Description
The Business Intelligence Model Strategic Modelling
The Business Intelligence Model:
Strategic Modelling
version 1.0
Daniele Barone
1
, John Mylopoulos
1
, Lei Jiang
1
, and Daniel
Amyot
2
1
Department of Computer Science, University of Toronto,
Toronto, ON, Canada
barone/jm/[email protected]
2
SITE, University of Ottawa, Ottawa, ON, Canada
[email protected]
April 14, 2010
Contents
1 Introduction 4
2 Primitive Concepts 8
2.1 Situations, Intentions and Processes . . . . . . . . . . . . . . . . . . . . 8
2.2 Actors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Objects’ Life-cycle and Behavior: A Finite State Machine Model . . . . 19
2.5 Evolution Timeline and Time Constrains for Objects . . . . . . . . . . . 22
3 The BIM Repository and the Abstraction Mechanisms 25
3.1 Aggregation: Mandatory, Optional, OR and Alternative . . . . . . . . . 26
3.2 Classi?cation: The OMG Four-Layer Metamodel Architecture . . . . . . 26
3.3 Generalization: Subtypes and Business Terms Specialization . . . . . . . 30
3.4 An UML Class Diagram for the BIM’s Abstraction Mechanisms . . . . . 30
4 Strategic Analysis through Mappings 34
4.1 Goal (Intention) Reasoning: A Formal (Axiomatic) Model . . . . . . . . 34
4.2 The SWOT Analysis with the BIM . . . . . . . . . . . . . . . . . . . . . 39
4.3 De?ne Strategic Map, Balanced Scorecard and Key Performance Indi-
cators with BIM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.4 Probabilistic Graphical Model for Intention Reasoning . . . . . . . . . . 44
5 A Case Study 46
6 Related Work 48
7 Conclusions 50
8 Appendix 51
8.1 BIM Taxonomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2
Business Intelligence (BI) consists of a range of technologies intended to assist large
organizations in determining the state and quality of their operations. BI activities
are meaningful in the context of a business organization and its objectives, strategies
and tactics, as well as a broader (external) context involving regulations, competitors,
customers, markets, etc. This business context (internal and external) de?nes the
e?ectiveness of business processes, and the things to monitor to ensure that business
objectives are being met and regulations and policies are complied with. The Business
Intelligence Model (BIM) provides a set of constructs for modeling and analyzing a
business context consisting of intentions, situations, processes, actors, in?uences, key
performance indicators, and more. It is intended to support the modeling and analysis
of a business organization at both a strategic and a tactical level. BIM schemas can
be used for governance activities, including analysis, monitoring and auditing. This
report presents some of the main innovations of BIM, including its primitive concepts,
its structuring mechanisms, analysis examples, as well as an overview of an illustrative
case study.
1 Introduction
Business intelligence (BI) consists of a range of technologies for using information
within organizations to ensure compliance to strategic and tactical objectives, as well
as to laws and regulations. As a research ?eld, it encompasses data and knowledge man-
agement, modeling of processes and policies, data quality, data privacy and security,
data cleaning and integration, data exchange, inconsistency management, information
retrieval, data mining, analytics, and decision support.
This interest in technologies and services that improve organizational governance
has caused dramatic growth for the BI market and the industry that serves it. By
now, most competitive organizations have a signi?cant investment in BI, much of it
technology-related, based on software tools and artifacts. However, as summarized by
Gartner [35], one important problem of BI technologies is that information generated
by BI systems and other decision inputs are rarely linked to business decisions and
outcomes. In addition, business people – be they executives, managers, consultants, or
analysts – are in general agreement that what they are looking for is not new gadgets
producing a dizzying array of largely useless statistics. Instead, they are interested
in having their business data analyzed in their terms: strategic objectives, business
models and strategies, business processes, markets, trends and risks. This gap between
the worlds of business and data remains today the greatest barrier to the adoption of
BI technology, and the greatest factor in the cost of applying BI technology.
We propose to bridge this gap by extending the notion of conceptual schema to
include concepts beyond entities and relationships. In particular, we propose the Busi-
ness Intelligence Model (BIM) as a business-level counterpart to the Entity-Relationship
or Relational Model, so that strategic objectives, business processes, risks and trends
can all be represented in a business schema. Users can query this schema, much
like conventional database schemas but with business terms, to perform analysis, to
track decisions and their impacts, or to explore suitable strategies to problems at
hand. Such queries are to be translated through schema mappings into queries de?ned
over databases and data warehouses, and the answers are to be translated back into
business-level concepts.
The main objective of this report is to introduce BIM’s constructs for modeling
business organizations at a strategic level. In particular, we present a set of primitive
concepts consisting of actors, intentions (e.g., goals), situations (strengths / weaknesses
/ opportunities / threats, a.k.a. SWOT), in?uences, processes, key performance indi-
cators and more. These concepts can be used in tandem with abstraction mechanisms,
such as generalization, aggregation and classi?cation, to develop global models of busi-
ness organizations for purposes of analysis, monitoring and auditing.
Our work is founded on modeling techniques from diverse sources. Abstract con-
cepts for describing all things are inspired from DOLCE[11]. The intentional and so-
4
cial concepts used in BIM are adopted from concepts in Goal-Oriented Requirements
Engineering, notably [7, 42, 18]. The notion of in?uence is adopted from in?uence
diagrams [16], a well-known and accepted decision analysis technique. Concepts re-
lated to SWOT analysis [9] and others have also been adopted from OMG’s Business
Motivation Model standard [3].
As shown in Figure 1.1, the ?nal aim of the BIM is to represent the internal and
external business and environment, and to support managers in making decisions at
each level of management providing answers to questions such as “What will happen
next?”. Indeed, the BIM is based on the idea that you cannot measure what you
cannot represent, and you cannot improve what you cannot measure.
As de?ned in [28, 24] and summarized in Table 1.1, people at di?erent levels in an
organization have di?erent types of decision-making responsibilities.
Strategic decisions, which are typically made by executive managers, a?ect the long-
term direction of an organization and are often complex and characterized by uncer-
tainty due to the limited availability of information. Usually, managers at this level
depend on their past experiences and instincts for making a decision.
Examples of strategic decisions can be to decide whether it is time to discontinue a
product line or to launch a new one.
Tactical decisions regard more intermediate-term issues and are typically made by
middle managers. The decisions made at this level attempt to move the organization
closer to reaching the strategic goals.
Examples of tactical decisions can be to hire an advertising agency in order to
promote a new product or to create an incentive plan for encouraging employees in
increasing the organization’s production.
Operational decisions concentrate on day-to-day organization activities and are typ-
ically made by lower-level managers. Decisions made at this level attempt to ensure
that the daily activities are conform with respect to business targets and standards in
order to achieve the strategic goals.
Examples of operational decisions include scheduling employees, purchasing raw
materials needed for production, or answering questions such as “Do we extend credit
to this customer?”.
Level of
Management
Core
Requirement
Nature of the Decision
Executive Strategic Planning Long term, unstructured, di?-
cult to develop speci?c decision
models
Mid-level Management Control Shorter term, semi-structured,
modeling possible.
Operational Operational Control Short term, structured, model-
ing possible.
Table 1.1: A taxonomy of management decision-making.
In this report, we focus on a particular use of the model for supporting the strate-
gic planning process, which is the process of de?ning an organization’s strategy (or
5
Organi zat on
St ruct ure
Model
Strategi c Model
Tactical Model
Operati onal Model
What i s t he best t hat can happen? (Opt i mi zat i on)
What wi l l happen next ? (Predi ct i ve model i ng)
What i f these trends conti nue? (Forecasti ng/extrapol ati on)
Why i s thi s happeni ng? (Stati sti cal anal ysi s)
What acti ons are needed? (Al erts)
Where exact l y i s t he probl em? (Query/ dri l l down)
How many, how often, where? (Ad hoc reports)
What happened? (St andard report s)
Answers
t o
Busi ness I nt el l i gence Model
Technolgies
Organi zati on
Market
More Intel l i gence
Less Intel l i gence
Figure 1.1: The Business Intelligence Model (“Questions” have been adapted from [8]).
direction) and making decisions on allocating the organization’s resources in order to
pursue this strategy.
In particular, an organization can follow di?erent approaches for strategic planning
such as the Situation-Target-Path, the Draw-See-Think-Plan, the See-Think-Draw, etc.
[36] which can be summarized by the following activities:
1. Formulate Vision & Mission: an organization must clearly de?ne its Vision and
Mission statements and the associated hierarchy of goals and objectives;
2. Situational Analysis: an analysis of the organization and its environment must
be conducted in order to identify in?uences among external and internal factors
and the organization goals, e.g., the SWOT analysis [9].
3. Develop Strategies: a set of alternatives for the ful?llment of goals must be
formulated in terms of actions and processes to be taken to achieve goals and in
terms of the resources required to execute these actions;
4. Implement Strategies: the “best” strategies must be chosen and implemented
using organizational processes and resources;
5. Evaluate and Monitor Performance: the implemented strategies must be evalu-
ated to see if they are working successfully.
In the next sections we present the BIM metamodel and how it can support the
above strategic planning activities.
6
In particular, Section 2 describes BIM’s primitive concepts and their use for strategic
modeling. Section 3 presents the mechanisms (generalization, aggregation, classi?ca-
tion) used to structure BIM models. Section 4 o?ers an overview of how BIM models
can be used to support a range of analyses, such as strategic map, goal analysis and
SWOT analysis. In Section 5, we present an illustrative case study, while Sections 6
and 7 respectively discuss related work and conclusions.
7
2 Primitive Concepts
Table 2.1 introduces the primitive concepts in BIM for modeling strategic objectives
and strategies.
As an ongoing example, we introduce BestTech Inc., a company in the market of cel-
lular phones and home computers
1
. Figure 2.1 shows its Strategic Map (SM) [22] and
Balanced Scorecard (BSC) [21], which are strategic planning instruments commonly
used in industry.
The former (SM) is a visual representation of the strategy of an organization showing
plans used to achieve missions and visions. In particular, it illustrates the cause-and-
e?ect relationships between di?erent strategic goals and associated measures, the key
performance indicators (KPIs) [33].
These measures are included in the latter, the BSC, which represents a “balanced”
range of metrics against which to measure the Organization’s performance. “Balance”
here means that the broader view of leading performance indicators includes also
non-?nancial concerns, such as “learning and growth of employees” and “customer
satisfaction”. Both SMs and BSCs describe and measure organizational performance
across four balanced perspectives: ?nancial, customers, internal business processes,
and learning and growth [22, 21].
The following subsections detail how BIM’s primitive concepts are used to describe
BestTech and its business environment, from a strategic viewpoint.
2.1 Situations, Intentions and Processes
Primitive concepts are represented as metaclasses (having names that end in -Class).
In particular (Figure 2.2
2
), Intention can be used to de?ne a hierarchy of Vision,
Strategic/Tactical goals, etc., which represents the desired end state of an organization.
Process models an organization’s Mission and the di?erent Strategies and, at tactical
and operational levels, their decomposition into Business processes and Activities.
Resource models what resources are required to execute such strategies.
In any strategic planning setting, a scan of the internal and external environment is a
fundamental issue [9]. Accordingly, BIM provides the Situation concept for modeling
internal and external situations that can be helpful or harmful to an organization’s
goals. In addition, we adopt the SWOT classi?cation [9] (see Section 4) in order to
de?ne the types of in?uence that a Situation may exercise on an Intention.
1
Adapted from a generic strategy map for a credit card company provided by the Balanced Scorecard
Institute –http://www.balancedscorecard.org (2010)
2
Dashed inheritance arrows indicate the existence of hidden metaclasses in between.
8
Concept Description Example Superclass
Thing BIM’s most general concept. This abstract meta-
class has everything as an instance.
Object Abstract metaclass whose instances per-
sist/endure over time [11].
Thing
Event Instantaneous happening (perdurant) that
changes an Object; can be described by a
Proposition that was false before the Event and
is true after [11, 15].
New order
received
Thing
Situation Partial state of the world described by a
Proposition. Situations can have structure con-
sisting of relations and Things standing in those
relations [27].
Christmas
season
Object
Proposition Describes a Situation. In general, Propositions
are true/false/unde?ned in a Situation [27].
Object
Intention A Proposition an Actor wants to make true [42]. More
products
sold
Proposition
Domain
Assumption
A Proposition assumed by an Actor to be true
for purposes of ful?lling an Intention.
Market
increases
at least
10%
annually
Proposition
Directive A Proposition prescribed by an authority in-
tended to constraint, guide, govern, or in?uence
elements of an organization such as Actors and
Processes [3].
Pizza de-
livery in
30 mins.
Proposition
Entity Object whose existence does not depend on that
of others.
Object
Actor Entity that carries out Actions to achieve
Intentions [42].
Sales
manager
Entity
Action Entity performed by an Actor that produces
Events and can have pre- and postconditions [7].
Deliver
product
Entity
Process Entity consisting of coordinated Actions to
achieve an Intention [3].
Sales
process
Entity
Resource Entity of value to an Actor [42]. Money,
informa-
tion
Entity
Indicator Measurable Object that gives information about
the state of an associated Object. Can be used
to quantify the satisfaction level of an Intention
or the operational performance of an Actor, a
Process or a Resource [33].
Number
of prod-
ucts sold
per week
Object
Relationship An Object that relates two or more Things and
whose existence depends on that of the Things it
relates [5].
Object
Influence A Relationship between two Things t
1
and t
2
,
where the state of t
1
constraints the state of t
2
in
a probabilistic or causal sense.
Relationship
Table 2.1: BIM primitive concepts.
9
Figure 2.1: A strategic map and a balanced scorecard for BestTech Inc.
Figure 2.2: Intention, Process and Situation metaclasses and associated relation-
ships in the BIM metamodel.
In this classi?cation, internal factors are situations further classi?ed into Strengths
or Weaknesses while external factors are Opportunities or Threats. For example, a
Threat for Wal-Mart may be “Exposure to political problems in the countries that
we operate in”. Notice that a strength with respect to one Intention may well be a
weakness for another.
Figure 2.3 shows a small example that instantiates the metaclasses of Figure 2.2.
Goals (i.e., Intentions), such as “Shareholder value increased” are decomposed into
subgoals. Subgoals may be mandatory or optional (the notation used here is adapted
from feature diagrams [37]). Goals may be achieved by carrying out Processes. In
10
turn, Processes require Resources; there are various types of Resources with as-
sociated icons. Situations can in?uence goals but also Processes, Resources, etc.
Opportunities and weaknesses are particular kinds of in?uences. In?uence relation-
ships
3
can actually exist between any two Things and play a pivotal role in governance
models.
Figure 2.3: An example of Situation, Intention and Process concepts at schema
level.
2.2 Actors
Figure 2.4 shows the Actor type and its relationships. An important relationship is
the “responsible for” that de?nes which Actor is liable to be called on to (legally)
answer when Intentions, Processes, Resources, and Directives are not acting
in accordance with the business of an organization. For example, an Actor can be
responsible for enforcing a work-site safety policy (i.e. a Directive). Indeed, if an
accident occurs due to inappropriate safety conditions or there is a problem during a
safety inspection, the responsibility will fall upon the Actor.
Actors must also comply with Directives, e.g., an employee must wear the pro-
tective equipments, and are able to de?ne them, e.g., the executive board can de?ne
a policy such as “Pizza must be delivered within 20 minutes from the order or it will
be free.”.
3
The instantiation of the EnumerationClass type for the qualitativeStrength attribute can vary
depending on the nature of Things involved in the relationship.
