have not been
investigated in depth. Answers to these questions
are explored in an action research undertaken in a
large university in Australia. The findings of semi-
structured interviews show that success of a BI
system relies more on the metadata for business
users than on technical metadata. As a result, a
business-users-oriented metadata solution on object
and element level has been developed to improve
the users? satisfaction. The main issues that were
found during the implementation process are also
discussed. The research findings in conjunction with
the conclusion of the study may help BI
stakeholders in identifying the critical metadata
needs required for a more optimised BI deployment.
It also may assist BI vendors in improving the
business-side metadata which are crucial for a
successful BI endeavour.
Keywords: Business Intelligence, Metadata,
Object, Element, Glossary, Action Research
1.0 Introduction
Recently Business Intelligence (BI) applications
have been dominating the technology priority list of
many CIOs [3, 4, 5]. Gartner Research predicts that
the BI market will be in strong growth till 2011
[11]. According to Hancock and Toren [6],
“Business Intelligence is a set of concepts, methods,
and technologies for turning separated data in an
organization into useful information in order to
improve business performance”. The Main element
of a typical BI environment is integration process,
which means moving data from different sources
into one integrated place, storing the data, analysing
data and presenting the data to the end users. In
other words, BI is a system that allows business
users to leverage the data for making informed
business decisions[9].
Being a powerful approach for working with diverse
data within a large organization, the BI system is a
complex and resourceful undertaking. For instance,
Gartner formulated 12 key technical capabilities that
should be delivered by the BI environment and
advocates that the BI vendors implement at least 8
of these 12 capabilities [11]. Nevertheless, the main
indicator of success in implementing BI system is
the level of end user satisfaction. Understanding of
the complex BI environment by business users is
critical issue because “a lack of skills in using
information, tools and applications presents a big
barrier to the success of BI” [12]. The difficulties in
understanding BI are associated with inadequate
training [13] and disconnect between business and
technical users [14]. Focus on the technological
issues without balancing with business orientation is
a frequent reason for unsuccessful outcome [13].
Another problem is the understanding of the data
that business users work with in the BI environment.
Allowing end users to handle sets of data from
different data sources creates a problem because
most users are not familiar with those data. On one
hand, BI is a technique that allows the delivery of
required data from various sources to the end users.
On the other hand, business users have problems
with understanding the BI environment and the data
provided by BI applications. Without solving this
issue BI will not help users to make decisions
effectively. That is, the understanding of the data by
business users is a pressing issue in BI environment.
Making decisions based on the results of BI tools is
not surprisingly the biggest challenge for end users
[9, 16].
In response to this, metadata serves as a mechanism
that provides the context about the data and
information of a BI application [15]. It addresses the
how, when, why and what questions in a technical
environment [7]. Without metadata the data in an
enterprise cannot be understood properly [8]. To
date various BI tools allow the gathering and display
of data to users, which was not possible before.
However, the complexity of the technical
environment increases constantly with the number
and the diversity of BI applications. So the problem
of data misunderstanding within a BI environment is
getting even more critical. Using Metadata allows
us to solve this problem. Gartner Research argues
that metadata management is one of the most
important functionalities that BI environment should
deliver [11].
Therefore, the purpose of this research is to identify
why BI needs metadata and to investigate what
metadata BI requires. Subsequently a metadata
solution will be developed to address the research
findings. Most existing BI literature does not focus
on practical reasons why BI needs metadata.
Workshop on Advances in Intelligent Computing 2009
43
Fig 1. Areas of BI environment for this research
There are also several studies that discuss the
different types of metadata, but none has addressed
the actual metadata required for systems
implementation. This research is going to overcome
the problem by exploring the why and the elements
of metadata that are required through an action
research in a large organisation in Australia. The
next section discusses the research approach before
elaborating on data collection method and case
background. The third section presents and
discusses the findings of the study. Then follows the
conclusion and proposal for further research.
2.0 Research Approach
Despite the daunting complexities in implementing
BI systems, there has been little empirical research
about the metadata specifically for BI environment.
This study investigates the issues of metadata
implementation in the BI environment of a large
university in Australia. It identifies the specific
metadata needs of the organisation, and
subsequently proposes and implements a solution by
developing a prototype metadata solution. When a
process of change is the subject of research, the
most appropriate methodology is action research,
which was deemed to be the most suitable approach
for this project [1].
The research was conducted in collaboration with
the Australian university?s BI team. Having more
than thirty thousand students, it is crucial that the
institution exploit a powerful data analysis and
performance management tool. That is why the
university acknowledges the importance of the BI
technology and strives to deliver it to the wider
organisational community. It also admits the fact
that it does not have a consistent metadata
implementation which benefits business users of BI
environment. These factors make the university an
ideal environment for investigating the metadata
issues in BI environment. This paper presents one of
the key elements of the action research – diagnosing
the environment to define the problem - with the
following main research questions: Why does BI
need metadata implementation? And what metadata
does BI require?
2.1 Data Collection
To answer these two questions, several semi-
structured interviews with the university?s BI
stakeholders have been conducted. The interview
began with a predefined list of questions. Hence it
allows some latitude in getting detailed information
from interviewees during the interview, but at the
same time to have general questions so that all
participants are treated in a consistent manner [10].
The interview questions can be divided into two
main parts. The initial part of general questions
focus on discovering various situations where
metadata can be needed. Whilst second part relates
to more concrete questions which try to determine
the exact elements of metadata users would like to
have. To investigate the second research questions
and to explore the required metadata elements, the
BI environment under study was divided into 5
logical areas. As shown in figure 1, these areas are:
Data Sources, ETL process, Data Warehouse,
Business Rules and End user?s environment.
The interview was conducted with 9 research
participants from different departments and with
different roles in BI environment of the university: 6
from BI team, 2 are business analysts, and 1
business user. In general, there are four main roles
in the BI environment: business users, business
analysts, members of BI team and technical users.
Different roles reflect the different ability and
responsibility in implementing an enterprise-scale
BI system. Business users can only view the
reporting objects, such as cubes and reports.
Business analysts can produce and view the
reporting objects. BI team is responsible for data
modelling in BI, administration of data warehouse,
data quality, implementation of ETL and business
rules. In addition, the BI team develop complex
cubes and reports on request from business analysts
or business users. Technical users are responsible
for providing and maintaining the technical
infrastructure of the BI environment, such as data
source and data warehouse. Moreover, the different
roles also mean that they require different metadata
for their work. For example, business users do not
wish to see any technical descriptions, they only
want to understand and trust the data that they see in
a reporting object. At the meantime technical users
need technical information in performing their
work. All these differences were taken into account
during the interviews and analysis process. It is
important to highlight that business analysts are in
fact business users with additional skills in
manipulating the data. Business analysts are usually
the first point of contact for business users in
relation to complaints and requests for change.
Therefore, business analysts of the organisation
Workshop on Advances in Intelligent Computing 2009
44
could understand in-depth problems and needs of
various business users.
2.2 The Metadata Status of Case Organisation
At the time of the study, the organisation was
migrating to Cognos BI environment version 8.4
from the previous Cognos version 7. With the new
version of BI environment some novel possibilities
for metadata implementation appeared. Also the
university was changing its way of using the BI
environment. Before this front end BI environment
was encapsulated only in Cognos Upfront
application that showed cubes, reports and related
documents for users. Nonetheless, new approach
allows the delivering of BI objects into Staff portal
so BI will be embedded into the general user?s
environment. This new approach of delivering BI to
end users should be also considered during the
metadata implementation process.