11
An Actor can desire Intentions which are satis?ed by other Actors; for example,
the “More products sold” Intention is desired by the executive board and is satis?ed
by the sales sta? under the responsibility of the sales manager. An Actor is also
capable of a Process (or an Action) that can actually performing or not.
Figure 2.4: The Actor primitive type and the associated relationships.
Figure 2.5 shows an example of Actors and their relationships with the business
environment.
Notice how cardinality values can be used for enriching the semantic of relationships
in order to describe the minimum and maximum number of associated elements within
a set. For example, in the ?gure, the sales department is constituted by a sales manager
and from a minimum of six to twelve sales employees. In particular, regarding the
“Package product” Action, a minimum of three employees are capable of the “Package
product” Action and at least two must perform this Action on a maximum of ten
that can be allocated to such Action.
Moreover, as described in Figure 2.6, an Actor can be specialized in Agent, Role,
and Position. As de?ned in i
?
[41], a Role is an abstract characterization of the
behavior of a social Actor within some specialized context or domain of endeavor, and
a Position is a set of Roles, usually played by one Agent. An Agent is an Actor
with concrete, physical manifestations, such as a human individual or an arti?cial
component of a system (hardware/software agents). Finally, an Agent can occupy a
Position, while a Position covers a Role. Figure 2.6 shows an example of Agent,
Role and Position.
2.3 Indicators
Indicators evaluate the quality of Objects to ensure compliance to internal policies
and external directives.
Figure 2.8(a) shows the Indicator metaclass, consisting of attributes such as target,
thresholds, extreme values, etc.[33]. The attributes ’s description is provided in Table
2.2.
12
Figure 2.5: An example of Actors and their relationships with the business
environment.
Figure 2.6: Agent, Role and Position in the BIM metamodel.
Notice that, in Table 2.2, the evaluationTime represents the timestamp associated
to a speci?c current value instance, i.e., the time when the instance is calculated and
created (see Section 2.5 for more details). This “system” time is di?erent from the
“domain” time (e.g. a Time dimension) used to navigate and calculate Indicators in
a historical period of time, e.g., the Time dimension’s value “March, 2009” to calculate
13
Figure 2.7: An example for Agent, Role and Position at schema level.
(a) (b) (c)
Figure 2.8: (a) The Indicator primitive type in the BIM metamodel. (b) An example
of visual notation for Indicators at instance level. (c) An Indicator ’s
Trend example.
the number of products sold in March.
An example of Indicator and the use of its attributes is shown in Figure 2.8(b).
Notice how the current value (currentValue) is found between Lower Threshold and
Target ; therefore, the company has almost reached its target in selling computers.
The current value is calculated at evaluationTime using the metrics de?ned in the
expression ?eld. In general, BIM relies on the Object Constraint Language (OCL) [31]
14
Meta-attribute’s
name
Description
currentValue the current value of an Indicator which is calculated
through the evaluation of a metric’s expression.
unitOfMeasure the unit of measure associated with the current value of an
Indicator.
expression the metric’s expression used to calculate the current value
of an Indicator.
dimension It allows to navigate (calculate the value of) an Indicator
along speci?c directions of interest.
target It is a value which allows to quantify the satisfaction level of
an Intention (or, more in general, a performance level de-
sired for a property of an Object). If an Indicator reaches
the Target the associated Intention is satis?ed.
(upper & lower)
threshold
It de?nes a threshold (upper or lower) value that can be as-
sumed by an Indicator that represents a critical situation
for the Organization’s business.
(upper & lower)
extremeValue
It de?nes the worst (upper or lower) value that can be as-
sumed by an Indicator which represents a critical (ex-
treme) situation for the Organization’s business.
evaluationTime The time in which the current value of an Indicator is
calculated.
Table 2.2: The description of Indicator’s attributes.
4
to de?ne such expressions. Moreover, one or more dimensions can be used to calculate
and navigate an Indicator along speci?c directions of interest, e.g., the number of
“Apple” computers sold.
BIM also supports the de?nition of operations, such as Trend, Risk, Reward or
Con?dence, at the class level (i.e., at the schema level), which calculate additional in-
formation (meta values) for the current values assumed by an Indicator. The result
is in-depth information on the subject of an Indicator. Table 2.3 provides a brief
description of such methods while Figure 2.8(c) is an example of a Trend operation
based on a linear regression method. The result is a trend with a slope of +0.6 which
means that: i) the trend is positive, and ii) the prediction of computers sold on Dec.
6, 2009 is about four units.
In order to better understand the use of Indicators, take a look at the following
example:
Suppose that a Car Dealer has in store 16 units of “Luxury car” that it desires to
sell over a period of one month (from December 1
st
, 2009 to December 31
st
, 2009).
4
The OCL is a declarative language for describing rules that apply to UML models developed by
the Object Management Group (OMG) and now part of the UML standard.
15
Method’s name Description
trend It represents a general movement over time of a statistically
detectable Indicators’s change.
risk It provides the actual loss associated to the current value
assumed by an Indicator.
reward It provides the actual gain associated to the current value
assumed by an Indicator.
con?dence It provides the actual con?dence associated to the current
value assumed by an Indicator. The con?dence is based
on i) the quality of the information used to calculate the
current value, ii) the reputation of the source who provided
the information and iii) the reliability of the method used
to calculate the current value.
Table 2.3: The description for the Indicator’s operations.
Suppose also that the actual (partial) state of the world, i.e. on December 6
th
,2009,
is as described in table 2.4 in which each car is identi?ed by a Vehicle Identi?cation
Number (VIN).
(Desired) State of the world Evaluation
Car (VIN 1HGBH41J8MN109180) Sold True
Car (VIN 1HGBH41J8MN109181) Sold False
Car (VIN 1HGBH41J8MN109182) Sold True
Car (VIN 1HGBH41J8MN109183) Sold False
Car (VIN 1HGBH41J8MN109184) Sold True
Car (VIN 1HGBH41J8MN109185) Sold False
Car (VIN 1HGBH41J8MN109186) Sold False
Car (VIN 1HGBH41J8MN109187) Sold True
Car (VIN 1HGBH41J8MN109188) Sold False
Car (VIN 1HGBH41J8MN109189) Sold True
Car (VIN 1HGBH41J8MN109190) Sold False
Car (VIN 1HGBH41J8MN109191) Sold True
Car (VIN 1HGBH41J8MN109192) Sold True
Car (VIN 1HGBH41J8MN109193) Sold False
Car (VIN 1HGBH41J8MN109194) Sold False
Car (VIN 1HGBH41J8MN109195) Sold True
Table 2.4: The (desired) states of the world for the “Car sold” Intention.
We have the following Indicator as described in Figure 2.9 and in Table 2.5. Notice
the use of the cardinality (0..16) to express the number of cars the organization desires
to sell. Moreover, Figure 2.10 shows the di?erent zones de?ned by using the extreme,
16
threshold and target values; and the current value for December 6
th
, 2009 lying on the
red zone.
Figure 2.9: An example of Indicators at schema level.
Meta-Level Schema-Level Instance-Level
Indicator Number of Cars Sold Inst 1:Number of Cars Sold
currentValue Integer 8
unitOfMeasure cars sold -
expression “context Car
select (car | car.status=’Sold’ ?
car.type=’Luxury car’ ) ? >
size()”
-
target Integer 13
(Lower)
threshold
Integer 10
(Lower)
extremeValue
Integer 7
dimension TypeOfCar “Luxury Car”
evaluationTime Time Dec 6, 2009
Table 2.5: An example of an Indicator at di?erent levels of modelling.
Figure 2.11 and Figure 2.12 show examples of Indicator’s operations. In particular,
the former is an example of a Trend Operation based on a linear regression
5
method.
The operation calculates the slope (or gradient) of the trend line ?tting the time series
provided in input, where the time series are the number of cars sold from Dec. 1
st
to
2009 to Dec. 6
th
, 2009. The result of the operation is a slope of ?0.7 which means
that on Dec. 6, 2009 the Trend for the number of cars sold is negative.
The latter, Figure 2.12, is an example of risk and reward operations which allow
to associate speci?c loss or gain values to any current value. In fact, if we suppose
that 30,000 USD is the total cost sustained for a car while 50,000 USD is the revenue
obtained by its selling, we have that: i) a total gain of 170,000 USD is obtained when
5
Notice that, di?erent ways for evaluating a Trend exist, such as logarithmic, exponential, etc.; in
this case we use the simple one, namely the linear regression.
17
Figure 2.10: The current value of the “Number of cars sold” Indicator on December,
6
th
.
the target is reached, ii) a loss of 130,000 USD is endured when the current value
assumes the extreme value and iii) a loss of 80,000 USD is endured on the current
value calculated on December 6, 2009.
Figure 2.11: The Trend for the “Number of cars sold” Indicator.
Figure 2.9 shows another Indicator, namely the SWOT Indicator, which is used to
evaluate the in?uences among Situations and Intentions. As described in Section
2.1, we have four kind on in?uences, namely, Strength, Weakness, Threat, and Oppor-
tunity. For each type, it is possible to de?ne a quality or quantity power scale, e.g.,
< high, medium, low > or [0, 100], in order to determine the degree of the in?uence.
Therefore, a SWOT Indicator can assume a current value that ranges among values of
the power scale and can have a unit of measure equal to Strength, Weakness, Threat,
or Opportunity. As described in Section 4, the SWOT Indicator’s current value is used
18
Risk and Reward
(
Y
)
T
h
o
u
s
a
n
d
s
-
U
S
D
0
100
200
300
400
500
600
(X) Number of Cars Sold
0 2 4 6 8 10 12 14 16
Gain
Loss
Threshold = 10
Target = 13
Worst Value = 7
Figure 2.12: The Loss and Gain values for the “Number of cars sold” Indicator.
by the In?uenceClass metaclass to support di?erent kind of analysis.
2.4 Objects’ Life-cycle and Behavior: A Finite State
Machine Model
A Finite State Machine (FSM) [13] provides a simple and e?ective means to control
the life-cycle and overall behavior of BIM’s Objects. In fact, a FSM is an abstract
computational model which allows to de?ne for each Object: i) a set of di?erent states
assumed by Objects’ instances in the real world; ii) the transitions among these
di?erent states in order to de?ne the (computational) behavior; iii) the events/inputs
which express the stimuli taken into account; and, iv) the actions/outputs which are
the possible responses that can be generated.
Formally, a FSM is a multi-tuple FSM = (?, ?, S, s
0
, ?, ?), where:
• (?) is the input alphabet of symbols representing external stimuli (inputs or
events) that are used by transition functions;
• (?) is the output alphabet of symbols representing responses (outputs or ac-
tions) that are provided by output functions;
• (S) is the set of possible states which are conditions of the state machine at a
certain time;
19
• (s
0
? S) is the start state;
• (? : S × ? ? S) is the state transition function. Based on the current state
s
c
? S and an input symbol i ? ?, it computes the transition to the next state
s
n
? S;
• (? : S ×? ? ?) is the output function (as de?ned in the Mealy model ).
As shown in the following subsection, the BIM also allows to associate to each state
a Time attribute which stores a timestamp value representing the last time in which
an Object enter in that speci?c state.
In general, for all Objects, it is possible to identify a START state and a (pos-
sible) END state. The START state allows to de?ne when an instance of the real
world becoming an instance of a speci?c type of the BIM. For example, the “Car
VIN=1HGBH41J8MN109180 sold” instance stars to be an Intention’s instance when
the car dealer de?nes a clear statement to sell that speci?c car, i.e. the car with
VIN=1HGBH41J8MN109180. Than, the car dealer will be able to pursue it, i.e., the
state BEING PURSUED.
In the opposite way, the END state allows to de?ne when the same instance stops to
be an (active) instance of that type
6
. For example, the “Car VIN=1HGBH41J8MN109180
sold” Intention’s instance stops to be an Intention’s instance when the car dealer
stops to pursued it either because it is achieved, failed or aborted (i.e.,the car dealer
has changed his mind).
Moreover, for each speci?c Object, such as the Intention, it is also possible to
de?ne speci?c states which have a particular semantic into the business environment.
For example, the above BEING PURSUED state can be de?ned for Intentions to
express the continuing activity by an Actor to achieve a speci?c Intention.
Figure 2.13 and Figure 2.14 show, respectively, possible FSM diagrams for Intention
and Resource. Notice that, a Vision cannot be ?tted in the diagram illustrated in
Figure 2.13 since, usually, it is not possible to de?ne a SATISFIED state or an END
state for such a concept. Therefore, if it is necessary, a speci?c FSM diagram must
be de?ned or, alternatively, the same FSM diagram can be used in which some states
will be never assumed by the instances.
Moreover, it is also important to notice the two “pass Deadline” and “pass Expir-
ing date” events which, respectively, lead to a FAILED Intention’s state and to an
EXPIRED Resource’s state.
These events are ?red when: i) the current time in the system is greater than, re-
spectively, the “deadline date” and the “expiring date” de?ned in the business environ-
ment (see next subsection); and, ii) the actual states of Intention’s and Resource’s
instances are, respectively, BEING PURSED and BEING CONSUMED. FSMs allows
to de?ne these events as guards which can be expressed by using constrains.
In this way, for example, it is possible to represent situations where a goal’s deadline
has passed and, since it is not longer important whether or not the goal will be satis?ed,
we transit to a FAILED state.
6
Stop to be an “active“ instance means that the instance is still recorded in the system for “history“
purpose (i.e., for analysis and queries) but is no more used for operative tasks.
20
Figure 2.13: A FSM diagram for Intention.
A similar situation exists for Resources in the case we have passed the expiring date.
In fact, the Resource stops to be a valid Resource loosing its intrinsic properties
that make it such as a Resource in the business environment; however, we can still
consume it with all the relative consequences.
The use of states and timestamps enable the de?nition of interesting queries over
Objects belonging to a schema. For example, we can express queries such as (1)=“List
all goals that are not yet satis?ed” or (2)=“Show the time in which
was satis?ed” that can be de?ned as:
(1) = context Intention
select (intention | not(intention.state =
?
SATISFIED
?
)
(2) = context Intention
select (intention | intention.name =
?
All car sold
?
and
intention.state =
?
SATISFIED
?
).time
21
Figure 2.14: A FSM diagram for Resource.
2.5 Evolution Timeline and Time Constrains for Objects
The previous subsection presented the concept of State to address issues related to the
life cycle and behavior of Objects. However, since we want to describe the “evolu-
tion” of Objects within a business environment, we need to introduce the concept of
Evolution Timeline. Figure 2.15 shows the fragment of the BIM metamodel aimed
to describe such aspect.
As described above, each object can have an Evolution Timeline which represents
its “lifetime”. On this lifetime line, two types of Timepoints can be de?ned: i) time
constrains for the de?nition of time constrains, such as a deadline for an Intention,
and ii) timestamps to de?ne when an instance enters into a speci?c state, such as when
an Intention’s instance assumes a SATISFIED state.
Notice that, for now, we are not constraining an Evolution Timeline to be as-
sociated to only one Object. In fact, as shown by the model, a same timeline can
be shared among di?erent Objects. Indeed, we think that having a unique timeline
for the entire organization on which Timepoints belonging to di?erent Objects can
coexist, will be useful for analysis activities.
Finally, notice the relationship among State and Timepoint which was also de-
scribed in the previous subsection. Each State can have multiple Timepoints repre-
senting all the times (timestamps) an Object entered in that particular State. More-
22
Figure 2.15: The Evolution Timeline concept.
over, a State might not have a Timepoint associated, i.e., the Object never assumed
that State, and a Timepoint might not have a State associated, i.e., the Timepoint
represents a time constrain for the associated Object and not a timestamp for a par-
ticular State.