Within Cognos version 7 BI environment, some
metadata techniques are in place but they are in
immature condition. Technical users use Excel
spreadsheets with links between technical and
business objects in the BI environment. Business
users exploit structured descriptions for cubes and
reports in Cognos Upfront environment. Also, they
are able to use „Glossary? application which
represents an ASP page with a list of terms and
descriptions, as illustrated in figure 2. While this
metadata provide some helpful information for the
business users, the BI team realises that they require
a more consistent, attractive and user-friendly
solution for metadata implementation. The
following section presents and discusses the
research findings.
Fig 2. Metadata in „Glossary? application
3.0 Research Findings and Discussions
One of the main questions of this research – whether
there is a need for metadata in BI environment –
was transformed to an interview question. Research
participants were asked to describe the situations
where he or she thinks that metadata would help
users of BI environment. The following section
presents the findings.
Q1: Can you tell me about the occasions when you
thought that metadata or additional description
would help users or yourself to understand data
better?
Participant 1
I suppose the main reason we want metadata is for end users looking at reports to be able to
understand where these numbers come from and what processing has done to them
Participant 2
I just think anywhere where people use reports.
Participant 3
I think probably in all instances while I publish reports we are trying to put as metadata
around the report as we can
Participant 4
But again it comes to interpretations of object and what does it actually represent. That is
where we need some sort of clarity and consistency.
Workshop on Advances in Intelligent Computing 2009
45
Participant 5
For developers understanding of the data is fine. For users it could be helpful. To clarify the
rules of presentation in the business tools, the scope of the data, to resolve different
definitions e.g. definition of mode for course is different from definition of mode for
program.
Participant 6
So for the people who use this dimension or column, they need some description – what
exactly this, how it was derived.
Participant 7
In general it is about data what you are looking at… I think to have some proms around that
that says “this is what you are looking at”.
Participant 8
Knowing where the information comes from, what the information represents
Participant 9
First of all, to help people to understand what they actually looking at
The answers to this question showed that different
staff with different roles almost identically
understands the importance and role of metadata.
The most common opinion regarding metadata was
that metadata should explain data on the reporting
object for business users. For this metadata should
show where the data comes from and the data
processing. Some participants from BI team said
that metadata will also help with testing, impact
analysis and documentation.
The next question relates to the finding that BI
efforts generally focus only on a limited set of data
points for semantic reconciliation, and therefore
miss the overall enterprise view [2]. Beyer said that
it is vital to develop a wider metadata model which
will be applied to the whole organisation in the
future rather than trying to connect differing data
using “tunnelling” approach.
Q2: What about the problem of seeing the overall
enterprise view?
Participant 1
Absolutely. When you look at the number and get more information of that. That needs to be
linked back to the source.
Participant 2
Yes, certainly that would be ideal – the enterprise view, which means that you have one
central location where thinks are defined than they linked to anywhere where they are used.
We don?t have a lot of that. That is part of our problem.
Participant 4
That would be useful, definitely. Probably far more from the developer perspective and
possibly from the user perspective.
Participant 5
That clarifies whole implementation process. It would be more helpful for BI developers.
Participant 6
Yes. I think there is problem.
Participant 7
I think there are two parts of it from end user point of view I don?t think it matters as long as
they got what they look for whereas these links are more important for technical roles, for
people who are building reports, building cubes where they are clear about the extract of
the data particularly when you have multiple systems.
Participant 8
I don?t think we have enterprise view of the information in the university at all… It is
organizational issue and cultural issue that we need to overcome.
Participant 9
I don?t think it is a problem. I think it is difficult in an organization which as complex as this
one.
According to the majority of the interviewees, they
do not have the overall enterprise view and so it
would be useful to have it. Nonetheless, some of the
arguments for overall enterprise view were related
to more technical side. Participants generally
believed that it would be useful for technical and
business analytics roles. The next question is about
understanding data or any elements in the BI
environment.
Q3: Are there problems with understanding data or
any elements in the environment?
Participant 1
Well, yes. That example with SATAC (reports) … What we do. We manually put the
metadata on the report… But often people will not see that.
Participant 2
Yes, I think so and I can give you classic example … o it is that basic that people in higher
and all over the uni, they don?t really understand the data because there is no place where
they can go to look up for consistent definition. I think it would help.
Participant 3
I think it is varying decrease of knowledge and understanding through all the levels.
Participant 4
Probably not. I think there is confusion about it. Most of the users understand the context
around it but if you take some infrequent users who are not necessarily know what the
difference is.
Workshop on Advances in Intelligent Computing 2009
46
Participant 5
Yes. The user should understand the business environment that they are dealing with.
Business users have different understanding of the data.
Participant 6
Yes, off course... you need a common description for each attribute or column or whatever.
Participant 7
I think there are two needs not only for business analysts who works with data but also for
someone who is not business analysts but uses the data he needs interpretation.
Participant 8
That is a huge barrier in trying to identify, especially from the metadata point of view, how
you define the information if it is used across the university with the same terminology but
the definitions are completely different.
Participant 9
Absolutely, yes. That is one of the things that we do all the time.
Almost all participants said that it is one of the key
problems that they deal with or want to solve. The
problem appears due to technical limitations,
complex business requirements in university,
organisational structure of the university and
because the consistent metadata implementation has
not been done yet. To continue the interview in the
area of data understanding, the next question
provided to participants an additional possible
ground why the data understanding could be
difficult.
Q4: Are there problems with interpretation of some
terms which could have different meaning for staff
members with different roles?
Participant 1 Absolutely. The way we get around that is again labelling the report with metadata and a
column but again it is manual metadata what ideally should be automated
Participant 2 Yes, definitely. Things like that actually can make quite a difference to the business
decisions that have been made at the uni ...
Participant 3 That is another big problem and it always go hand in hand with the question before –
understanding what the data actually is and what is used for.
Participant 4 Yes. People want to see things from the different perspective.
Participant 5 Yes, I think there are problems.
Participant 6 Yes.
Participant 7 Someone who uses their own system in their environment still needs to understand that that
terminology is for their group and when it is different system it could be different.
Participant 9 Yes.
All participants agreed that this is a problem and
that everybody sees the data from his/her
perspective. The problem is not only between
different departments within the organisation which
could interpret the same term differently but also
regarding the collaboration between the firm and
external organisations, such as government
agencies. The problem of translation between
business area and technical area is usually very
common. Next question has a goal to find out such
problems.
Q5: Do you recall any problems relating to a lack of
translation between technical terms and business
terms?
Participant 1
Yes … that is why we have that huge mapping spreadsheet which maps all of the business
descriptions, all the business names to be actually technical names.
Participant 2
Yes, probably… We have not standardized basically on one, and say that is what we mean
when we say that.
Participant 3
I have not encountered any by myself.
Participant 4
Users really don?t know what they want until they see it … Developers also need to
understand the business and how they use it. So you do get the mismatch between these two.
Participant 5
I am not sure about it.
Participant 7
The business person knows the business processes so it is to provide right information to
technical person because technical person quite often will not know business processes.
Participant 8
Every single day… So translation of something stored in a technical sense to a business
sense is difficult to understand.
Participant 9
You have to be very careful whenever you use language what particularly language you
using and who is going to read it. Certainly in terms of writing reports we write them very
differently depending on for what audience we are writing.
Workshop on Advances in Intelligent Computing 2009
47
The majority of the participants said that translation
between technical and business terms is a problem
that they see during their work. But as one
participant said it is an issue particularly for
technical people and business analysts who try to
implement business tasks in technical environment.