Figure 2.16 shows an example describing the above concepts with respect to the
Intention primitive type.
In this example, we can see how an Intention was de?ned on March 1, 2009 with
a deadline ?xed on March 30, 2009. Moreover, the Intention was started to be
pursed on March 3, 2009 to be paused on March 11, 2009 and ?nally satis?ed on
March 23, 2009 (after it was re-pursed on March 13, 2009). Notice that, all the above
Timempoints represent timestamps with the exception of the deadline Timepoint con-
strain.
23
Figure 2.16: An example of Evolution Timeline for the Intention primitive type.
24
3 The BIM Repository and the
Abstraction Mechanisms
A BIM Repository is a persistent location in which organization and business data
are stored and maintained in order to be fetched to perform some particular task,
e.g., analytics tasks (see Section 4). In particular, a BIM Repository consists of struc-
tured classes(or types) and objects (or instances) which are de?ned using the BIM
metamodel.
In general, as with other modelling languages, classes and objects can be organized
along the three dimensions of aggregation, classi?cation and generalization (see [17]).
As described in [20], the act of “abstracting a collection of units into a new unit is
called aggregation”; indeed, an aggregation is a special type of association in which
objects or classes are assembled or con?gured together to create a more complex object
or class.
For example, the John object can be aggregated into the Pizza Pizza Sales Depart-
ment object and respectively, the Employee class can be aggregated into the Depart-
ment class. As we will describe in Section 3.1, we adopt the feature model [20] in order
to allow users to be ?exible during the aggregation activity.
The classi?cation dimension calls for each object or class to be an instance of one or
more generic classes or metaclasses. In fact, referring to the previous example, John
is an instance of Employee while Employee is an instance of the ActorClass metaclass.
The Classi?cation dimension is also used for relationships belonging to the model;
in fact, we can have an HomeAddress object which can be classi?ed by an Address
class which, similarly, can be classi?ed by an AddressClass metaclass. More details
about Classi?cation is provided in Section 3.1 which describes the four meta layer
metamodelling architecture of the Meta Object Facility (MOF) [30] and the features
inherited from Telos [29].
Classes and Metaclasses can be specialized along generalization or ISA hierarchies.
As de?ned in [20], generalization is the act of “abstracting the commonalities among
a collection of units into a new conceptual unit suppressing detailed di?erences”. For
example, an Employee class may have subclasses such as Clerk, Sales Person, etc.;
similarly, an ObjectClass metaclass may have submetaclasses such as EntityClass, Sit-
uationClass, etc. Notice that ISA hierarchies are orthogonal to the classi?cation di-
mension; therefore, all the above subclasses, i.e., Clerk, Sales Person, etc., should
be instances of the ActorClass.
In the next subsections, further details of aggregation, classi?cation, and general-
ization are provided. In particular, in Section 3.4 a possible implementation of the
abstraction mechanisms for the BIM metamodel is illustrated and described.
25
3.1 Aggregation: Mandatory, Optional, OR and
Alternative
The BIM provides a ?exible way for the aggregation activity that allows to have a
direct control in the choice of the parts (i.e., the “partOf“ relationship) constituting
an Object.
Indeed, the BIM aggregation mechanism uses the same approach described in [20]
in which the relationships between a parent and its children are categorized as:
• Mandatory – a child is required,
• Optional – a child is optional,
• Or – at least one of the children must be selected,
• Alternative (xor) – exact one of the children must be selected.
Figure 3.1 provides an example which shows the visual notation used to aggregate
di?erent Intentions. The example, also show how cardinalities can be used to add
more semantic during the aggregation activity; e.g., the “iPods produced” aggregates
from one to one million “one iPod produced”. Moreover, since the “one iPod produced”
is mandatory the range of cardinality must start from “one”.
Figure 3.1: An Intention aggregation hierarchy.
In Section 3.4, an underlying model for the aggregation mechanism is described.
3.2 Classi?cation: The OMG Four-Layer Metamodel
Architecture
The BIM metamodel is designed to be aligned with the OMG four-layer metamodel
architecture [30] which is summarized in Table 3.1.
26
Layer Description Example of model elements
M3: meta-
metamodel
De?nes the language for spec-
ifying metamodels.
MetaClass, MetaAttribute, MetaOp-
eration
M2: meta-
model
An instance of a meta-
metamodel. De?nes the lan-
guage for specifying a model.
Class, Attribute, Operation, Compo-
nent
M1: model An instance of a metamodel.
De?nes a language to de-
scribe an information do-
main.
StockShare, askPrice, sellLimi-
tOrder, StockQuoteServer
M0: user
objects (user
data)
An instance of a model. De-
?nes a speci?c information
domain.
< Acme Software Share 98789 >,
654.56, sell limit order,
< Stock Quote Svr 32123 >
Table 3.1: The OMG four-layer metamodel architecture.
In this architecture, a model at one layer is used to specify models belonging to the
layer below. Similarly, a model at one layer can be seen as an instance of a particular
model in the layer above.
Usually, models at M
n
layer have an higher level of abstraction and are typically
more compacts than models at M
n?1
layer. In fact, models at M
n?1
are more elaborate
than the models at M
n
layer that describe them. Figure 3.2 shows an example where
the top layers M3 and M2 of the architecture are represented and speci?ed, respectively,
by the MOF meta-metamodel and by the UML metamodel.
The BIM metamodel is at the same layer of the UML metamodel, i.e, at the M2
layer. Therefore, we would consider the BIM metamodel to be an instance of the MOF
meta-metamodel. We want this for two reasons: i) MOF enables the interoperability
of model and metadata driven systems; and, ii) MOF is quite spread across industry.
In this way, our model can exploit the interoperability provided by MOF in order to
facilitate its eventual integration with industry’s models and systems.
However, a major problem is that MOF (and UML) su?ers from the “shallow”
instantiation problem [6].
Basically, a class can only de?ne the semantic of its direct instances, but it has no
e?ect on entities created by further instantiation steps.
This is caused by the old “two-levels only” modeling philosophy which does not
adequately support a multi layer architecture. In fact, although model elements in a
multiple layer architecture can represent both objects and classes, i.e., an object at M2
layer can be seen as a class for objects at M1 layer, a class can never receive attributes
and associations from its classi?er, but only slots and links, thus leading to the shallow
instantiation problem (see [6] for further details).
The BIM metamodel needs to be able to in?uence both the M1 layer and the M0
layer in order to constrain designers in the choice of domain concepts and relationships
at M1 layer but also to propagate such “semantic” constrains (when requested) on
instances at M0 layer.
27
Figure 3.2: An example for the four-layer model architecture which uses MOF and
UML.
Figure 3.3 shows an example of such deep instantiation in which some “instance of”
links are missing to simplify the illustration. Notice how:
1. associations, such as “evaluates”, can be propagated across multiple layers;
2. some attributes, such as “currentValue”, can be propagated across multiple layers
while re?ning their types, e.g., the type of currentValue is NumberClass at M2
while is Integer at M1;
3. some attributes, such as “metric”, can be limited to speci?c layer, i.e., metric is
instantiated at M1 while is disappearing at M0;
4. some attributes, such as “director”, can be freely de?ned by the designer at M1
but are not speci?ed at M2, i.e., they are domain speci?c and are not mandatory
by the BIM metamodel.
In order to support such deep instantiation, the BIM is inspired by the Telos lan-
guage [29] in which classes, attributes, and associations
1
are collectively referred to
1
In Telos associations are represented using attributes which are binary relationships between enti-
ties, i.e. classes, or other relationships.
28
Figure 3.3: An example of the deep instantiation concept required in the BIM.
by the term “proposition” and are treated uniformly by the structuring mechanisms
of aggregation, classi?cation, and generalization. Therefore, as shown in Figure 3.4,
we can have metaclasses, classes, objects; but also: i) metattributes, attributes, slots;
and, ii) metassociations, associations, links
2
.
Figure 3.4: The Metaclass, Class and Objects instantiation.
A careful reader can observe that the UML metamodel introduces Instance meta-
classes in order to “link” objects of di?erent types at the M0 Layer. Although this can
2
In Telos is also possible to have meta-metaclasses, meta-metattributes, meta-metassociations and
so on (although for our aim it is not necessary).
29
resolve the instantiation of associations at the M0 layer, this approach arises issues
such as the “ambiguous classi?cation” and the “replication of concepts” [6]. Moreover,
it can increase the complexity of the model and lead to inconsistencies and the losing of
precision; and, it does not satisfy the “need” of propagating attributes across di?erent
layers.
3.3 Generalization: Subtypes and Business Terms
Specialization
The primitive types presented in Section 2 represent the model elements provided by
the BIM metamodel for the description of the di?erent concepts belonging to a generic
business environment. Moreover, in order to describe particular instances of the real
world, the BIM metamodel de?nes a set of subtypes whose semantic is described in
Table 3.2.
Moreover, in order to cover and map to business terminologies, such as a vision,
mission, or strategy, we use metaproperties such as (i) short-/long-term, (ii) many/few
instances, (iii) formal/informal de?nition, and (iv) chances of success. Clusters of
terms from a business glossary, such as Vision, Strategic/Tactical goal, Softgoal, Ob-
jective are then represented in terms of a single BIM primitive concept (Intention)
but each has di?erent combinations of values for the four metaproperties. For exam-
ple, a Vision is a long-term Intention without a formal de?nition, which is likely to
only have a few instances (usually one) whose chances of success are low (depending
on many uncertain factors).
To represent business terms, the BIM metamodel de?nes an attribute associated to
the ThingClass metaclass, called type which is inherited by all the other metaclass
in the metamodel to store speci?c terminology for each Thing. Table 3.3
3
provides
common clusters for some of BIM’s primitive concepts.
Figure 3.5 shows an example of subtypes and business terms specialization for the
Intention primitive type.
3.4 An UML Class Diagram for the BIM’s Abstraction
Mechanisms
In this section we present a possible UML class diagram for the BIM metamodel, with
respect to the aggregation, classi?cation, and generalization mechanisms described in
the previous sections.
The model is shown in Figure 3.6. The RefinementLinkClass, the NodeElementClass
and the AggregationClass are parts of the aggregation mechanism. In particular,
3
To de?ne the set of terms illustrated in Table 3.3 we analyzed both the scienti?c litera-
ture, e.g., the Business Motivation Model (BMM) [3], and the business world, e.g., the
www.businessdictionary.com [4] site; however, the business terminology can be easily customized
for the domain at hand.
30
Primitive
Type
Subtype Subtype Description
Intention Operational Intention
An atomic Intention which has a
very strict and clear logical criterion of
satis?ability and can be achieved by an
operational process or activity.
Qualitative Intention
An atomic Intention which has not
a clear-cut criterion for its satisfaction
and can be claimed only when there is
su?cient positive and little negative ev-
idences (or unsatisfaction in the oppo-
site case).
Actor Agent
Actor with concrete, physical manifes-
tations, such as a human individual.
We use the term agent instead of per-
son for generality, so that it can be used
to refer to human as well as arti?cial
(hardware/software agents). An agent
has dependencies that apply regardless
of what roles he/she/it happens to
be playing. These characteristics are
typically not easily transferable to
other individuals, e.g. its skills and ex-
periences, and its physical limitations
[1].
Role
Abstract characterization of the be-
havior of a social actor within some
specialized context or domain of en-
deavor. Its characteristics are easily
transferable to other social actors. The
dependencies associated with a role
apply regardless of the agent who plays
the role [1].
Position
Intermediate abstraction that can be
used between a role and an agent. It
is a set of roles typically played by one
agent (e.g., assigned jointly to that one
agent). We say that an agent occupies
a position. A position is said to cover a
role [1].
Table 3.2: Subtypes belonging to the BIM metamodel.
31
BIM Concept Business Terms
Intention Vision, Strategic/Tactical Goal, SoftGoal, Objective
Process Mission, Strategy, Tactic, Initiative, Business Process, Ac-
tivity
Actor Organization, Business Unit, Human person, System Appli-
cation
Resource Monetary / Infrastructure / Economic Good / Information
/ Human / Capability Resource
Directive Policy, Rule
Table 3.3: An example of business terms captured with the Thing’s attribute type.
Figure 3.5: An example of subtypes and business terms specialization for the
Intention primitive type.
the attribute mandatory of the RefinementLinkClass allows to specify if the “re-
?ner” component in the re?nement relationship is mandatory or optional; while, the
attribute type of the AggregationClass allows to specify the type of the aggregation,
i.e., OR, AND, XOR, Alternative, performed on the sub-components.
In regard to the generalization mechanism, the class named “...” (specializing the
ThingClass) represents the di?erent primitive types presented in Section 2. Instead,
the attribute type belonging to the ThingClass is used to specify the business termi-
nology shown in table 3.3.
Finally, the right part of the model represents the state and the evolution timeline
concepts describe in Subsection 2.4 and Subsection 2.5.
Figure 3.7 shows an example of how the model works underlying the feature model
visual notation (see Subsection 3.1).
32
Figure 3.6: An UML Class diagram for the BIM’s abstraction mechanism.
Figure 3.7: An example of aggregation using (b) the feature model visual notation and
(a) the underlying UML Class Diagram.
33
4 Strategic Analysis through Mappings
We illustrate how the richness and ?exibility of BIM can be used to represent widespread
strategic planning models. Moreover, since the ?nal aim of BIM is to support anal-
ysis activities for answering questions such as “What will happen next?” or “Where
exactly is the problem?” [8], we describe how BIM can be projected onto di?erent
analysis models.
In particular, the following target models are considered:
• a goal reasoning model based on a formal goal model [14],
• the SWOT analysis model [9],
• the Strategic Map [22] model,
• the Balanced SCorecards (BSCs)[21](and Key Performance Indicators [33]) model.
4.1 Goal (Intention) Reasoning: A Formal (Axiomatic)
Model
In the BIM, Intention analysis and reasoning are given a prominent role to help
stakeholders in the de?nition of their intentions and relationships among them, such
as con?icts and negative or positive contributions.
As we described in Section 2.1, the Intention primitive type can be used to de?ne
the hierarchy of the Vision, Goals, and Objectives of an organization in which nodes
can be connected by in?uence links.
In this section, we want to provide an underneath algorithm which enables the
reasoning on the Intentions belonging to such a hierarchy.
At this aim, we project the BIM toward the goal reasoning model described in [14].
In this work, the authors adopts a formal goal model to make the goal analysis process
concrete through the use of forward and backward reasoning. Notice that, the model
is used in the context of the Tropos methodology [2] which adopts the i
?
[41] modelling
framework for analyzing requirements (Early Requirements and Late Requirements
1
).
In particular, the formal model goals is used by the software engineer to cope with
qualitative relationships and inconsistencies among goals during the early requirements
phase.
1
The former is concerned with understanding the organizational context within which the system-
to-be will eventually function; the latter, on the other hand, is concerned with a de?nition of the
functional and non-functional requirements of the system-to-be.
34
The formal model description resides in the de?nition of the notions of goal graphs
and the axiomatic representation of goal relations. The goal graphs is de?ned trough a
set of goal nodes G
i
and of relations (G
1
, ..., G
n
)
r
? G over them, including the (n+1)-
any relations and, or and the binary relations +S, -S, +D, -D, ++S, - -S, +,-++, - -.
For a in depth description we remand to [14] while here we brie?y recall the intuitive
meaning of such relationships.
For and and or we have that:
• (G
1
, ..., G
n
)
and
? G means that G is satis?ed (resp. denied) if all G
1
, ..., G
n
are
satis?ed (resp. if at least one G
i
is denied);
• (G
1
, ..., G
n
)
or
? G means that G is denied (resp. satis?ed) if all G
1
, ..., G
n
are
denied (resp. if at least one G
i
is satis?ed);
For the other binary relationships, an example is provided by: G
2
+S
? G
1
(resp.
G
2
++S
? G
1
) means that if G
2
is satis?ed, then there is some (resp. a full) evidence
that G
1
is satis?ed, but if G
2
is denied, then nothing is said about the denial of G
1
.
To generalize the previous G
2
+S
? G
1
relationship, we said that, the “S” (resp. “D”)
symbol denotes the fact that the satis?ability (resp. deniability) value of the source
goal, e.g., G
2
, is propagated; the “+” (resp. “-”) symbol denotes the fact that the
propagation is positive (resp. negative), in the sense that satis?ability propagates
to satis?ability (resp. deniability) and deniability propagates to deniability (resp.
satis?ability).
Finally, the relations +, -, ++, - - are de?ned such that G
2
r
? G
1
is a shorthand for
the combination of the two corresponding relationships G
2
rS
? G
1
and G
2
rD
? G
1
, e.g.,
G
2
+
? G
1
is a shorthand for the combination of G
2
+S
? G
1
and G
2
+D
? G
1
. The ?rst
kind of relationships are called symmetric and the latter two asymmetric.
Now, a set of four distinct predicates over goals are introduced to be used with
ground axioms in order to reasoning on the goal model. They are: FS(G), FD(G)
and PS(G), PD(G); which mean, respectively, that there is (at least) full evidence
that goal G is satis?ed and that G is denied, and that there is at least partial evidence
that G is satis?ed and that G is denied. In their work, the authors provide a set of
ground axioms for the propagation rules which are soundness and completeness. An
example of of relation axiom is: G
2
+S
? G
1
: PS(G
2
) ? PS(G
1
).
Given a goal graph and an initial values assignment to some goals, the underlying
algorithm exploits the ground axioms for forward and backward reasoning tasks. In
particular, for the forward reasoning the assigned goals are called input goals (typically
the leaf goals) while for the backward reasoning the assigned goals are called target
goals (typically root goals).
The aim of the forward reasoning is the propagation of initial values (i.e., the input
goals) to all other goals of the graph; the user can look the ?nal values of the goals of
interest (i.e., the target goals).
Instead, the aim of the backward reasoning is the backward search of the possible
input values (i.e., the input goals) leading to some desired ?nal value (i.e.the target
values), under desired constrains, e.g., avoiding con?icts among goals.
35
In general, the forward reasoning is used for evaluating the impact of the adoption of
the di?erent alternatives with respect to the root goals; while, the backward reasoning,
is used to analyze goal models and ?nd the set of goals at the minimum costs that if
achieved can guarantee the achievement of the desired top goals and softgoals.
The algorithm of the formal goal model can be used within the BIM to allow such
reasoning. In fact, the goal relationships are accounted for within BIM through the
In?uenceClass metaclass illustrated in Figure 2.2.
Notice how the qualitativeStrength and the quantitativeStrengh allow, respectively,
to record the qualitative (e.g., + or - -) or quantitative (e.g., 0.7 or -0.3) strength of an
in?uence. The type attribute allows to specify whether the satis?ability or deniability is
propagated, i.e., S or D. The StateClass, which is inherited from the ObjectClass,
is used to record the four states associated to FS(G), FD(G), PS(G) and PD(G)
predicates.
Finally, the ResourceClass, can help in the backward “search“ when we desire
to ?nd the set of Intentions at the minimum cost that, if achieved, can guarantee
the achievement of the desired top Intentions. Indeed, the ResourceClass can
represent the monetary resource required for the achievement (through a Process) of
an Intention which is used in the minimum cost analysis
2
.
Figure 4.1 shows an example of Intentions reasoning with respect to the example
described in Figure 2.3.
Figure 4.1: An example of Intentions reasoning with BIM.
Notice how the formal model is used for both Situations and Intentions. In the
?gure, the semantic of the in?uence relationships is the following:
• the satis?ability of “Outsourcing advertising company hired” Situation is prop-
agated negatively (-S) to the “Cost decreased” Intention; this means that if the
former holds the latter is partial denied; nothing is said about the denial of the
“Outsourcing advertising company hired”;
2
Alternatively, a redundant attribute called cost can be added in the de?nition of the IntentionClass
metaclass.
36
• the satis?ability (resp. deniability) of “Best customers attracted and retained”
Intention is propagated positively (++) to the “Outsourcing advertising com-
pany hired” Situation; this means that if the former is satis?ed (resp. denied)
the latter holds (resp. does not hold);
• the satis?ability (resp. deniability) of “Best customers attracted and retained”
Intention is propagated positively (+) to the “More products sold” Intention;
this means that if the former is satis?ed (resp. denied) the latter is partial
satis?ed (resp. denied);
• the satis?ability of “Sta? need training” Situation is propagated negatively (-
S) to the “More products sold” Intention; this means that if the former holds
the latter is partial denied; nothing is said about the denial of the “Sta? need
training” Situation;
• the satis?ability of “Christmas season” Situation is propagated positively (++S)
to the “More products sold” Intention; this means that if the former holds the
latter is (at least partial) satis?ed; nothing is said about the denial of the “Christ-
mas season” Situation;
In order to show an example of forward reasoning on the model de?ned in Figure
4.1
3
, we input such a model in the same tool used in [14]. The result is as shown in
Figure 4.2.
Table 4.1 shows the results obtained by applying forward reasoning. The ?rst three
rows correspond to Situations, followed by three rows for the top Intentions and
three rows for the bottom Intentions with respect to the Intentions hierarchy. In
the table, three experiments are described through initial values (Init ) and ?nal values
(Fin) for satis?ability (S) and deniability (D) of Situations/Intentions. In partic-
ular, these values can be: full (F); partial (P); an empty cell when the corresponding
element is not involved in the reasoning; or, a question mark symbol (?) when a result
cannot be calculated.
A brief description of the experiments is the following:
• Exp 1 : The “Christmas season” Situation is satis?ed (see the F value for the
-Init column) so is the “Best customers attracted and retained” Intention.
As result, we have that the “Shareholder value increased” is partial denied due
to the partial denying of the “Cost decreased” Intention.
• Exp 2 : The “Christmas season”, the “Sta? need training”, and “Focused on
career and skills development” initial values are set to full satis?ed (F). As result,
we have a full “Revenue increased” satisfaction (see below for the semantic of the
con?icts) but no information for the “Shareholder value increased” Intention
(represented by the question mark). This result is due to the fact that the ground
3
Notice that: i) we need to add an extra node (namely, “-”) to simulate the feature model approach
for the decomposition; ii) we need to select at least one of the OR sub-Intentions to properly use
the tool while preserving the semantic of our model.
37
Figure 4.2: An example of Intention reasoning using the tool described in [14].
axioms, in this case (G
1
, ..., G
n
)
and
? G, are not able to work with uncertainties
(see Subsection 4.4 for how to address such issues). In fact, we have no informa-
tion (see the question mark symbol ?) for the “Cost decreased” Intention.
• Exp 3 : The “Christmas season”, the “Sta? need training”, “Best customers
attracted and retained” and “Focused on career and skills development” initial
values are set to full satis?ed (F). As result, we have that the “Shareholder value
increased” is partial denied due to the partial denying of the “Cost decreased”
Intention and a con?ict (i.e., full satis?ed and partial denied) on the “Revenue
increased” Intention.
Therefore, in the three experiments we use di?erent strategies to satisfy the top
“Shareholder value increased” Intention which lead to di?erent results.
To conclude this section a ?nal observation regarding the in?uence from intentions
towards Situations must be made. In fact, an Intention can lead to (++) or
avoid/mitigate (- -) a Situation.
A clear example is shown in the analysis performed in Table 4.1 where the “Out-
sourcing advertising company hired” can hold as the result of the satisfaction of the
“Best customers attracted and retained” Intention; vice-versa, we have also that
“Sta? need training” Situation is avoided or mitigated by the satisfaction of “Fo-
cused on career and skills development”.
The latter is the semantic associated to the con?icting values (S=F and D=F) for
38
Situation / Intention Exp 1 Exp 2 Exp 3
Init Fin Init Fin Init Fin
S D S D S D S D S D S D
Outsourcing advertising
company hired
F F
Christmas season F F F F F F
Sta? need training F F F F F F
Shareholder value
increased
P ? P
Cost decreased P ? P
Revenue increased F F P F P
Best customers attracted
and retained
F F F F
Focused on career and
skills development
F F F F
More products sold F F P F
Table 4.1: A formal forward reasoning example.
the “Sta? need training” Situation which is propagated towards the“More products
sold” Intention. Notice also that, the formal model is not able to deal with with
uncertainty when some Intentions have not an initial value since the ground axioms
require a complete information for the reasoning algorithm.
Finally, a similar analysis for the backward reasoning can be performed using similar
experiments as shown in [14] both considering or not a cost criteria.
As a summary, we can said that the Intention reasoning model enables to:
• perform forward reasoning, in order to evaluate di?erent strategies for the satis-
faction of top Intentions elements;
• perform backward reasoning (considering also cost constrains), in order to eval-
uate the optimal input values leading to some desired ?nal value;
• perform analysis on Intention inconsistencies and con?icts in the Intention
hierarchy.
4.2 The SWOT Analysis with the BIM
The SWOT analysis [9] is a strategic planning method which is used to evaluate the
Strengths, the Weaknesses, the Opportunities, and Threats which are involved in
a business environment. The purpose of the analysis is to specify the goals of the
39
organization, business venture or project and identifying those internal and external
factors that are favorable and unfavorable to achieve these goals.
Since a scan of the internal and external environment covers an fundamental role in
the strategic planning process, the SWOT analysis can be considered as the ?rst stage
of such a process in which an organization is helped to focus on key issues.
Therefore, a SWOT analysis starts with the de?nition of a desired state of the world
in terms of a set of strategic goals. Than, the identi?cation of SWOTs with respect
to the these strategic goals is performed. The result is an essential information which
helps the decision makers in understanding the attainability of the selected strategic
goals given such SWOTs. If the goals are not attainable di?erent objectives must be
selected and the process repeated.
In detail, the description of SWOT factors is:
1. Strengths are resources and capabilities of an organization which can be used
as a basis for developing a competitive advantage since they are are helpful to
achieve the strategic goals;
2. Weaknesses are absence of (certain) strengths as resources and capabilities which
may be viewed as a weakness since they are are harmful to achieve strategic goals;
3. Opportunities are external conditions which can be helpful to achieve the strate-
gic goals since represent favorable circumstances for pro?t and growth;
4. Threats are external conditions, usually due to changes in the external environ-
ment, which can be harmful to the strategic goals.
The results of a SWOT analysis are often presented in the form of a matrix as
illustrated in Figure 4.3.
Notice how, some factors may be viewed as strengths/opportunities or weaknesses/threats
depending upon their impact on the organization’s goals, e.g, the opportunity or threat
“changing of customer tastes”.
Another way to use SWOT is for the matching and converting activities. The match-
ing is used to ?nd competitive advantages by “matching” the strengths to opportuni-
ties, while converting is the act of guide strategies in order to convert weaknesses or
threats into strengths or opportunities. Usually, if the threats or weaknesses cannot
be converted an organization should try to minimize or avoid them.
In particular, an organization can use a SWOT analysis to de?ne:
• S-O strategies, which pursue opportunities that ?t good to the organization’s
strengths.
• W-O strategies, which overcome or avoid weaknesses to pursue opportunities.
• S-T strategies, which identify ways to use organization’s strengths to reduce its
vulnerability to external threats.
• W-T strategies, which establish a defensive plan to avoid that organization’s
weaknesses accentuate external threats.
40
Figure 4.3: An example of SWOT matrix.
The BIM provides a formal way to perform the SWOT analysis since: i) allows to link
the SWOT factors directly to the strategic goals they impact upon; ii) allows a formal
reasoning on the set of strategic goals, SWOT factors and in?uences relationships
among them.
The latter can be very useful for the de?nition of S-O, W-O, S-T, W-T strategies
since make feasible the exploration of the di?erent alternatives relying on the forward
reasoning and backward reasoning approaches presented in Subsection 4.1.
As shown in Section 2.1, we use Situation to represent those internal and external
factors which can contribute positively or negatively to the achievement of Intention,
i.e., strategic goals.
Notice that, as described in Subsection 4.1, when a schema is de?ned, some Intention
can be introduced to mitigate or avoid some Situations and some (harmful) Situations
can arise due to the presence, in the schema, of speci?c Intentions.
Moreover, it must be said that, in the BIM, we characterized as strength, weakness,
opportunity or threat the “in?uence” that exist from a Situation to an Intention.
This allows to represent those cases in which the same Situation can represent, for
example, a strength with respect to an Intention while representing a weakness with
respect to another.
Table 4.2 shows how to map the SWOT in?uences to the formal model presented in
Subsection 4.1, while Figure 4.4 illustrates an example.
In the ?gure, the “More products sold” is de?ned to exploit the “Christmas season”
external opportunity. This opportunity is matched by the“E?cient and e?ective dis-
tribution channels” internal strength that allows to deal with the high demand during
41
SWOT In?uence Formal Model In?uence
Strength +S, ++S
Weakness -S, –S
Opportunity +S, ++S
Threat -S, - -S
Table 4.2: SWOT and formal model mapping.
Figure 4.4: A SWOT analysis example with BIM
Christmas.
Moreover, in order to avoid and mitigate weaknesses and threats, two strategic goals
are also de?ned in the schema, namely “Focused on career and skills development” and
“New set of products researched” strategic goals.
The former attempts: i) to mitigate the lack of Sta?’s skills in order to be prepared
for the Christmas; and ii) to reduce the organization vulnerability to the external threat
helping the Sta? to turn the customer’s taste toward the Organization’s products.
The latter, the “New set of products researched”, is de?ned and pursued as a de-
fensive plan to match the new customer’s taste.
4.3 De?ne Strategic Map, Balanced Scorecard and Key
Performance Indicators with BIM
Important instruments for strategic planning are Strategic Maps (SMs) [22] and Bal-
anced SCorecards (BSCs)[21]. The former are visual representation of the strategy of
an organization which shows organization plans used to achieve missions and visions.
In particular, a Strategic map illustrates the cause-and-e?ect relationships between
di?erent strategic goals and the associated measures, the key performance indicators
(KPIs).
These measures are included in the latter, the BSC, which represents a “balanced”
range of metrics against which to measure the Organization’s performance. The mean-
ing of “balance” is provided by the fact that the broader view of leading indicators
42
of performance includes also non-?nancial metrics, such as “learning and growth of
employees”, “customer satisfaction”, etc.
The combination of SMs and BSCs follows the principle of “you cannot measure what
you cannot describe”. In fact, SMs aim to describe the direction of an organization
while BSCs aim to de?ne a comprehensive set of performance measures that provides
the framework for a strategic measurement and a management system.
Both the SMs and BSCs describe and measure organizational performance across
four balanced perspectives: ?nancial, customers, internal business processes, and learn-
ing and growth (for their descriptions and further details see [21] and [22]). In general,
these perspectives, allow to see the organization and the business environment from
di?erent viewpoints and not only from the ?nancial aspects.