So for them translation between technical terms and
business terms is critical whereas for business
people it is not so important because they usually do
not need to understand technical side at all.
Q6: What about descriptions for calculations?
Participant 1
We call it business rules… Ideally more automated would be better. I can't think of anything
which would make more useful than links.
Participant 2
We don?t probably have a lot of them documented at the moment. Report writer could go
and act on this and see what calculation is... we have not actually, you know, put that out. It
is not available to anyone except the people who are actually working on that project at the
moment.
Participant 3
I can?t think about the cases where we do calculations as standard. If we use it in the report
we always put appropriate labels and description of the calculations.
Participant 4
We might have some for DW. I think they are useful from the technical perspective. But also
from the users perspective to see how is something actually derived.
Participant 5 For me the descriptions are okay. Maybe for users there are problems.
Participant 6 Yes, there is need for that.
Participant 7
Most of my usages in a cubes or SATAC cubes so it is more about totals. But clearly where
there is need for other measures these formulas need to be standard and clear.
Participant 8
From a business point of view the user don?t understand or don?t really need understand the
technical stuff, they just have to have the confidence. That kind of definition, business
definition should be the first that the user sees. And under that business definition it is
technical information.
Participant 9 To be honest I did not see that many, I am not sure.
It is noted that the participants from BI team require
detailed calculations to be shown because they are
responsible for that. However, there is no such
pressing need for business users to see a full
description of the calculation.
Q7: What about descriptions for aggregated metrics
that are used in different reports, for example in
quarterly or annual reports?
Participant 1
We don't do many aggregates, we just started. We don't say when they are the same as the
source. If they are not the same as the source than yes we need to be able to report that.
Descriptions in that aggregate so far have been assigned as actual source because we don't
introduce any new.
Participant 2
It is basically a calculation but we have come across the problems with the reporting where
it is not feasible to aggregate a certain measure
Participant 3
The aggregation tends to happen by default anyway. Others aggregations tend to be quite
straightforward e.g. by division or program.
Participant 4
I would say it could be a problem. Personally I have not experienced a lot because I have
not deal with that sort of information.
Participant 6 Yes, what I said is perfectly valued for aggregated metrics.
Participant 8
Aggregation is critical. But aggregation does not mean the sum of everything underneath
e.g. student count. So we use count them uniquely for the grouping of programs or classes. A
lot of people get confused with that example.
Participant 9
We use a lot of aggregated data but you need to be very clear what it is involved otherwise
people can misinterpret it. And the other thing I found – people will take some data from one
report and apply it into another report what is not applicable in that situation.
Most participants agreed that there could be complex
aggregations that need further explanations but such
cases are very infrequent. Usually aggregations are
straightforward and do not require explanations.
Q8: Is it worth to include the descriptions of reasons
why a term, calculation or metric was developed?
Workshop on Advances in Intelligent Computing 2009
48
Participant 1 Absolutely. And not just reasons but who authorised that.
Participant 2
You need to have things like the scope. Because often you will filter out certain pieces of
data for, I don?t how many reasons, depending on who is using it, what the rule is.
Participant 3
If you looking at the data I expect to see definitions and descriptions so I could know what
is happening. If I want to know why I would always think you will go to another place.
Participant 4
I think any information you give is worthwhile because it gives you picture of how things
work together.
Participant 5 In some cases - yes you probably need to, but in major cases you do not need to.
Participant 6 Yes, you need that.
Participant 7 I think the more important aspect, once it has been defined, is being clear what it is.
Participant 8 I think that sometimes yes.
Participant 9 I think it depends on situation.
To some interviewees, the explanation of the reason
„why? and „how? term and calculation were
developed is considered as necessary. Whilst other
participants viewed them as needed only in some
cases, but it is more important to have a clear
description - what it is. Furthermore after these
questions some participants were asked to categorise
all situations when they think metadata is required.
Below are the findings.
Participant 1
From start we need metadata on the report which describes what it is all about. We also
need metadata on the columns again including the scope, the business rules attached to
them. I did not think about metadata on the row level, we probably do not. We also need
metadata on the actual source systems and source tables. And most importantly we need
metadata for the links between them.
Participant 2
You could go into technical side more as well but we are mainly looking at the business side.
And then complete business processes where we tend to categorize reports into particular
area like Enrolment or Student Evaluation. I will probably just start with the basic set and
say – let?s do that.
Participant 8
You can overwhelm people with too much information. First thing we do is we start with
small amount of information. And then you go to the people and realize that they don?t use
something because the description is not informative enough which need to be improved. I
do not want to go too much first. I would rather go a little and built rather than too much
and people will not use it.
It is observed that from the early stage, the BI team
was more focused on implementing metadata for
business users because the satisfaction of end users
is the priority. This answer also suggests that to
implement metadata, it is advisable to apply several
iterations starting from business metadata and to
finish whole metadata implementation process with
that chosen part of metadata.
4.0 Developing a Metadata Solution on Object
and Element Level
The result of the data analysis indicates that the
metadata solution ought to consist only of metadata
for business users due to the following two main
reasons. First, the interview results clearly showed
that the majority of participants considered business
users as important, and they were more likely to
provide detailed elements of metadata for the
business users? BI environment. Although the
remaining metadata is useful for various purposes it
is not as crucial as metadata for business people.
This proves that business orientation is paramount in
case organisation. Second reason is that the existing
metadata in the university?s BI environment was
taken as the basis of the future metadata model. The
existing metadata model consists of two levels which
relates to end user?s environment – reporting objects
such as cubes, reports and elements such as terms
and columns used in cubes and reports. Metadata on
the level of objects was implemented as additional
description for each element in Cognos Upfront
environment. Metadata on the level of elements was
implemented as ASP page („Glossary? application)
where all the terms used in cubes and reports are
presented. These two levels are both related to
business metadata – metadata that is important and
used by business users. As a result these two parts of
metadata model were enhanced in the first place.
Figure 3 depicts the proposed metadata model on
object level. The enhanced object metadata level has
a number of new metadata fields which were
mentioned during the interviews. „Type? metadata
defines whether the object is report or cube. This
metadata will allow using this kind of classification
for search, sorting and presentation purposes.
„Scope? metadata defines what is included, what is
excluded from the data, for instance, a report
Workshop on Advances in Intelligent Computing 2009
49
contains all students or only domestic students.
„Source systems? metadata shows from what source
applications or external sources the data was taken.
„Primary audience? metadata describes who
supposed to be the key users of the reporting object.
„History? metadata stores previous name of the
object or some historical comments. „Report run
date? metadata shows when the report was run and
relevant only for the report (not relevant for cube).
Current object metadata consists of 5 metadata
fields: „Name of the object?, „Description?,
„Contact/author?, „Refreshed? and „Modified?. The
information in description metadata field can be
divided into 4 separate metadata fields –
„Description? metadata that provides general
overview of the object, „Time period? metadata that
explains what time period is presented in the data,
„Usage? metadata that discusses purpose of the
reporting object, how and for what it should be used
and „Notes? metadata that presents critical notes
regarding the object. „Contact/author? metadata field
shows the main contact person and the author (or
developer) of the report. For consistency this
metadata field was divided into two metadata –
„Contact person? and „Report designer?. „Report
designer? metadata is also relevant only for the
report. „Refreshed? metadata field shows refresh
frequency of the reporting object and has been
renamed to „Refresh frequency?. „Modified?
metadata field shows the date of the last data refresh
and also has been renamed as „Data Refresh date? of
Name of the object
Type of the object
Description
Time period
Scope
Usage
Primary audience
Critical notes
Source systems
History
Contact person
Report designer
Data Refresh date
Refresh frequency
Report run date
Fig 3. Proposed object metadata level
The proposed metadata model on element level has 5
new metadata fields: „Business acronym?, „Places of
use?, „History?, „Owner? and „Refresh date?.
„Business acronym? shows the business acronym for
the element if it exists. „Places of use? metadata field
shows where the element is commonly used across
the BI environment. „History? metadata stores
previous name of the element or some historical
comments. „Owner? metadata file represents the
owner of that particular field, it can be the same as in
„Contact person? for the parent object but it also can
be different. „Refresh date? shows the date when that
element was updated last time.
Current element metadata level consists of 4
metadata fields: „Name of the element?, „Primary
system?, „Comments? and „Type?. „Primary system?
metadata field indicates where the term originated
from. „Comments? field describes what the element
mean. „Type? indicates is the term used as a measure
or dimension. „Comments? field has been renamed to
„Description?. „Description? field should also include
business rule that was applied to the element in
business terms. The completed element metadata
level is shown in figure 4.
Name of the element
Business acronym
Primary system
Description
Type
Places of use
History
Owner
Refresh date
Fig 4. Proposed element metadata level
5.0 Implementation Issues
The implementation process showed that, while
some part of the metadata can be extracted from the
current BI environment, the remaining part is not
stored anywhere so support by the business users is
required for that part of the metadata. Moreover, the
relations between object metadata level and element
metadata level should be fully automated because
the corresponding changes would be very often
within the BI environment. Thus, these relations
should be extracted from the BI environment which
requires more detailed work with the BI
environment.
The main problem relating to the extraction of
metadata from the BI environment is that the BI
environment is not designed for the extraction of the
extensive metadata. That is why this process can be
time consuming or not achieved in full. Another
issue for the metadata implementation is the lack of
functionality in the BI environment that would
allow integrating seamlessly the metadata
application within the BI environment. The
integration mechanism should also transmit the
current context from the BI environment to the
metadata application. The good integration with the
Workshop on Advances in Intelligent Computing 2009
50
BI environment means that the business user would
be able to get the metadata from within the BI
environment.
As was discovered during the implementation
process these two issues - extraction of metadata
from the BI environment, and integration with the
BI environment – are the key requirements for the
successful metadata implementation and the most
challenging ones. Thus, the ability of the BI
environment to support them should be investigated
at the early stages.
6.0 Concluding Comments
The research findings showed that there is a lack of
rigorous metadata implementation in BI
environment of the case organisation. More
importantly, there is a common understanding on
the necessity of business-oriented metadata among
organisational employees with different roles. In
general, most participants agreed that they need
metadata in BI environment for various situations.
The most important issue was related to the
metadata for business users which would allow
them to understand the data and to be sure that it is
exactly what they want. Also it was mentioned that
the university needs a single, consistent and
independent metadata repository. It will provide the
flexibility for technical users to offer more powerful
solution in the near future. This study explores the
metadata needs of the university and substantiates
that the current BI environment requires a
comprehensive metadata implementation from
business and cross-functional enterprise perspective.
The result of the interview analysis demonstrated
that the following reasons drive the needs of
metadata in BI environment:
? To provide consistency for descriptions and
definitions of the data in BI environment;
? To provide an overall enterprise view;
? To solve the problem of misinterpretations of
some terms which could have different
meanings for staff with different roles; and
? To provide the translation between technical
and business terms.
The research also reaffirms the importance of
metadata for business users. Many technical
metadata elements are not comparatively critical for
the metadata implementation because they are not
that useful for business users. The level of
satisfaction by the business users is the main
indicator of success of a BI environment and,
therefore, metadata implementation. In the next
phase of the action research, the prototype metadata
solution will be implemented in a controlled
environment to gather the feedback of general users.
6.0 References
[1] Benbasat, I., D. Goldstein, et al. "The Case
Research Strategy in Studies of Information Systems,"
MIS Quarterly (11:3), 1987, pp. 369-386.
[2] Beyer, M. A. Why Metadata Matters to Business
Intelligence Initiatives, 2007, retrieved May 1, 2009,
from:http://www.gartner.com/DisplayDocument?doc_cd=14
4814&ref=g_fromdoc
[3] Gartner. Gartner EXP Survey of More than 1,400
CIOs Shows CIOs Must Create Leverage to Remain
Relevant to the Business 2007. Retrieved May 1,
2009, from:http://www.gartner.com/it/page.jsp?id=501189
[4] Gartner. Gartner EXP Worldwide Survey of 1,500
CIOs Shows 85 Percent of CIOs Expect "Significant
Change" Over Next Three Years 2008 Retrieved May
1, 2009, from:http://www.gartner.com/it/page.jsp?id=587309
[5] Gartner. Gartner EXP Worldwide Survey of More
than 1,500 CIOs Shows IT Spending to Be Flat in
2009." Retrieved May 1, 2009, from:http://www.gartner.com/it/page.jsp?id=855612
[6] Hancock, J. and Toren. R. Practical Business
Intelligence with SQL Server 2005, Addison Wesley
Professional, 2006
[7] Hess, T. J. and J. D. Wells. "Understanding how
metadata and explanations can better support data
warehousing and related decision support systems: an
exploratory case study," 35th Annual Hawaii
International Conference on System Sciences, 2002.
[8] Inmon, W., B. O'Neil, et al. Business Metadata,
Capturing Enterprise Knowledge, Elsevier. 2008.
[9] Lawton, G. "Making Business Intelligence More
Useful," Computer (39), IEEE Computer Society.
2006, pp. 14-16.
[10] McMurray, A., W. Pace, et al. Research: a
commonsense approach, Thomson/Social Science
Press, 2004, pp. 260-271.
[11] Richardson, J., K. Schlegel, et al. Magic Quadrant
for Business Intelligence Platforms, Gartner, 2008
[12] Schlegel, K. and Rayner. N. Key Issues for
Business Intelligence and Performance Management
Initiatives, Gartner, 2009
[13] Sheina, M. What went wrong with business
intelligence? Retrieved April 4, 2009 from:http://www.cbronline.com/article_cbr.asp?guid=BE8B
E7D0-CD23-48CC-9FD4-42EF5486A846
[14] Sherman, R. "Business Intelligence Goes Back to
the Future, Part 2: Couples Therapy for IT and
Business Users." Information Management Online,
2005
[15] Tvrdíková, M. "Support of Decision Making by
Business Intelligence Tool," 6th International
Conference on Computer Information Systems and
Industrial Management Applications (CISIM'07),
IEEE, 2007
[16] Xu, L., L. Zeng, et al. "Research on Business
Intelligence in enterprise computing environment,"
IEEE International Conference on Systems,
Management and Cybernetics, ISIC, Montreal,
Canada, 2007.
Workshop on Advances in Intelligent Computing 2009
51
doc_618026998.pdf
The ultimate goal of Business Intelligence (BI) initiative is to deliver data in the most appropriate way for making informed business decisions.