As described in [21], the four perspectives have been found to be robust across a
wide variety of companies and industries but should be considered a template. Indeed,
no mathematical theorem exists to proof that four perspectives are both necessary and
su?cient.
Within each of the four perspectives, the organization must de?ne the following
elements:
1. Strategic goals
4
– strategies which must be achieved in that perspective;
2. Measures – the progresses toward that particular strategic goals;
3. Targets – the target value sought for each measure;
4. Initiatives – what should be done to facilitate the achievement of the target;
5. Cause-e?ect relationships – in?uences among strategic goals (or measures).
Figure 4.5 illustrates an example of such elements in which only Targets are missing.
A typical target can be, for example, a value of $10,000 for the Revenue measure for
satisfying the “Revenue increased” strategic goal.
A common approach to evaluate the performance of an organization and how suc-
cessful it is in achieving short and long-term goals, is the use of KPIs [33]. KPIs are
quanti?able measurements which re?ect the performance of an organization towards
its goals. Therefore, BSCs can express measures and targets through a set of KPIs.
BIM integrates in a single conceptual framework the primitive concepts that charac-
terize SMs, BSCs and KPIs, as well as requirements models in Software Engineering.
Through projection mappings on a global BIM model, it is possible to obtain par-
tial models that can be analyzed through SM, BSC, KPI and formal goal reasoning
techniques [14] as described in previous sections.
Using the fragment in Figure 2.2 and the IndicatorClass described in Figure 2.8(a),
we are able to represent both SMs and BSCs (i.e., a set of KPIs). In particular,
Intentions and Indicators represent strategic goals, their measures and associated
targets. Processes with the type attribute set to Initiative (see Table 32) describe
4
We use the strategic goal term instead of the objective as used in the BSC.
43
initiatives used to reach targets. Instances of the Influence metaclass address cause-
e?ect relationships (both in quantitative and qualitative ways). Finally, the perspective
attribute helps to characterize elements along the four di?erent perspectives.
An example of such mapping is shown in Figure 4.5, corresponding to the model of
Figure 2.1.
Figure 4.5: The BestTech Strategic Map and Balanced Scorecard de?ned with BIM.
Notice how, the BIM model can represent a possible underneath formal schema for
the SM and BSC described in Figure 2.1. Therefore, SM and BSC should be used for
illustration purpose, since familiar to executives, middle managers, etc., while BIM
should be used to formalize such abstracted human-language to a machine-readable
language on which queries, in depth analysis, etc., can be performed.
In conclusion we can a?rm that, as SMs and BSCs do, the BIM is: i) a way of
providing a macro view of an organization’s strategy using the Intention primitive
type to describe strategic and tactical goals; and ii), a way of constructing metrics to
evaluate performance against these strategies using the Indicator primitive type.
However, at the contrary of SMs and BSCs, the BIM allows more in depth analysis
on the schema obtained after the designing activity.
4.4 Probabilistic Graphical Model for Intention
Reasoning
In Section 4.1, we described a solution based on formal logic model to provide a reason-
ing mechanism on Intentions. However, we also highlighted that such kind of model
44
is able to provide only partial results in condition of uncertainty. For example, we
would recall the experiment two in Table 4.1 in which a question mark symbol (?) was
introduced for the “Cost decreased” and “Shareholder value increased” Intentions.
The issue of treat with uncertainty is an inescapable aspect of most real-world
applications; indeed, it is quite common to have not a complete information during
an analysis activity. Future works for BIM, include the investigation of probabilistic
(graphical) models [23], which make the uncertainty explicit and provide models that
are more faithful to reality.
Probabilistic graphical models are approaches model-based which allow interpretable
models to be constructed and then manipulated by reasoning algorithms. These models
can be de?ned by an analyst or can be learned automatically from data in order
to facilitate their construction when a manual design is di?cult or even impossible.
Di?erent Probabilistic graphical models have been de?ned in the scienti?c community,
such as Bayesian networks, undirected Markov networks, In?uence Diagrams, etc. (see
[23] a comprehensive discussion).
One of our goals within BIM, is to adapt such models in order to manage uncer-
tainty to perform causal reasoning and decision making under such circumstances. In
particular, we are concentrating on Bayesian networks and in providing a ?rst step
toward the use of In?uence Diagrams.
45
5 A Case Study
In this section, we sketch a case study for BestTech Inc. for which we constructed
a complete BIM schema. Part of the schema is shown in Figure 5.1. This schema
provides a comprehensive description of the business and its environment, balanced
along the four perspectives discussed earlier. For example, from the Financial Per-
spective, the top-level intention is Shareholder value increased; one of its sub-intention
Cost decreased is further re?ned into Management cost decreased and Supply chain
cost decreased. In general, for each perspective, Intentions have their associated
Indicators, e.g., Market share for Market share increased (from the Customer per-
spective), and they are related to high-level processes (strategies), e.g., Rewards pro-
gram.
The BestTech schema can be queried by the business analyst to answer questions
such as “Which are the in?uencers and sub-intentions for Revenue Increased”, or
“Which are the Intentions whose performance is poor (red zone) and whose deadline
is at the end of the month”. Since data often resides in and scatters across databases,
such queries are translated through schema mappings into database queries, and the
answers are then translated back into business-level concepts. Schema mapping be-
tween a BIM schema and database schema is a ongoing research in our group. More-
over, this schema can be projected along di?erent views. An example is illustrated by
the SM of Figure 4.5 which is a useful view when communicating an organization’s
strategies with the BestTech executives. Moreover, if the need is to perform analysis,
we can project the schema towards a variety of analysis models, as discussed in Sec-
tion 4. With such projections, we can respond to queries such as “Show me all the
Intentions which are in con?ict with at least one other Intention” or “Show me the
impact of denial of the Marketing improved Intention”.
46
Figure 5.1: Part of the BestTech BIM schema.
47
6 Related Work
The use of business-level concepts—such as business objects, rules and processes—has
been researched widely for at least 15 years and is already practiced to some extent
in both Data Engineering and Software Engineering [38, 25, 19]. These e?orts have
more recently resulted in standards, e.g., OMG’s Business Process Modeling Notation
(BPMN) [32]. Such proposals focus on modeling objects and processes, with little
attention paid to objectives.
Enterprise modeling languages (enterprise ontologies, to some) have also been re-
searched for a long time, with the express intention of aligning business and IT con-
cerns. Examples of this line of research include TOVE [10], REA [26] and the Zachman
Framework for Enterprise Architecture [43], as well as TOGAF [39]. Of those, BMM [3]
is closest in spirit to BIM. Our proposal places the BIM concepts we adopted from
BMM on an ontological foundation adopted from DOLCE [11] and also integrates
those with state-of-the-art abstraction mechanisms.
Notably, our concept of Situation is akin to the notions of description and situation
proposed in [12], but the authors there envisioned semantic web applications, rather
than business ones.
The Zachman Framework for Enterprise Architecture is one of the oldest proposals
for enterprise modeling. The framework consists of a table of 5 rows and 6 columns.
The rows de?ne an IT system and its context from di?erent perspectives ranging from
scope (top row), to business model, information system model, technology model and
detailed description (lowest row). Each row of the table uses a di?erent language.
Columns de?ne common questions that need to be answered about each perspective:
what, how, where, who, when and why. The public part of the Zachman framework
consists of this table, with no stand taken on what notation or modeling method to
use. Issues of notation and method to use are addressed in the proprietary part. This
modeling framework has had considerable in?uence on enterprise modeling practice,
including recent work on Service-Oriented Architectures (SOAs). BIM ?ts within the
Zachman framework, focusing on the why column, but o?ers a di?erent set of primitive
concepts for capturing why concerns than other proposals in the literature.
As indicated in the introduction, the other modeling proposals that relate to our
work are i* [42], URN/GRL [18] and KAOS [7, 40], all from the general area of Goal-
Oriented Requirements Engineering. From these we have adopted intentional and
social concepts. These models lack primitive constructs for in?uence relationships,
indicators, and various types of situations integrated in the BIM modeling framework.
Recent proposals extending URN do include indicators [34], but BIM’s indicators are
more general and they can be used to measure any model object, including other
indicators.
From a business perspective, BIM models can capture what is commonly found in
48
Strategic Maps and Balanced Scorecards. They can also be mapped to other languages
that enable goal analysis and SWOT analysis, and we expect other mappings to prob-
abilistic frameworks such as Bayesian networks and Analytics [8] to enable reasoning
under uncertainties.
49
7 Conclusions
One important problem of Business Intelligence technologies is that information re-
quired and generated by such technologies is rarely explicitly linked to business con-
cepts, decisions and outcomes, and is therefore hard to interpret and use. In this report
we have proposed the Business Intelligence Model (BIM), as ?rst step towards bridg-
ing the gap between the worlds of business and data analytics. The proposed model
extends the notion of conceptual schema to accommodate business concepts such as
strategic objectives, business processes, in?uences, indicators, risks and trends. We
have showed, through examples taken from a case study how a BIM schema can sup-
port governance activities, including monitoring, auditing and analysis at the strategic
level. As mentioned before, for BIM to be useful, we also need technologies for trans-
lating queries speci?ed over a BIM schema into queries over database schemas, also for
translating answers back into business terms. Such work is being carried out within
the context of the strategic network for Business Intelligence, funded by the Natural
Sciences and Engineering Research Council (NSERC) of Canada
1
.
As for future work, along one direction, we are further evaluating and re?ning BIM
with a large scale, real-world case study. Along another, we are extending to cover
the tactical level of business organizations, and along a third, we plan to extend our
model to incorporate uncertainty in strategic modeling and analysis through the use
of Bayesian networks. This will enable BIM to support statistical decision making [23]
and will complement the logic-based analysis techniques currently within BIM’s scope.
Acknowledgments
This work was supported by the Business Intelligence Network (BIN) and NSERC.
We are grateful to G. Mussbacher, G. Richards, E. Yu and many others for useful
discussions.
1http://bin.cs.toronto.edu/home/index.php andhttp://www.nserc-crsng.gc.ca/Partners-Partenaires/Networks-Reseaux/BIN-RVE eng.asp
50
8 Appendix
8.1 BIM Taxonomy
Refer to Table 2.1 and Table 3.2 for the taxonomy’s description.
Figure 8.1: The BIM ’s taxonomy.
51
Bibliography
[1] istarquickguide.http://istar.rwth-aachen.de/tiki-
index.php?page=iStarQuickGuide, 2009.
[2] P. Bresciani, P. Giorgini, F. Giunchiglia, J. Mylopoulos, and A. Perini. Tropos:
An agent-oriented software development methodology. Autonomous Agents and
Multi-Agent Systems, 8(3):203–236, 2004.
[3] Business Rules Group. The business motivation model: Business governance in
a volatile world.http://www.businessrulesgroup.org/bmm.shtml, 2007. Release
1.3.
[4] BusinessDictionary.com. Businessdictionary.http://www.businessdictionary.com,
2009.
[5] P. P.-S. Chen. The entity-relationship model—toward a uni?ed view of data.
ACM Trans. Database Syst., 1(1):9–36, 1976.
[6] A. Colin and K. Thomas. The essence of multilevel metamodeling. In Proc. of
the 4th Int. Conf. on The UML, Model. Lang., Concepts, and Tools, pages 19–33,
London, UK, 2001. Springer-Verlag.
[7] A. Dardenne, A. van Lamsweerde, and S. Fickas. Goal-directed requirements
acquisition. Sci. Comput. Program., 20(1-2):3–50, 1993.
[8] T. H. Davenport and J. G. Harris. Competing on Analytics: The New Science of
Winning. Harvard Business School Press, March 6, 2007.
[9] T. R. Dealtry. Dynamic Swot Analysis. Dynamic Swot Associates, 1994.
[10] M. S. Fox. The tove project towards a common-sense model of the enterprise. In
IEA/AIE ’92: Proceedings of the 5th international conference on Industrial and
engineering applications of arti?cial intelligence and expert systems, pages 25–34,
London, UK, 1992. Springer-Verlag.
[11] A. Gangemi, N. Guarino, C. Masolo, A. Oltramari, and L. Schneider. Sweetening
ontologies with dolce. In Proc. of the 13th Int. Conf. on Knowl. Engineer. and
Knowl. Manage. Ontol. and the Semant. Web, pages 166–181, London, UK, 2002.
Springer-Verlag.
[12] A. Gangemi and P. Mika. Understanding the semantic web through descrip-
tions and situations. In Proceedings of ODBASE03 Conference, pages 689–706.
Springer, 2003.
52
[13] A. Gill. Introduction to the Theory of Finite-State Machines. McGraw Hill, 1962.
[14] P. Giorgini, J. Mylopoulos, and R. Sebastiani. Goal-oriented requirements analysis
and reasoning in the tropos methodology. Eng. App. Artif. Intel., 18:159–171,
2005.
[15] C. Heitmeyer. Software cost reduction. In Encyclopedia of Software Engineering,
Two Volumes. Wiley, January 2002.
[16] R. Howard and J. Matheson. In?uence diagrams. Readings on the Principles and
Applications of Decision Analysis, Vol. II, 1984.
[17] R. Hull and R. King. Semantic database modeling: survey, applications, and
research issues. ACM Comput. Surv., 19(3):201–260, 1987.
[18] International Telecommunication Union: Recommendation Z.151 (11/08). User
Requirements Notation (URN) – Language de?nition.http://www.itu.int/rec/T-
REC-Z.151/en.
[19] S. Jablonski. On the complementarity of work?ow management and business
process modeling. SIGOIS Bull., 16(1):33–38, 1995.
[20] K. Kang, S. Cohen, J. Hess, W. Novak, and A. Peterson. Feature-oriented domain
analysis (foda) feasibility study. Technical report, CMU/SEI-90-TR-021, SEI,
Carnegie Mellon University, November 1990.
[21] R. S. Kaplan and D. P. Norton. Balanced Scorecard: Translating Strategy into
Action. Harvard Business School Press, 1996.
[22] R. S. Kaplan and D. P. Norton. Strategy maps: Converting intangible assets into
tangible outcomes. Harvard Business School Press, 2004.
[23] D. Koller and N. Friedman. Probabilistic Graphical Models: Principles and Tech-
niques. The MIT Press, August 2009.
[24] N.-S. Koutsoukis and G. Mitra. Decision Modelling and Information Sys-
tems: The Information Value Chain, volume 26. Springer: Operations Re-
search/Computer Science Interfaces Series, 2003.
[25] P. Loucopoulos and E. Katsouli. Modelling business rules in an o?ce environment.
SIGOIS Bull., 13(2):28–37, 1992.
[26] W. E. McCarthy. The rea accounting model: A generalized framework for ac-
counting systems in a shared data environment. The Accounting Review, 1982.
[27] Metaphysics Research Lab. Stanford Encyclopedia of Philosophy.http://plato.stanford.edu/, 2010.
[28] H. Morris. Analytic applications and decision-centric bi. Information Management
Magazine, June 1 2004.
53
[29] J. Mylopoulos, A. Borgida, M. Jarke, and M. Koubarakis. Telos: representing
knowledge about information systems. ACM Trans. Inf. Syst., 8(4):325–362, 1990.
[30] Object Management Group. Meta-object facility (MOF) 2.0 speci?cation.http://www.omg.org/spec/MOF/2.0/, 2006.
[31] Object Management Group. Object constraint language.http://www.omg.org/spec/OCL/2.0/, May 2006.
[32] Object Management Group. Business process modeling notation (bpmn).http://www.omg.org/spec/BPMN/1.2/, January 2009. version 1.2.