An Empirical Research into Metadata of Business Intelligence Systems
Yuriy Verbitskiy, University of South Australia, Adelaide, Australia, [email protected]
William Yeoh, University Tunku Abdul Rahman, Perak, Malaysia, [email protected]
Andy Koronios, University of South Australia, Adelaide, Australia, [email protected]
Abstract
The ultimate goal of Business Intelligence (BI)
initiative is to deliver data in the most appropriate
way for making informed business decisions. To
attain this goal, metadata for end users plays a
critical role in understanding the data provided by a
BI system. However, there has been little empirical
research about the metadata issues impacting the
implementation of BI systems. In particular, such
critical questions as „why does BI need metadata??
and „What metadata does BI need?? have not been
investigated in depth. Answers to these questions
are explored in an action research undertaken in a
large university in Australia. The findings of semi-
structured interviews show that success of a BI
system relies more on the metadata for business
users than on technical metadata. As a result, a
business-users-oriented metadata solution on object
and element level has been developed to improve
the users? satisfaction. The main issues that were
found during the implementation process are also
discussed. The research findings in conjunction with
the conclusion of the study may help BI
stakeholders in identifying the critical metadata
needs required for a more optimised BI deployment.
It also may assist BI vendors in improving the
business-side metadata which are crucial for a
successful BI endeavour.
Keywords: Business Intelligence, Metadata,
Object, Element, Glossary, Action Research
1.0 Introduction
Recently Business Intelligence (BI) applications
have been dominating the technology priority list of
many CIOs [3, 4, 5]. Gartner Research predicts that
the BI market will be in strong growth till 2011
[11]. According to Hancock and Toren [6],
“Business Intelligence is a set of concepts, methods,
and technologies for turning separated data in an
organization into useful information in order to
improve business performance”. The Main element
of a typical BI environment is integration process,
which means moving data from different sources
into one integrated place, storing the data, analysing
data and presenting the data to the end users. In
other words, BI is a system that allows business
users to leverage the data for making informed
business decisions[9].
Being a powerful approach for working with diverse
data within a large organization, the BI system is a
complex and resourceful undertaking. For instance,
Gartner formulated 12 key technical capabilities that
should be delivered by the BI environment and
advocates that the BI vendors implement at least 8
of these 12 capabilities [11]. Nevertheless, the main
indicator of success in implementing BI system is
the level of end user satisfaction. Understanding of
the complex BI environment by business users is
critical issue because “a lack of skills in using
information, tools and applications presents a big
barrier to the success of BI” [12]. The difficulties in
understanding BI are associated with inadequate
training [13] and disconnect between business and
technical users [14]. Focus on the technological
issues without balancing with business orientation is
a frequent reason for unsuccessful outcome [13].
Another problem is the understanding of the data
that business users work with in the BI environment.
Allowing end users to handle sets of data from
different data sources creates a problem because
most users are not familiar with those data. On one
hand, BI is a technique that allows the delivery of
required data from various sources to the end users.
On the other hand, business users have problems
with understanding the BI environment and the data
provided by BI applications. Without solving this
issue BI will not help users to make decisions
effectively. That is, the understanding of the data by
business users is a pressing issue in BI environment.
Making decisions based on the results of BI tools is
not surprisingly the biggest challenge for end users
[9, 16].
In response to this, metadata serves as a mechanism
that provides the context about the data and
information of a BI application [15]. It addresses the
how, when, why and what questions in a technical
environment [7]. Without metadata the data in an
enterprise cannot be understood properly [8]. To
date various BI tools allow the gathering and display
of data to users, which was not possible before.
However, the complexity of the technical
environment increases constantly with the number
and the diversity of BI applications. So the problem
of data misunderstanding within a BI environment is
getting even more critical. Using Metadata allows
us to solve this problem. Gartner Research argues
that metadata management is one of the most
important functionalities that BI environment should
deliver [11].
Therefore, the purpose of this research is to identify
why BI needs metadata and to investigate what
metadata BI requires. Subsequently a metadata
solution will be developed to address the research
findings. Most existing BI literature does not focus
on practical reasons why BI needs metadata.
Workshop on Advances in Intelligent Computing 2009
43
Fig 1. Areas of BI environment for this research
There are also several studies that discuss the
different types of metadata, but none has addressed
the actual metadata required for systems
implementation. This research is going to overcome
the problem by exploring the why and the elements
of metadata that are required through an action
research in a large organisation in Australia. The
next section discusses the research approach before
elaborating on data collection method and case
background. The third section presents and
discusses the findings of the study. Then follows the
conclusion and proposal for further research.
2.0 Research Approach
Despite the daunting complexities in implementing
BI systems, there has been little empirical research
about the metadata specifically for BI environment.
This study investigates the issues of metadata
implementation in the BI environment of a large
university in Australia. It identifies the specific
metadata needs of the organisation, and
subsequently proposes and implements a solution by
developing a prototype metadata solution. When a
process of change is the subject of research, the
most appropriate methodology is action research,
which was deemed to be the most suitable approach
for this project [1].
The research was conducted in collaboration with
the Australian university?s BI team. Having more
than thirty thousand students, it is crucial that the
institution exploit a powerful data analysis and
performance management tool. That is why the
university acknowledges the importance of the BI
technology and strives to deliver it to the wider
organisational community. It also admits the fact
that it does not have a consistent metadata
implementation which benefits business users of BI
environment. These factors make the university an
ideal environment for investigating the metadata
issues in BI environment. This paper presents one of
the key elements of the action research – diagnosing
the environment to define the problem - with the
following main research questions: Why does BI
need metadata implementation? And what metadata
does BI require?
2.1 Data Collection
To answer these two questions, several semi-
structured interviews with the university?s BI
stakeholders have been conducted. The interview
began with a predefined list of questions. Hence it
allows some latitude in getting detailed information
from interviewees during the interview, but at the
same time to have general questions so that all
participants are treated in a consistent manner [10].
The interview questions can be divided into two
main parts. The initial part of general questions
focus on discovering various situations where
metadata can be needed. Whilst second part relates
to more concrete questions which try to determine
the exact elements of metadata users would like to
have. To investigate the second research questions
and to explore the required metadata elements, the
BI environment under study was divided into 5
logical areas. As shown in figure 1, these areas are:
Data Sources, ETL process, Data Warehouse,
Business Rules and End user?s environment.
The interview was conducted with 9 research
participants from different departments and with
different roles in BI environment of the university: 6
from BI team, 2 are business analysts, and 1
business user. In general, there are four main roles
in the BI environment: business users, business
analysts, members of BI team and technical users.
Different roles reflect the different ability and
responsibility in implementing an enterprise-scale
BI system. Business users can only view the
reporting objects, such as cubes and reports.
Business analysts can produce and view the
reporting objects. BI team is responsible for data
modelling in BI, administration of data warehouse,
data quality, implementation of ETL and business
rules. In addition, the BI team develop complex
cubes and reports on request from business analysts
or business users. Technical users are responsible
for providing and maintaining the technical
infrastructure of the BI environment, such as data
source and data warehouse. Moreover, the different
roles also mean that they require different metadata
for their work. For example, business users do not
wish to see any technical descriptions, they only
want to understand and trust the data that they see in
a reporting object. At the meantime technical users
need technical information in performing their
work. All these differences were taken into account
during the interviews and analysis process. It is
important to highlight that business analysts are in
fact business users with additional skills in
manipulating the data. Business analysts are usually
the first point of contact for business users in
relation to complaints and requests for change.
Therefore, business analysts of the organisation
Workshop on Advances in Intelligent Computing 2009
44
could understand in-depth problems and needs of
various business users.
2.2 The Metadata Status of Case Organisation
At the time of the study, the organisation was
migrating to Cognos BI environment version 8.4
from the previous Cognos version 7. With the new
version of BI environment some novel possibilities
for metadata implementation appeared. Also the
university was changing its way of using the BI
environment. Before this front end BI environment
was encapsulated only in Cognos Upfront
application that showed cubes, reports and related
documents for users. Nonetheless, new approach
allows the delivering of BI objects into Staff portal
so BI will be embedded into the general user?s
environment. This new approach of delivering BI to
end users should be also considered during the
metadata implementation process.
Within Cognos version 7 BI environment, some
metadata techniques are in place but they are in
immature condition. Technical users use Excel
spreadsheets with links between technical and
business objects in the BI environment. Business
users exploit structured descriptions for cubes and
reports in Cognos Upfront environment. Also, they
are able to use „Glossary? application which
represents an ASP page with a list of terms and
descriptions, as illustrated in figure 2. While this
metadata provide some helpful information for the
business users, the BI team realises that they require
a more consistent, attractive and user-friendly
solution for metadata implementation. The
following section presents and discusses the
research findings.
Fig 2. Metadata in „Glossary? application
3.0 Research Findings and Discussions
One of the main questions of this research – whether
there is a need for metadata in BI environment –
was transformed to an interview question. Research
participants were asked to describe the situations
where he or she thinks that metadata would help
users of BI environment. The following section
presents the findings.
Q1: Can you tell me about the occasions when you
thought that metadata or additional description
would help users or yourself to understand data
better?
Participant 1
I suppose the main reason we want metadata is for end users looking at reports to be able to
understand where these numbers come from and what processing has done to them
Participant 2
I just think anywhere where people use reports.
Participant 3
I think probably in all instances while I publish reports we are trying to put as metadata
around the report as we can
Participant 4
But again it comes to interpretations of object and what does it actually represent. That is
where we need some sort of clarity and consistency.
Workshop on Advances in Intelligent Computing 2009
45
Participant 5
For developers understanding of the data is fine. For users it could be helpful. To clarify the
rules of presentation in the business tools, the scope of the data, to resolve different
definitions e.g. definition of mode for course is different from definition of mode for
program.
Participant 6
So for the people who use this dimension or column, they need some description – what
exactly this, how it was derived.
Participant 7
In general it is about data what you are looking at… I think to have some proms around that
that says “this is what you are looking at”.
Participant 8
Knowing where the information comes from, what the information represents
Participant 9
First of all, to help people to understand what they actually looking at
The answers to this question showed that different
staff with different roles almost identically
understands the importance and role of metadata.
The most common opinion regarding metadata was
that metadata should explain data on the reporting
object for business users. For this metadata should
show where the data comes from and the data
processing. Some participants from BI team said
that metadata will also help with testing, impact
analysis and documentation.
The next question relates to the finding that BI
efforts generally focus only on a limited set of data
points for semantic reconciliation, and therefore
miss the overall enterprise view [2]. Beyer said that
it is vital to develop a wider metadata model which
will be applied to the whole organisation in the
future rather than trying to connect differing data
using “tunnelling” approach.
Q2: What about the problem of seeing the overall
enterprise view?
Participant 1
Absolutely. When you look at the number and get more information of that. That needs to be
linked back to the source.
Participant 2
Yes, certainly that would be ideal – the enterprise view, which means that you have one
central location where thinks are defined than they linked to anywhere where they are used.
We don?t have a lot of that. That is part of our problem.
Participant 4
That would be useful, definitely. Probably far more from the developer perspective and
possibly from the user perspective.
Participant 5
That clarifies whole implementation process. It would be more helpful for BI developers.
Participant 6
Yes. I think there is problem.
Participant 7
I think there are two parts of it from end user point of view I don?t think it matters as long as
they got what they look for whereas these links are more important for technical roles, for
people who are building reports, building cubes where they are clear about the extract of
the data particularly when you have multiple systems.
Participant 8
I don?t think we have enterprise view of the information in the university at all… It is
organizational issue and cultural issue that we need to overcome.
Participant 9
I don?t think it is a problem. I think it is difficult in an organization which as complex as this
one.
According to the majority of the interviewees, they
do not have the overall enterprise view and so it
would be useful to have it. Nonetheless, some of the
arguments for overall enterprise view were related
to more technical side. Participants generally
believed that it would be useful for technical and
business analytics roles. The next question is about
understanding data or any elements in the BI
environment.
Q3: Are there problems with understanding data or
any elements in the environment?
Participant 1
Well, yes. That example with SATAC (reports) … What we do. We manually put the
metadata on the report… But often people will not see that.
Participant 2
Yes, I think so and I can give you classic example … o it is that basic that people in higher
and all over the uni, they don?t really understand the data because there is no place where
they can go to look up for consistent definition. I think it would help.
Participant 3
I think it is varying decrease of knowledge and understanding through all the levels.
Participant 4
Probably not. I think there is confusion about it. Most of the users understand the context
around it but if you take some infrequent users who are not necessarily know what the
difference is.
Workshop on Advances in Intelligent Computing 2009
46
Participant 5
Yes. The user should understand the business environment that they are dealing with.
Business users have different understanding of the data.
Participant 6
Yes, off course... you need a common description for each attribute or column or whatever.
Participant 7
I think there are two needs not only for business analysts who works with data but also for
someone who is not business analysts but uses the data he needs interpretation.
Participant 8
That is a huge barrier in trying to identify, especially from the metadata point of view, how
you define the information if it is used across the university with the same terminology but
the definitions are completely different.
Participant 9
Absolutely, yes. That is one of the things that we do all the time.
Almost all participants said that it is one of the key
problems that they deal with or want to solve. The
problem appears due to technical limitations,
complex business requirements in university,
organisational structure of the university and
because the consistent metadata implementation has
not been done yet. To continue the interview in the
area of data understanding, the next question
provided to participants an additional possible
ground why the data understanding could be
difficult.
Q4: Are there problems with interpretation of some
terms which could have different meaning for staff
members with different roles?
Participant 1 Absolutely. The way we get around that is again labelling the report with metadata and a
column but again it is manual metadata what ideally should be automated
Participant 2 Yes, definitely. Things like that actually can make quite a difference to the business
decisions that have been made at the uni ...
Participant 3 That is another big problem and it always go hand in hand with the question before –
understanding what the data actually is and what is used for.
Participant 4 Yes. People want to see things from the different perspective.
Participant 5 Yes, I think there are problems.
Participant 6 Yes.
Participant 7 Someone who uses their own system in their environment still needs to understand that that
terminology is for their group and when it is different system it could be different.
Participant 9 Yes.
All participants agreed that this is a problem and
that everybody sees the data from his/her
perspective. The problem is not only between
different departments within the organisation which
could interpret the same term differently but also
regarding the collaboration between the firm and
external organisations, such as government
agencies. The problem of translation between
business area and technical area is usually very
common. Next question has a goal to find out such
problems.
Q5: Do you recall any problems relating to a lack of
translation between technical terms and business
terms?
Participant 1
Yes … that is why we have that huge mapping spreadsheet which maps all of the business
descriptions, all the business names to be actually technical names.
Participant 2
Yes, probably… We have not standardized basically on one, and say that is what we mean
when we say that.
Participant 3
I have not encountered any by myself.
Participant 4
Users really don?t know what they want until they see it … Developers also need to
understand the business and how they use it. So you do get the mismatch between these two.
Participant 5
I am not sure about it.
Participant 7
The business person knows the business processes so it is to provide right information to
technical person because technical person quite often will not know business processes.