[33] D. Parmenter. Key Performance Indicators. John Wiley & Sons, 2007.
[34] A. Pourshahid, D. Amyot, L. Peyton, S. Ghanavati, P. Chen, M. Weiss, and
A. J. Forster. Business process management with the user requirements notation.
Electronic Commerce Research, 9(4):269–316, 2009.
[35] R. L. Sallam and K. Schlegel. Overcoming the gap between Business Intelligence
and Decision Support.http://www.gartner.com/it/page.jsp?id=927913, 2009.
[36] P. K. Saxena. Principles of Management: A Modern Approach. Global India
Publications Pvt Ltd., 2009.
[37] P.-Y. Schobbens, P. Heymans, J.-C. Trigaux, and Y. Bontemps. Generic semantics
of feature diagrams. Comput. Netw., 51(2):456–479, 2007.
[38] J. Sutherland. Business objects in corporate information systems. ACM Comput.
Surv., 27(2):274–276, 1995.
[39] The Open Group. TOGAF 9 - The Open Group Architecture Framework Version
9, 2009.
[40] A. van Lamsweerde. Requirements engineering: From System Goals to UML
Models to Software Speci?cations. John Wiley & Sons, 2009.
[41] E. Yu. Modelling strategic relationships for process reengineering. PhD thesis,
University of Toronto, Dept. of Computer Science, University of Toronto., 1995.
[42] E. Yu. Towards modelling and reasoning support for early-phase requirements
engineering. In Proc. 3rd IEEE Int. Symp. on Requirements Engineering, IEEE
CS, Washington, USA, 1997.
[43] J. A. Zachman. A framework for information system architecture. IBM Systems
Journal, 26(3):277–293, 1987.
54
doc_835622236.pdf
The Business Intelligence Model Strategic Modelling
The Business Intelligence Model:
Strategic Modelling
version 1.0
Daniele Barone
1
, John Mylopoulos
1
, Lei Jiang
1
, and Daniel
Amyot
2
1
Department of Computer Science, University of Toronto,
Toronto, ON, Canada
barone/jm/[email protected]
2
SITE, University of Ottawa, Ottawa, ON, Canada
[email protected]
April 14, 2010
Contents
1 Introduction 4
2 Primitive Concepts 8
2.1 Situations, Intentions and Processes . . . . . . . . . . . . . . . . . . . . 8
2.2 Actors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Objects’ Life-cycle and Behavior: A Finite State Machine Model . . . . 19
2.5 Evolution Timeline and Time Constrains for Objects . . . . . . . . . . . 22
3 The BIM Repository and the Abstraction Mechanisms 25
3.1 Aggregation: Mandatory, Optional, OR and Alternative . . . . . . . . . 26
3.2 Classi?cation: The OMG Four-Layer Metamodel Architecture . . . . . . 26
3.3 Generalization: Subtypes and Business Terms Specialization . . . . . . . 30
3.4 An UML Class Diagram for the BIM’s Abstraction Mechanisms . . . . . 30
4 Strategic Analysis through Mappings 34
4.1 Goal (Intention) Reasoning: A Formal (Axiomatic) Model . . . . . . . . 34
4.2 The SWOT Analysis with the BIM . . . . . . . . . . . . . . . . . . . . . 39
4.3 De?ne Strategic Map, Balanced Scorecard and Key Performance Indi-
cators with BIM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.4 Probabilistic Graphical Model for Intention Reasoning . . . . . . . . . . 44
5 A Case Study 46
6 Related Work 48
7 Conclusions 50
8 Appendix 51
8.1 BIM Taxonomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2
Business Intelligence (BI) consists of a range of technologies intended to assist large
organizations in determining the state and quality of their operations. BI activities
are meaningful in the context of a business organization and its objectives, strategies
and tactics, as well as a broader (external) context involving regulations, competitors,
customers, markets, etc. This business context (internal and external) de?nes the
e?ectiveness of business processes, and the things to monitor to ensure that business
objectives are being met and regulations and policies are complied with. The Business
Intelligence Model (BIM) provides a set of constructs for modeling and analyzing a
business context consisting of intentions, situations, processes, actors, in?uences, key
performance indicators, and more. It is intended to support the modeling and analysis
of a business organization at both a strategic and a tactical level. BIM schemas can
be used for governance activities, including analysis, monitoring and auditing. This
report presents some of the main innovations of BIM, including its primitive concepts,
its structuring mechanisms, analysis examples, as well as an overview of an illustrative
case study.
1 Introduction
Business intelligence (BI) consists of a range of technologies for using information
within organizations to ensure compliance to strategic and tactical objectives, as well
as to laws and regulations. As a research ?eld, it encompasses data and knowledge man-
agement, modeling of processes and policies, data quality, data privacy and security,
data cleaning and integration, data exchange, inconsistency management, information
retrieval, data mining, analytics, and decision support.
This interest in technologies and services that improve organizational governance
has caused dramatic growth for the BI market and the industry that serves it. By
now, most competitive organizations have a signi?cant investment in BI, much of it
technology-related, based on software tools and artifacts. However, as summarized by
Gartner [35], one important problem of BI technologies is that information generated
by BI systems and other decision inputs are rarely linked to business decisions and
outcomes. In addition, business people – be they executives, managers, consultants, or
analysts – are in general agreement that what they are looking for is not new gadgets
producing a dizzying array of largely useless statistics. Instead, they are interested
in having their business data analyzed in their terms: strategic objectives, business
models and strategies, business processes, markets, trends and risks. This gap between
the worlds of business and data remains today the greatest barrier to the adoption of
BI technology, and the greatest factor in the cost of applying BI technology.
We propose to bridge this gap by extending the notion of conceptual schema to
include concepts beyond entities and relationships. In particular, we propose the Busi-
ness Intelligence Model (BIM) as a business-level counterpart to the Entity-Relationship
or Relational Model, so that strategic objectives, business processes, risks and trends
can all be represented in a business schema. Users can query this schema, much
like conventional database schemas but with business terms, to perform analysis, to
track decisions and their impacts, or to explore suitable strategies to problems at
hand. Such queries are to be translated through schema mappings into queries de?ned
over databases and data warehouses, and the answers are to be translated back into
business-level concepts.
The main objective of this report is to introduce BIM’s constructs for modeling
business organizations at a strategic level. In particular, we present a set of primitive
concepts consisting of actors, intentions (e.g., goals), situations (strengths / weaknesses
/ opportunities / threats, a.k.a. SWOT), in?uences, processes, key performance indi-
cators and more. These concepts can be used in tandem with abstraction mechanisms,
such as generalization, aggregation and classi?cation, to develop global models of busi-
ness organizations for purposes of analysis, monitoring and auditing.
Our work is founded on modeling techniques from diverse sources. Abstract con-
cepts for describing all things are inspired from DOLCE[11]. The intentional and so-
4
cial concepts used in BIM are adopted from concepts in Goal-Oriented Requirements
Engineering, notably [7, 42, 18]. The notion of in?uence is adopted from in?uence
diagrams [16], a well-known and accepted decision analysis technique. Concepts re-
lated to SWOT analysis [9] and others have also been adopted from OMG’s Business
Motivation Model standard [3].
As shown in Figure 1.1, the ?nal aim of the BIM is to represent the internal and
external business and environment, and to support managers in making decisions at
each level of management providing answers to questions such as “What will happen
next?”. Indeed, the BIM is based on the idea that you cannot measure what you
cannot represent, and you cannot improve what you cannot measure.
As de?ned in [28, 24] and summarized in Table 1.1, people at di?erent levels in an
organization have di?erent types of decision-making responsibilities.
Strategic decisions, which are typically made by executive managers, a?ect the long-
term direction of an organization and are often complex and characterized by uncer-
tainty due to the limited availability of information. Usually, managers at this level
depend on their past experiences and instincts for making a decision.
Examples of strategic decisions can be to decide whether it is time to discontinue a
product line or to launch a new one.
Tactical decisions regard more intermediate-term issues and are typically made by
middle managers. The decisions made at this level attempt to move the organization
closer to reaching the strategic goals.
Examples of tactical decisions can be to hire an advertising agency in order to
promote a new product or to create an incentive plan for encouraging employees in
increasing the organization’s production.
Operational decisions concentrate on day-to-day organization activities and are typ-
ically made by lower-level managers. Decisions made at this level attempt to ensure
that the daily activities are conform with respect to business targets and standards in
order to achieve the strategic goals.
Examples of operational decisions include scheduling employees, purchasing raw
materials needed for production, or answering questions such as “Do we extend credit
to this customer?”.
Level of
Management
Core
Requirement
Nature of the Decision
Executive Strategic Planning Long term, unstructured, di?-
cult to develop speci?c decision
models
Mid-level Management Control Shorter term, semi-structured,
modeling possible.
Operational Operational Control Short term, structured, model-
ing possible.
Table 1.1: A taxonomy of management decision-making.
In this report, we focus on a particular use of the model for supporting the strate-
gic planning process, which is the process of de?ning an organization’s strategy (or
5
Organi zat on
St ruct ure
Model
Strategi c Model
Tactical Model
Operati onal Model
What i s t he best t hat can happen? (Opt i mi zat i on)
What wi l l happen next ? (Predi ct i ve model i ng)
What i f these trends conti nue? (Forecasti ng/extrapol ati on)
Why i s thi s happeni ng? (Stati sti cal anal ysi s)
What acti ons are needed? (Al erts)
Where exact l y i s t he probl em? (Query/ dri l l down)
How many, how often, where? (Ad hoc reports)
What happened? (St andard report s)
Answers
t o
Busi ness I nt el l i gence Model
Technolgies
Organi zati on
Market
More Intel l i gence
Less Intel l i gence
Figure 1.1: The Business Intelligence Model (“Questions” have been adapted from [8]).
direction) and making decisions on allocating the organization’s resources in order to
pursue this strategy.
In particular, an organization can follow di?erent approaches for strategic planning
such as the Situation-Target-Path, the Draw-See-Think-Plan, the See-Think-Draw, etc.
[36] which can be summarized by the following activities:
1. Formulate Vision & Mission: an organization must clearly de?ne its Vision and
Mission statements and the associated hierarchy of goals and objectives;
2. Situational Analysis: an analysis of the organization and its environment must
be conducted in order to identify in?uences among external and internal factors
and the organization goals, e.g., the SWOT analysis [9].
3. Develop Strategies: a set of alternatives for the ful?llment of goals must be
formulated in terms of actions and processes to be taken to achieve goals and in
terms of the resources required to execute these actions;
4. Implement Strategies: the “best” strategies must be chosen and implemented
using organizational processes and resources;
5. Evaluate and Monitor Performance: the implemented strategies must be evalu-
ated to see if they are working successfully.
In the next sections we present the BIM metamodel and how it can support the
above strategic planning activities.
6
In particular, Section 2 describes BIM’s primitive concepts and their use for strategic
modeling. Section 3 presents the mechanisms (generalization, aggregation, classi?ca-
tion) used to structure BIM models. Section 4 o?ers an overview of how BIM models
can be used to support a range of analyses, such as strategic map, goal analysis and
SWOT analysis. In Section 5, we present an illustrative case study, while Sections 6
and 7 respectively discuss related work and conclusions.
7
2 Primitive Concepts
Table 2.1 introduces the primitive concepts in BIM for modeling strategic objectives
and strategies.
As an ongoing example, we introduce BestTech Inc., a company in the market of cel-
lular phones and home computers
1
. Figure 2.1 shows its Strategic Map (SM) [22] and
Balanced Scorecard (BSC) [21], which are strategic planning instruments commonly
used in industry.
The former (SM) is a visual representation of the strategy of an organization showing
plans used to achieve missions and visions. In particular, it illustrates the cause-and-
e?ect relationships between di?erent strategic goals and associated measures, the key
performance indicators (KPIs) [33].
These measures are included in the latter, the BSC, which represents a “balanced”
range of metrics against which to measure the Organization’s performance. “Balance”
here means that the broader view of leading performance indicators includes also
non-?nancial concerns, such as “learning and growth of employees” and “customer
satisfaction”. Both SMs and BSCs describe and measure organizational performance
across four balanced perspectives: ?nancial, customers, internal business processes,
and learning and growth [22, 21].
The following subsections detail how BIM’s primitive concepts are used to describe
BestTech and its business environment, from a strategic viewpoint.
2.1 Situations, Intentions and Processes
Primitive concepts are represented as metaclasses (having names that end in -Class).
In particular (Figure 2.2
2
), Intention can be used to de?ne a hierarchy of Vision,
Strategic/Tactical goals, etc., which represents the desired end state of an organization.
Process models an organization’s Mission and the di?erent Strategies and, at tactical
and operational levels, their decomposition into Business processes and Activities.
Resource models what resources are required to execute such strategies.
In any strategic planning setting, a scan of the internal and external environment is a
fundamental issue [9]. Accordingly, BIM provides the Situation concept for modeling
internal and external situations that can be helpful or harmful to an organization’s
goals. In addition, we adopt the SWOT classi?cation [9] (see Section 4) in order to
de?ne the types of in?uence that a Situation may exercise on an Intention.
1
Adapted from a generic strategy map for a credit card company provided by the Balanced Scorecard
Institute –http://www.balancedscorecard.org (2010)
2
Dashed inheritance arrows indicate the existence of hidden metaclasses in between.
8
Concept Description Example Superclass
Thing BIM’s most general concept. This abstract meta-
class has everything as an instance.
Object Abstract metaclass whose instances per-
sist/endure over time [11].
Thing
Event Instantaneous happening (perdurant) that
changes an Object; can be described by a
Proposition that was false before the Event and
is true after [11, 15].
New order
received
Thing
Situation Partial state of the world described by a
Proposition. Situations can have structure con-
sisting of relations and Things standing in those
relations [27].
Christmas
season
Object
Proposition Describes a Situation. In general, Propositions
are true/false/unde?ned in a Situation [27].
Object
Intention A Proposition an Actor wants to make true [42]. More
products
sold
Proposition
Domain
Assumption
A Proposition assumed by an Actor to be true
for purposes of ful?lling an Intention.
Market
increases
at least
10%
annually
Proposition
Directive A Proposition prescribed by an authority in-
tended to constraint, guide, govern, or in?uence
elements of an organization such as Actors and
Processes [3].
Pizza de-
livery in
30 mins.
Proposition
Entity Object whose existence does not depend on that
of others.
Object
Actor Entity that carries out Actions to achieve
Intentions [42].
Sales
manager
Entity
Action Entity performed by an Actor that produces
Events and can have pre- and postconditions [7].
Deliver
product
Entity
Process Entity consisting of coordinated Actions to
achieve an Intention [3].
Sales
process
Entity
Resource Entity of value to an Actor [42]. Money,
informa-
tion
Entity
Indicator Measurable Object that gives information about
the state of an associated Object. Can be used
to quantify the satisfaction level of an Intention
or the operational performance of an Actor, a
Process or a Resource [33].
Number
of prod-
ucts sold
per week
Object
Relationship An Object that relates two or more Things and
whose existence depends on that of the Things it
relates [5].
Object
Influence A Relationship between two Things t
1
and t
2
,
where the state of t
1
constraints the state of t
2
in
a probabilistic or causal sense.
Relationship
Table 2.1: BIM primitive concepts.
9
Figure 2.1: A strategic map and a balanced scorecard for BestTech Inc.
Figure 2.2: Intention, Process and Situation metaclasses and associated relation-
ships in the BIM metamodel.