Participant 8
Every single day… So translation of something stored in a technical sense to a business
sense is difficult to understand.
Participant 9
You have to be very careful whenever you use language what particularly language you
using and who is going to read it. Certainly in terms of writing reports we write them very
differently depending on for what audience we are writing.
Workshop on Advances in Intelligent Computing 2009
47
The majority of the participants said that translation
between technical and business terms is a problem
that they see during their work. But as one
participant said it is an issue particularly for
technical people and business analysts who try to
implement business tasks in technical environment.
So for them translation between technical terms and
business terms is critical whereas for business
people it is not so important because they usually do
not need to understand technical side at all.
Q6: What about descriptions for calculations?
Participant 1
We call it business rules… Ideally more automated would be better. I can't think of anything
which would make more useful than links.
Participant 2
We don?t probably have a lot of them documented at the moment. Report writer could go
and act on this and see what calculation is... we have not actually, you know, put that out. It
is not available to anyone except the people who are actually working on that project at the
moment.
Participant 3
I can?t think about the cases where we do calculations as standard. If we use it in the report
we always put appropriate labels and description of the calculations.
Participant 4
We might have some for DW. I think they are useful from the technical perspective. But also
from the users perspective to see how is something actually derived.
Participant 5 For me the descriptions are okay. Maybe for users there are problems.
Participant 6 Yes, there is need for that.
Participant 7
Most of my usages in a cubes or SATAC cubes so it is more about totals. But clearly where
there is need for other measures these formulas need to be standard and clear.
Participant 8
From a business point of view the user don?t understand or don?t really need understand the
technical stuff, they just have to have the confidence. That kind of definition, business
definition should be the first that the user sees. And under that business definition it is
technical information.
Participant 9 To be honest I did not see that many, I am not sure.
It is noted that the participants from BI team require
detailed calculations to be shown because they are
responsible for that. However, there is no such
pressing need for business users to see a full
description of the calculation.
Q7: What about descriptions for aggregated metrics
that are used in different reports, for example in
quarterly or annual reports?
Participant 1
We don't do many aggregates, we just started. We don't say when they are the same as the
source. If they are not the same as the source than yes we need to be able to report that.
Descriptions in that aggregate so far have been assigned as actual source because we don't
introduce any new.
Participant 2
It is basically a calculation but we have come across the problems with the reporting where
it is not feasible to aggregate a certain measure
Participant 3
The aggregation tends to happen by default anyway. Others aggregations tend to be quite
straightforward e.g. by division or program.
Participant 4
I would say it could be a problem. Personally I have not experienced a lot because I have
not deal with that sort of information.
Participant 6 Yes, what I said is perfectly valued for aggregated metrics.
Participant 8
Aggregation is critical. But aggregation does not mean the sum of everything underneath
e.g. student count. So we use count them uniquely for the grouping of programs or classes. A
lot of people get confused with that example.
Participant 9
We use a lot of aggregated data but you need to be very clear what it is involved otherwise
people can misinterpret it. And the other thing I found – people will take some data from one
report and apply it into another report what is not applicable in that situation.
Most participants agreed that there could be complex
aggregations that need further explanations but such
cases are very infrequent. Usually aggregations are
straightforward and do not require explanations.
Q8: Is it worth to include the descriptions of reasons
why a term, calculation or metric was developed?
Workshop on Advances in Intelligent Computing 2009
48
Participant 1 Absolutely. And not just reasons but who authorised that.
Participant 2
You need to have things like the scope. Because often you will filter out certain pieces of
data for, I don?t how many reasons, depending on who is using it, what the rule is.
Participant 3
If you looking at the data I expect to see definitions and descriptions so I could know what
is happening. If I want to know why I would always think you will go to another place.
Participant 4
I think any information you give is worthwhile because it gives you picture of how things
work together.
Participant 5 In some cases - yes you probably need to, but in major cases you do not need to.
Participant 6 Yes, you need that.
Participant 7 I think the more important aspect, once it has been defined, is being clear what it is.
Participant 8 I think that sometimes yes.
Participant 9 I think it depends on situation.
To some interviewees, the explanation of the reason
„why? and „how? term and calculation were
developed is considered as necessary. Whilst other
participants viewed them as needed only in some
cases, but it is more important to have a clear
description - what it is. Furthermore after these
questions some participants were asked to categorise
all situations when they think metadata is required.
Below are the findings.
Participant 1
From start we need metadata on the report which describes what it is all about. We also
need metadata on the columns again including the scope, the business rules attached to
them. I did not think about metadata on the row level, we probably do not. We also need
metadata on the actual source systems and source tables. And most importantly we need
metadata for the links between them.
Participant 2
You could go into technical side more as well but we are mainly looking at the business side.
And then complete business processes where we tend to categorize reports into particular
area like Enrolment or Student Evaluation. I will probably just start with the basic set and
say – let?s do that.
Participant 8
You can overwhelm people with too much information. First thing we do is we start with
small amount of information. And then you go to the people and realize that they don?t use
something because the description is not informative enough which need to be improved. I
do not want to go too much first. I would rather go a little and built rather than too much
and people will not use it.
It is observed that from the early stage, the BI team
was more focused on implementing metadata for
business users because the satisfaction of end users
is the priority. This answer also suggests that to
implement metadata, it is advisable to apply several
iterations starting from business metadata and to
finish whole metadata implementation process with
that chosen part of metadata.
4.0 Developing a Metadata Solution on Object
and Element Level
The result of the data analysis indicates that the
metadata solution ought to consist only of metadata
for business users due to the following two main
reasons. First, the interview results clearly showed
that the majority of participants considered business
users as important, and they were more likely to
provide detailed elements of metadata for the
business users? BI environment. Although the
remaining metadata is useful for various purposes it
is not as crucial as metadata for business people.
This proves that business orientation is paramount in
case organisation. Second reason is that the existing
metadata in the university?s BI environment was
taken as the basis of the future metadata model. The
existing metadata model consists of two levels which
relates to end user?s environment – reporting objects
such as cubes, reports and elements such as terms
and columns used in cubes and reports. Metadata on
the level of objects was implemented as additional
description for each element in Cognos Upfront
environment. Metadata on the level of elements was
implemented as ASP page („Glossary? application)
where all the terms used in cubes and reports are
presented. These two levels are both related to
business metadata – metadata that is important and
used by business users. As a result these two parts of
metadata model were enhanced in the first place.
Figure 3 depicts the proposed metadata model on
object level. The enhanced object metadata level has
a number of new metadata fields which were
mentioned during the interviews. „Type? metadata
defines whether the object is report or cube. This
metadata will allow using this kind of classification
for search, sorting and presentation purposes.
„Scope? metadata defines what is included, what is
excluded from the data, for instance, a report
Workshop on Advances in Intelligent Computing 2009
49
contains all students or only domestic students.
„Source systems? metadata shows from what source
applications or external sources the data was taken.
„Primary audience? metadata describes who
supposed to be the key users of the reporting object.
„History? metadata stores previous name of the
object or some historical comments. „Report run
date? metadata shows when the report was run and
relevant only for the report (not relevant for cube).