In this classi?cation, internal factors are situations further classi?ed into Strengths
or Weaknesses while external factors are Opportunities or Threats. For example, a
Threat for Wal-Mart may be “Exposure to political problems in the countries that
we operate in”. Notice that a strength with respect to one Intention may well be a
weakness for another.
Figure 2.3 shows a small example that instantiates the metaclasses of Figure 2.2.
Goals (i.e., Intentions), such as “Shareholder value increased” are decomposed into
subgoals. Subgoals may be mandatory or optional (the notation used here is adapted
from feature diagrams [37]). Goals may be achieved by carrying out Processes. In
10
turn, Processes require Resources; there are various types of Resources with as-
sociated icons. Situations can in?uence goals but also Processes, Resources, etc.
Opportunities and weaknesses are particular kinds of in?uences. In?uence relation-
ships
3
can actually exist between any two Things and play a pivotal role in governance
models.
Figure 2.3: An example of Situation, Intention and Process concepts at schema
level.
2.2 Actors
Figure 2.4 shows the Actor type and its relationships. An important relationship is
the “responsible for” that de?nes which Actor is liable to be called on to (legally)
answer when Intentions, Processes, Resources, and Directives are not acting
in accordance with the business of an organization. For example, an Actor can be
responsible for enforcing a work-site safety policy (i.e. a Directive). Indeed, if an
accident occurs due to inappropriate safety conditions or there is a problem during a
safety inspection, the responsibility will fall upon the Actor.
Actors must also comply with Directives, e.g., an employee must wear the pro-
tective equipments, and are able to de?ne them, e.g., the executive board can de?ne
a policy such as “Pizza must be delivered within 20 minutes from the order or it will
be free.”.
3
The instantiation of the EnumerationClass type for the qualitativeStrength attribute can vary
depending on the nature of Things involved in the relationship.
11
An Actor can desire Intentions which are satis?ed by other Actors; for example,
the “More products sold” Intention is desired by the executive board and is satis?ed
by the sales sta? under the responsibility of the sales manager. An Actor is also
capable of a Process (or an Action) that can actually performing or not.
Figure 2.4: The Actor primitive type and the associated relationships.
Figure 2.5 shows an example of Actors and their relationships with the business
environment.
Notice how cardinality values can be used for enriching the semantic of relationships
in order to describe the minimum and maximum number of associated elements within
a set. For example, in the ?gure, the sales department is constituted by a sales manager
and from a minimum of six to twelve sales employees. In particular, regarding the
“Package product” Action, a minimum of three employees are capable of the “Package
product” Action and at least two must perform this Action on a maximum of ten
that can be allocated to such Action.
Moreover, as described in Figure 2.6, an Actor can be specialized in Agent, Role,
and Position. As de?ned in i
?
[41], a Role is an abstract characterization of the
behavior of a social Actor within some specialized context or domain of endeavor, and
a Position is a set of Roles, usually played by one Agent. An Agent is an Actor
with concrete, physical manifestations, such as a human individual or an arti?cial
component of a system (hardware/software agents). Finally, an Agent can occupy a
Position, while a Position covers a Role. Figure 2.6 shows an example of Agent,
Role and Position.
2.3 Indicators
Indicators evaluate the quality of Objects to ensure compliance to internal policies
and external directives.
Figure 2.8(a) shows the Indicator metaclass, consisting of attributes such as target,
thresholds, extreme values, etc.[33]. The attributes ’s description is provided in Table
2.2.
12
Figure 2.5: An example of Actors and their relationships with the business
environment.
Figure 2.6: Agent, Role and Position in the BIM metamodel.
Notice that, in Table 2.2, the evaluationTime represents the timestamp associated
to a speci?c current value instance, i.e., the time when the instance is calculated and
created (see Section 2.5 for more details). This “system” time is di?erent from the
“domain” time (e.g. a Time dimension) used to navigate and calculate Indicators in
a historical period of time, e.g., the Time dimension’s value “March, 2009” to calculate
13
Figure 2.7: An example for Agent, Role and Position at schema level.
(a) (b) (c)
Figure 2.8: (a) The Indicator primitive type in the BIM metamodel. (b) An example
of visual notation for Indicators at instance level. (c) An Indicator ’s
Trend example.
the number of products sold in March.
An example of Indicator and the use of its attributes is shown in Figure 2.8(b).
Notice how the current value (currentValue) is found between Lower Threshold and
Target ; therefore, the company has almost reached its target in selling computers.
The current value is calculated at evaluationTime using the metrics de?ned in the
expression ?eld. In general, BIM relies on the Object Constraint Language (OCL) [31]
14
Meta-attribute’s
name
Description
currentValue the current value of an Indicator which is calculated
through the evaluation of a metric’s expression.
unitOfMeasure the unit of measure associated with the current value of an
Indicator.
expression the metric’s expression used to calculate the current value
of an Indicator.
dimension It allows to navigate (calculate the value of) an Indicator
along speci?c directions of interest.
target It is a value which allows to quantify the satisfaction level of
an Intention (or, more in general, a performance level de-
sired for a property of an Object). If an Indicator reaches
the Target the associated Intention is satis?ed.
(upper & lower)
threshold
It de?nes a threshold (upper or lower) value that can be as-
sumed by an Indicator that represents a critical situation
for the Organization’s business.
(upper & lower)
extremeValue
It de?nes the worst (upper or lower) value that can be as-
sumed by an Indicator which represents a critical (ex-
treme) situation for the Organization’s business.
evaluationTime The time in which the current value of an Indicator is
calculated.
Table 2.2: The description of Indicator’s attributes.
4
to de?ne such expressions. Moreover, one or more dimensions can be used to calculate
and navigate an Indicator along speci?c directions of interest, e.g., the number of
“Apple” computers sold.
BIM also supports the de?nition of operations, such as Trend, Risk, Reward or
Con?dence, at the class level (i.e., at the schema level), which calculate additional in-
formation (meta values) for the current values assumed by an Indicator. The result
is in-depth information on the subject of an Indicator. Table 2.3 provides a brief
description of such methods while Figure 2.8(c) is an example of a Trend operation
based on a linear regression method. The result is a trend with a slope of +0.6 which
means that: i) the trend is positive, and ii) the prediction of computers sold on Dec.
6, 2009 is about four units.
In order to better understand the use of Indicators, take a look at the following
example:
Suppose that a Car Dealer has in store 16 units of “Luxury car” that it desires to
sell over a period of one month (from December 1
st
, 2009 to December 31
st
, 2009).
4
The OCL is a declarative language for describing rules that apply to UML models developed by
the Object Management Group (OMG) and now part of the UML standard.
15
Method’s name Description
trend It represents a general movement over time of a statistically
detectable Indicators’s change.
risk It provides the actual loss associated to the current value
assumed by an Indicator.
reward It provides the actual gain associated to the current value
assumed by an Indicator.
con?dence It provides the actual con?dence associated to the current
value assumed by an Indicator. The con?dence is based
on i) the quality of the information used to calculate the
current value, ii) the reputation of the source who provided
the information and iii) the reliability of the method used
to calculate the current value.
Table 2.3: The description for the Indicator’s operations.
Suppose also that the actual (partial) state of the world, i.e. on December 6
th
,2009,
is as described in table 2.4 in which each car is identi?ed by a Vehicle Identi?cation
Number (VIN).
(Desired) State of the world Evaluation
Car (VIN 1HGBH41J8MN109180) Sold True
Car (VIN 1HGBH41J8MN109181) Sold False
Car (VIN 1HGBH41J8MN109182) Sold True
Car (VIN 1HGBH41J8MN109183) Sold False
Car (VIN 1HGBH41J8MN109184) Sold True
Car (VIN 1HGBH41J8MN109185) Sold False
Car (VIN 1HGBH41J8MN109186) Sold False
Car (VIN 1HGBH41J8MN109187) Sold True
Car (VIN 1HGBH41J8MN109188) Sold False
Car (VIN 1HGBH41J8MN109189) Sold True
Car (VIN 1HGBH41J8MN109190) Sold False
Car (VIN 1HGBH41J8MN109191) Sold True
Car (VIN 1HGBH41J8MN109192) Sold True
Car (VIN 1HGBH41J8MN109193) Sold False
Car (VIN 1HGBH41J8MN109194) Sold False
Car (VIN 1HGBH41J8MN109195) Sold True
Table 2.4: The (desired) states of the world for the “Car sold” Intention.
We have the following Indicator as described in Figure 2.9 and in Table 2.5. Notice
the use of the cardinality (0..16) to express the number of cars the organization desires
to sell. Moreover, Figure 2.10 shows the di?erent zones de?ned by using the extreme,
16
threshold and target values; and the current value for December 6
th
, 2009 lying on the
red zone.
Figure 2.9: An example of Indicators at schema level.
Meta-Level Schema-Level Instance-Level
Indicator Number of Cars Sold Inst 1:Number of Cars Sold
currentValue Integer 8
unitOfMeasure cars sold -
expression “context Car
select (car | car.status=’Sold’ ?
car.type=’Luxury car’ ) ? >
size()”
-
target Integer 13
(Lower)
threshold
Integer 10
(Lower)
extremeValue
Integer 7
dimension TypeOfCar “Luxury Car”
evaluationTime Time Dec 6, 2009
Table 2.5: An example of an Indicator at di?erent levels of modelling.
Figure 2.11 and Figure 2.12 show examples of Indicator’s operations. In particular,
the former is an example of a Trend Operation based on a linear regression
5
method.
The operation calculates the slope (or gradient) of the trend line ?tting the time series
provided in input, where the time series are the number of cars sold from Dec. 1
st
to
2009 to Dec. 6
th
, 2009. The result of the operation is a slope of ?0.7 which means
that on Dec. 6, 2009 the Trend for the number of cars sold is negative.
The latter, Figure 2.12, is an example of risk and reward operations which allow
to associate speci?c loss or gain values to any current value. In fact, if we suppose
that 30,000 USD is the total cost sustained for a car while 50,000 USD is the revenue
obtained by its selling, we have that: i) a total gain of 170,000 USD is obtained when
5
Notice that, di?erent ways for evaluating a Trend exist, such as logarithmic, exponential, etc.; in
this case we use the simple one, namely the linear regression.
17
Figure 2.10: The current value of the “Number of cars sold” Indicator on December,
6
th
.
the target is reached, ii) a loss of 130,000 USD is endured when the current value
assumes the extreme value and iii) a loss of 80,000 USD is endured on the current
value calculated on December 6, 2009.
Figure 2.11: The Trend for the “Number of cars sold” Indicator.
Figure 2.9 shows another Indicator, namely the SWOT Indicator, which is used to
evaluate the in?uences among Situations and Intentions. As described in Section
2.1, we have four kind on in?uences, namely, Strength, Weakness, Threat, and Oppor-
tunity. For each type, it is possible to de?ne a quality or quantity power scale, e.g.,
< high, medium, low > or [0, 100], in order to determine the degree of the in?uence.
Therefore, a SWOT Indicator can assume a current value that ranges among values of
the power scale and can have a unit of measure equal to Strength, Weakness, Threat,
or Opportunity. As described in Section 4, the SWOT Indicator’s current value is used
18
Risk and Reward
(
Y
)
T
h
o
u
s
a
n
d
s
-
U
S
D
0
100
200
300
400
500
600
(X) Number of Cars Sold
0 2 4 6 8 10 12 14 16
Gain
Loss
Threshold = 10
Target = 13
Worst Value = 7
Figure 2.12: The Loss and Gain values for the “Number of cars sold” Indicator.
by the In?uenceClass metaclass to support di?erent kind of analysis.
2.4 Objects’ Life-cycle and Behavior: A Finite State
Machine Model
A Finite State Machine (FSM) [13] provides a simple and e?ective means to control
the life-cycle and overall behavior of BIM’s Objects. In fact, a FSM is an abstract
computational model which allows to de?ne for each Object: i) a set of di?erent states
assumed by Objects’ instances in the real world; ii) the transitions among these
di?erent states in order to de?ne the (computational) behavior; iii) the events/inputs
which express the stimuli taken into account; and, iv) the actions/outputs which are
the possible responses that can be generated.
Formally, a FSM is a multi-tuple FSM = (?, ?, S, s
0
, ?, ?), where:
• (?) is the input alphabet of symbols representing external stimuli (inputs or
events) that are used by transition functions;
• (?) is the output alphabet of symbols representing responses (outputs or ac-
tions) that are provided by output functions;
• (S) is the set of possible states which are conditions of the state machine at a
certain time;
19
• (s
0
? S) is the start state;
• (? : S × ? ? S) is the state transition function. Based on the current state
s
c
? S and an input symbol i ? ?, it computes the transition to the next state
s
n
? S;
• (? : S ×? ? ?) is the output function (as de?ned in the Mealy model ).
As shown in the following subsection, the BIM also allows to associate to each state
a Time attribute which stores a timestamp value representing the last time in which
an Object enter in that speci?c state.
In general, for all Objects, it is possible to identify a START state and a (pos-
sible) END state. The START state allows to de?ne when an instance of the real
world becoming an instance of a speci?c type of the BIM. For example, the “Car
VIN=1HGBH41J8MN109180 sold” instance stars to be an Intention’s instance when
the car dealer de?nes a clear statement to sell that speci?c car, i.e. the car with
VIN=1HGBH41J8MN109180. Than, the car dealer will be able to pursue it, i.e., the
state BEING PURSUED.
In the opposite way, the END state allows to de?ne when the same instance stops to
be an (active) instance of that type
6
. For example, the “Car VIN=1HGBH41J8MN109180
sold” Intention’s instance stops to be an Intention’s instance when the car dealer
stops to pursued it either because it is achieved, failed or aborted (i.e.,the car dealer
has changed his mind).
Moreover, for each speci?c Object, such as the Intention, it is also possible to
de?ne speci?c states which have a particular semantic into the business environment.
For example, the above BEING PURSUED state can be de?ned for Intentions to
express the continuing activity by an Actor to achieve a speci?c Intention.
Figure 2.13 and Figure 2.14 show, respectively, possible FSM diagrams for Intention
and Resource. Notice that, a Vision cannot be ?tted in the diagram illustrated in
Figure 2.13 since, usually, it is not possible to de?ne a SATISFIED state or an END
state for such a concept. Therefore, if it is necessary, a speci?c FSM diagram must
be de?ned or, alternatively, the same FSM diagram can be used in which some states
will be never assumed by the instances.
Moreover, it is also important to notice the two “pass Deadline” and “pass Expir-
ing date” events which, respectively, lead to a FAILED Intention’s state and to an
EXPIRED Resource’s state.
These events are ?red when: i) the current time in the system is greater than, re-
spectively, the “deadline date” and the “expiring date” de?ned in the business environ-
ment (see next subsection); and, ii) the actual states of Intention’s and Resource’s
instances are, respectively, BEING PURSED and BEING CONSUMED. FSMs allows
to de?ne these events as guards which can be expressed by using constrains.
In this way, for example, it is possible to represent situations where a goal’s deadline
has passed and, since it is not longer important whether or not the goal will be satis?ed,
we transit to a FAILED state.
6
Stop to be an “active“ instance means that the instance is still recorded in the system for “history“
purpose (i.e., for analysis and queries) but is no more used for operative tasks.
20
Figure 2.13: A FSM diagram for Intention.
A similar situation exists for Resources in the case we have passed the expiring date.
In fact, the Resource stops to be a valid Resource loosing its intrinsic properties
that make it such as a Resource in the business environment; however, we can still
consume it with all the relative consequences.