Current object metadata consists of 5 metadata
fields: „Name of the object?, „Description?,
„Contact/author?, „Refreshed? and „Modified?. The
information in description metadata field can be
divided into 4 separate metadata fields –
„Description? metadata that provides general
overview of the object, „Time period? metadata that
explains what time period is presented in the data,
„Usage? metadata that discusses purpose of the
reporting object, how and for what it should be used
and „Notes? metadata that presents critical notes
regarding the object. „Contact/author? metadata field
shows the main contact person and the author (or
developer) of the report. For consistency this
metadata field was divided into two metadata –
„Contact person? and „Report designer?. „Report
designer? metadata is also relevant only for the
report. „Refreshed? metadata field shows refresh
frequency of the reporting object and has been
renamed to „Refresh frequency?. „Modified?
metadata field shows the date of the last data refresh
and also has been renamed as „Data Refresh date? of
Name of the object
Type of the object
Description
Time period
Scope
Usage
Primary audience
Critical notes
Source systems
History
Contact person
Report designer
Data Refresh date
Refresh frequency
Report run date
Fig 3. Proposed object metadata level
The proposed metadata model on element level has 5
new metadata fields: „Business acronym?, „Places of
use?, „History?, „Owner? and „Refresh date?.
„Business acronym? shows the business acronym for
the element if it exists. „Places of use? metadata field
shows where the element is commonly used across
the BI environment. „History? metadata stores
previous name of the element or some historical
comments. „Owner? metadata file represents the
owner of that particular field, it can be the same as in
„Contact person? for the parent object but it also can
be different. „Refresh date? shows the date when that
element was updated last time.
Current element metadata level consists of 4
metadata fields: „Name of the element?, „Primary
system?, „Comments? and „Type?. „Primary system?
metadata field indicates where the term originated
from. „Comments? field describes what the element
mean. „Type? indicates is the term used as a measure
or dimension. „Comments? field has been renamed to
„Description?. „Description? field should also include
business rule that was applied to the element in
business terms. The completed element metadata
level is shown in figure 4.
Name of the element
Business acronym
Primary system
Description
Type
Places of use
History
Owner
Refresh date
Fig 4. Proposed element metadata level
5.0 Implementation Issues
The implementation process showed that, while
some part of the metadata can be extracted from the
current BI environment, the remaining part is not
stored anywhere so support by the business users is
required for that part of the metadata. Moreover, the
relations between object metadata level and element
metadata level should be fully automated because
the corresponding changes would be very often
within the BI environment. Thus, these relations
should be extracted from the BI environment which
requires more detailed work with the BI
environment.
The main problem relating to the extraction of
metadata from the BI environment is that the BI
environment is not designed for the extraction of the
extensive metadata. That is why this process can be
time consuming or not achieved in full. Another
issue for the metadata implementation is the lack of
functionality in the BI environment that would
allow integrating seamlessly the metadata
application within the BI environment. The
integration mechanism should also transmit the
current context from the BI environment to the
metadata application. The good integration with the
Workshop on Advances in Intelligent Computing 2009
50
BI environment means that the business user would
be able to get the metadata from within the BI
environment.
As was discovered during the implementation
process these two issues - extraction of metadata
from the BI environment, and integration with the
BI environment – are the key requirements for the
successful metadata implementation and the most
challenging ones. Thus, the ability of the BI
environment to support them should be investigated
at the early stages.
6.0 Concluding Comments
The research findings showed that there is a lack of
rigorous metadata implementation in BI
environment of the case organisation. More
importantly, there is a common understanding on
the necessity of business-oriented metadata among
organisational employees with different roles. In
general, most participants agreed that they need
metadata in BI environment for various situations.
The most important issue was related to the
metadata for business users which would allow
them to understand the data and to be sure that it is
exactly what they want. Also it was mentioned that
the university needs a single, consistent and
independent metadata repository. It will provide the
flexibility for technical users to offer more powerful
solution in the near future. This study explores the
metadata needs of the university and substantiates
that the current BI environment requires a
comprehensive metadata implementation from
business and cross-functional enterprise perspective.
The result of the interview analysis demonstrated
that the following reasons drive the needs of
metadata in BI environment:
? To provide consistency for descriptions and
definitions of the data in BI environment;
? To provide an overall enterprise view;
? To solve the problem of misinterpretations of
some terms which could have different
meanings for staff with different roles; and
? To provide the translation between technical
and business terms.
The research also reaffirms the importance of
metadata for business users. Many technical
metadata elements are not comparatively critical for
the metadata implementation because they are not
that useful for business users. The level of
satisfaction by the business users is the main
indicator of success of a BI environment and,
therefore, metadata implementation. In the next
phase of the action research, the prototype metadata
solution will be implemented in a controlled
environment to gather the feedback of general users.
6.0 References
[1] Benbasat, I., D. Goldstein, et al. "The Case
Research Strategy in Studies of Information Systems,"
MIS Quarterly (11:3), 1987, pp. 369-386.
[2] Beyer, M. A. Why Metadata Matters to Business
Intelligence Initiatives, 2007, retrieved May 1, 2009,
from:http://www.gartner.com/DisplayDocument?doc_cd=14
4814&ref=g_fromdoc
[3] Gartner. Gartner EXP Survey of More than 1,400
CIOs Shows CIOs Must Create Leverage to Remain
Relevant to the Business 2007. Retrieved May 1,
2009, from:http://www.gartner.com/it/page.jsp?id=501189
[4] Gartner. Gartner EXP Worldwide Survey of 1,500
CIOs Shows 85 Percent of CIOs Expect "Significant
Change" Over Next Three Years 2008 Retrieved May
1, 2009, from:http://www.gartner.com/it/page.jsp?id=587309
[5] Gartner. Gartner EXP Worldwide Survey of More
than 1,500 CIOs Shows IT Spending to Be Flat in
2009." Retrieved May 1, 2009, from:http://www.gartner.com/it/page.jsp?id=855612
[6] Hancock, J. and Toren. R. Practical Business
Intelligence with SQL Server 2005, Addison Wesley
Professional, 2006
[7] Hess, T. J. and J. D. Wells. "Understanding how
metadata and explanations can better support data
warehousing and related decision support systems: an
exploratory case study," 35th Annual Hawaii
International Conference on System Sciences, 2002.
[8] Inmon, W., B. O'Neil, et al. Business Metadata,
Capturing Enterprise Knowledge, Elsevier. 2008.
[9] Lawton, G. "Making Business Intelligence More
Useful," Computer (39), IEEE Computer Society.
2006, pp. 14-16.
[10] McMurray, A., W. Pace, et al. Research: a
commonsense approach, Thomson/Social Science
Press, 2004, pp. 260-271.
[11] Richardson, J., K. Schlegel, et al. Magic Quadrant
for Business Intelligence Platforms, Gartner, 2008
[12] Schlegel, K. and Rayner. N. Key Issues for
Business Intelligence and Performance Management
Initiatives, Gartner, 2009
[13] Sheina, M. What went wrong with business
intelligence? Retrieved April 4, 2009 from:http://www.cbronline.com/article_cbr.asp?guid=BE8B
E7D0-CD23-48CC-9FD4-42EF5486A846
[14] Sherman, R. "Business Intelligence Goes Back to
the Future, Part 2: Couples Therapy for IT and
Business Users." Information Management Online,
2005
[15] Tvrdíková, M. "Support of Decision Making by
Business Intelligence Tool," 6th International
Conference on Computer Information Systems and
Industrial Management Applications (CISIM'07),
IEEE, 2007
[16] Xu, L., L. Zeng, et al. "Research on Business
Intelligence in enterprise computing environment,"
IEEE International Conference on Systems,
Management and Cybernetics, ISIC, Montreal,
Canada, 2007.
Workshop on Advances in Intelligent Computing 2009
51
doc_618026998.pdf