The use of states and timestamps enable the de?nition of interesting queries over
Objects belonging to a schema. For example, we can express queries such as (1)=“List
all goals that are not yet satis?ed” or (2)=“Show the time in which
was satis?ed” that can be de?ned as:
(1) = context Intention
select (intention | not(intention.state =
?
SATISFIED
?
)
(2) = context Intention
select (intention | intention.name =
?
All car sold
?
and
intention.state =
?
SATISFIED
?
).time
21
Figure 2.14: A FSM diagram for Resource.
2.5 Evolution Timeline and Time Constrains for Objects
The previous subsection presented the concept of State to address issues related to the
life cycle and behavior of Objects. However, since we want to describe the “evolu-
tion” of Objects within a business environment, we need to introduce the concept of
Evolution Timeline. Figure 2.15 shows the fragment of the BIM metamodel aimed
to describe such aspect.
As described above, each object can have an Evolution Timeline which represents
its “lifetime”. On this lifetime line, two types of Timepoints can be de?ned: i) time
constrains for the de?nition of time constrains, such as a deadline for an Intention,
and ii) timestamps to de?ne when an instance enters into a speci?c state, such as when
an Intention’s instance assumes a SATISFIED state.
Notice that, for now, we are not constraining an Evolution Timeline to be as-
sociated to only one Object. In fact, as shown by the model, a same timeline can
be shared among di?erent Objects. Indeed, we think that having a unique timeline
for the entire organization on which Timepoints belonging to di?erent Objects can
coexist, will be useful for analysis activities.
Finally, notice the relationship among State and Timepoint which was also de-
scribed in the previous subsection. Each State can have multiple Timepoints repre-
senting all the times (timestamps) an Object entered in that particular State. More-
22
Figure 2.15: The Evolution Timeline concept.
over, a State might not have a Timepoint associated, i.e., the Object never assumed
that State, and a Timepoint might not have a State associated, i.e., the Timepoint
represents a time constrain for the associated Object and not a timestamp for a par-
ticular State.
Figure 2.16 shows an example describing the above concepts with respect to the
Intention primitive type.
In this example, we can see how an Intention was de?ned on March 1, 2009 with
a deadline ?xed on March 30, 2009. Moreover, the Intention was started to be
pursed on March 3, 2009 to be paused on March 11, 2009 and ?nally satis?ed on
March 23, 2009 (after it was re-pursed on March 13, 2009). Notice that, all the above
Timempoints represent timestamps with the exception of the deadline Timepoint con-
strain.
23
Figure 2.16: An example of Evolution Timeline for the Intention primitive type.
24
3 The BIM Repository and the
Abstraction Mechanisms
A BIM Repository is a persistent location in which organization and business data
are stored and maintained in order to be fetched to perform some particular task,
e.g., analytics tasks (see Section 4). In particular, a BIM Repository consists of struc-
tured classes(or types) and objects (or instances) which are de?ned using the BIM
metamodel.
In general, as with other modelling languages, classes and objects can be organized
along the three dimensions of aggregation, classi?cation and generalization (see [17]).
As described in [20], the act of “abstracting a collection of units into a new unit is
called aggregation”; indeed, an aggregation is a special type of association in which
objects or classes are assembled or con?gured together to create a more complex object
or class.
For example, the John object can be aggregated into the Pizza Pizza Sales Depart-
ment object and respectively, the Employee class can be aggregated into the Depart-
ment class. As we will describe in Section 3.1, we adopt the feature model [20] in order
to allow users to be ?exible during the aggregation activity.
The classi?cation dimension calls for each object or class to be an instance of one or
more generic classes or metaclasses. In fact, referring to the previous example, John
is an instance of Employee while Employee is an instance of the ActorClass metaclass.
The Classi?cation dimension is also used for relationships belonging to the model;
in fact, we can have an HomeAddress object which can be classi?ed by an Address
class which, similarly, can be classi?ed by an AddressClass metaclass. More details
about Classi?cation is provided in Section 3.1 which describes the four meta layer
metamodelling architecture of the Meta Object Facility (MOF) [30] and the features
inherited from Telos [29].
Classes and Metaclasses can be specialized along generalization or ISA hierarchies.
As de?ned in [20], generalization is the act of “abstracting the commonalities among
a collection of units into a new conceptual unit suppressing detailed di?erences”. For
example, an Employee class may have subclasses such as Clerk, Sales Person, etc.;
similarly, an ObjectClass metaclass may have submetaclasses such as EntityClass, Sit-
uationClass, etc. Notice that ISA hierarchies are orthogonal to the classi?cation di-
mension; therefore, all the above subclasses, i.e., Clerk, Sales Person, etc., should
be instances of the ActorClass.
In the next subsections, further details of aggregation, classi?cation, and general-
ization are provided. In particular, in Section 3.4 a possible implementation of the
abstraction mechanisms for the BIM metamodel is illustrated and described.
25
3.1 Aggregation: Mandatory, Optional, OR and
Alternative
The BIM provides a ?exible way for the aggregation activity that allows to have a
direct control in the choice of the parts (i.e., the “partOf“ relationship) constituting
an Object.
Indeed, the BIM aggregation mechanism uses the same approach described in [20]
in which the relationships between a parent and its children are categorized as:
• Mandatory – a child is required,
• Optional – a child is optional,
• Or – at least one of the children must be selected,
• Alternative (xor) – exact one of the children must be selected.
Figure 3.1 provides an example which shows the visual notation used to aggregate
di?erent Intentions. The example, also show how cardinalities can be used to add
more semantic during the aggregation activity; e.g., the “iPods produced” aggregates
from one to one million “one iPod produced”. Moreover, since the “one iPod produced”
is mandatory the range of cardinality must start from “one”.
Figure 3.1: An Intention aggregation hierarchy.
In Section 3.4, an underlying model for the aggregation mechanism is described.
3.2 Classi?cation: The OMG Four-Layer Metamodel
Architecture
The BIM metamodel is designed to be aligned with the OMG four-layer metamodel
architecture [30] which is summarized in Table 3.1.
26
Layer Description Example of model elements
M3: meta-
metamodel
De?nes the language for spec-
ifying metamodels.
MetaClass, MetaAttribute, MetaOp-
eration
M2: meta-
model
An instance of a meta-
metamodel. De?nes the lan-
guage for specifying a model.
Class, Attribute, Operation, Compo-
nent
M1: model An instance of a metamodel.
De?nes a language to de-
scribe an information do-
main.
StockShare, askPrice, sellLimi-
tOrder, StockQuoteServer
M0: user
objects (user
data)
An instance of a model. De-
?nes a speci?c information
domain.
< Acme Software Share 98789 >,
654.56, sell limit order,
< Stock Quote Svr 32123 >
Table 3.1: The OMG four-layer metamodel architecture.
In this architecture, a model at one layer is used to specify models belonging to the
layer below. Similarly, a model at one layer can be seen as an instance of a particular
model in the layer above.
Usually, models at M
n
layer have an higher level of abstraction and are typically
more compacts than models at M
n?1
layer. In fact, models at M
n?1
are more elaborate
than the models at M
n
layer that describe them. Figure 3.2 shows an example where
the top layers M3 and M2 of the architecture are represented and speci?ed, respectively,
by the MOF meta-metamodel and by the UML metamodel.
The BIM metamodel is at the same layer of the UML metamodel, i.e, at the M2
layer. Therefore, we would consider the BIM metamodel to be an instance of the MOF
meta-metamodel. We want this for two reasons: i) MOF enables the interoperability
of model and metadata driven systems; and, ii) MOF is quite spread across industry.
In this way, our model can exploit the interoperability provided by MOF in order to
facilitate its eventual integration with industry’s models and systems.
However, a major problem is that MOF (and UML) su?ers from the “shallow”
instantiation problem [6].
Basically, a class can only de?ne the semantic of its direct instances, but it has no
e?ect on entities created by further instantiation steps.
This is caused by the old “two-levels only” modeling philosophy which does not
adequately support a multi layer architecture. In fact, although model elements in a
multiple layer architecture can represent both objects and classes, i.e., an object at M2
layer can be seen as a class for objects at M1 layer, a class can never receive attributes
and associations from its classi?er, but only slots and links, thus leading to the shallow
instantiation problem (see [6] for further details).
The BIM metamodel needs to be able to in?uence both the M1 layer and the M0
layer in order to constrain designers in the choice of domain concepts and relationships
at M1 layer but also to propagate such “semantic” constrains (when requested) on
instances at M0 layer.
27
Figure 3.2: An example for the four-layer model architecture which uses MOF and
UML.
Figure 3.3 shows an example of such deep instantiation in which some “instance of”
links are missing to simplify the illustration. Notice how:
1. associations, such as “evaluates”, can be propagated across multiple layers;
2. some attributes, such as “currentValue”, can be propagated across multiple layers
while re?ning their types, e.g., the type of currentValue is NumberClass at M2
while is Integer at M1;
3. some attributes, such as “metric”, can be limited to speci?c layer, i.e., metric is
instantiated at M1 while is disappearing at M0;
4. some attributes, such as “director”, can be freely de?ned by the designer at M1
but are not speci?ed at M2, i.e., they are domain speci?c and are not mandatory
by the BIM metamodel.
In order to support such deep instantiation, the BIM is inspired by the Telos lan-
guage [29] in which classes, attributes, and associations
1
are collectively referred to
1
In Telos associations are represented using attributes which are binary relationships between enti-
ties, i.e. classes, or other relationships.
28
Figure 3.3: An example of the deep instantiation concept required in the BIM.
by the term “proposition” and are treated uniformly by the structuring mechanisms
of aggregation, classi?cation, and generalization. Therefore, as shown in Figure 3.4,
we can have metaclasses, classes, objects; but also: i) metattributes, attributes, slots;
and, ii) metassociations, associations, links
2
.
Figure 3.4: The Metaclass, Class and Objects instantiation.
A careful reader can observe that the UML metamodel introduces Instance meta-
classes in order to “link” objects of di?erent types at the M0 Layer. Although this can
2
In Telos is also possible to have meta-metaclasses, meta-metattributes, meta-metassociations and
so on (although for our aim it is not necessary).
29
resolve the instantiation of associations at the M0 layer, this approach arises issues
such as the “ambiguous classi?cation” and the “replication of concepts” [6]. Moreover,
it can increase the complexity of the model and lead to inconsistencies and the losing of
precision; and, it does not satisfy the “need” of propagating attributes across di?erent
layers.
3.3 Generalization: Subtypes and Business Terms
Specialization
The primitive types presented in Section 2 represent the model elements provided by
the BIM metamodel for the description of the di?erent concepts belonging to a generic
business environment. Moreover, in order to describe particular instances of the real
world, the BIM metamodel de?nes a set of subtypes whose semantic is described in
Table 3.2.
Moreover, in order to cover and map to business terminologies, such as a vision,
mission, or strategy, we use metaproperties such as (i) short-/long-term, (ii) many/few
instances, (iii) formal/informal de?nition, and (iv) chances of success. Clusters of
terms from a business glossary, such as Vision, Strategic/Tactical goal, Softgoal, Ob-
jective are then represented in terms of a single BIM primitive concept (Intention)
but each has di?erent combinations of values for the four metaproperties. For exam-
ple, a Vision is a long-term Intention without a formal de?nition, which is likely to
only have a few instances (usually one) whose chances of success are low (depending
on many uncertain factors).
To represent business terms, the BIM metamodel de?nes an attribute associated to
the ThingClass metaclass, called type which is inherited by all the other metaclass
in the metamodel to store speci?c terminology for each Thing. Table 3.3
3
provides
common clusters for some of BIM’s primitive concepts.
Figure 3.5 shows an example of subtypes and business terms specialization for the
Intention primitive type.
3.4 An UML Class Diagram for the BIM’s Abstraction
Mechanisms
In this section we present a possible UML class diagram for the BIM metamodel, with
respect to the aggregation, classi?cation, and generalization mechanisms described in
the previous sections.
The model is shown in Figure 3.6. The RefinementLinkClass, the NodeElementClass
and the AggregationClass are parts of the aggregation mechanism. In particular,
3
To de?ne the set of terms illustrated in Table 3.3 we analyzed both the scienti?c litera-
ture, e.g., the Business Motivation Model (BMM) [3], and the business world, e.g., the
www.businessdictionary.com [4] site; however, the business terminology can be easily customized
for the domain at hand.
30
Primitive
Type
Subtype Subtype Description
Intention Operational Intention
An atomic Intention which has a
very strict and clear logical criterion of
satis?ability and can be achieved by an
operational process or activity.
Qualitative Intention
An atomic Intention which has not
a clear-cut criterion for its satisfaction
and can be claimed only when there is
su?cient positive and little negative ev-
idences (or unsatisfaction in the oppo-
site case).
Actor Agent
Actor with concrete, physical manifes-
tations, such as a human individual.
We use the term agent instead of per-
son for generality, so that it can be used
to refer to human as well as arti?cial
(hardware/software agents). An agent
has dependencies that apply regardless
of what roles he/she/it happens to
be playing. These characteristics are
typically not easily transferable to
other individuals, e.g. its skills and ex-
periences, and its physical limitations
[1].
Role
Abstract characterization of the be-
havior of a social actor within some
specialized context or domain of en-
deavor. Its characteristics are easily
transferable to other social actors. The
dependencies associated with a role
apply regardless of the agent who plays
the role [1].
Position
Intermediate abstraction that can be
used between a role and an agent. It
is a set of roles typically played by one
agent (e.g., assigned jointly to that one
agent). We say that an agent occupies
a position. A position is said to cover a
role [1].
Table 3.2: Subtypes belonging to the BIM metamodel.
31
BIM Concept Business Terms
Intention Vision, Strategic/Tactical Goal, SoftGoal, Objective
Process Mission, Strategy, Tactic, Initiative, Business Process, Ac-
tivity
Actor Organization, Business Unit, Human person, System Appli-
cation
Resource Monetary / Infrastructure / Economic Good / Information
/ Human / Capability Resource
Directive Policy, Rule
Table 3.3: An example of business terms captured with the Thing’s attribute type.
Figure 3.5: An example of subtypes and business terms specialization for the
Intention primitive type.
the attribute mandatory of the RefinementLinkClass allows to specify if the “re-
?ner” component in the re?nement relationship is mandatory or optional; while, the
attribute type of the AggregationClass allows to specify the type of the aggregation,
i.e., OR, AND, XOR, Alternative, performed on the sub-components.
In regard to the generalization mechanism, the class named “...” (specializing the
ThingClass) represents the di?erent primitive types presented in Section 2. Instead,
the attribute type belonging to the ThingClass is used to specify the business termi-
nology shown in table 3.3.
Finally, the right part of the model represents the state and the evolution timeline
concepts describe in Subsection 2.4 and Subsection 2.5.
Figure 3.7 shows an example of how the model works underlying the feature model
visual notation (see Subsection 3.1).
32
Figure 3.6: An UML Class diagram for the BIM’s abstraction mechanism.
Figure 3.7: An example of aggregation using (b) the feature model visual notation and
(a) the underlying UML Class Diagram.
33
4 Strategic Analysis through Mappings
We illustrate how the richness and ?exibility of BIM can be used to represent widespread
strategic planning models. Moreover, since the ?nal aim of BIM is to support anal-
ysis activities for answering questions such as “What will happen next?” or “Where
exactly is the problem?” [8], we describe how BIM can be projected onto di?erent
analysis models.
In particular, the following target models are considered:
• a goal reasoning model based on a formal goal model [14],
• the SWOT analysis model [9],
• the Strategic Map [22] model,
• the Balanced SCorecards (BSCs)[21](and Key Performance Indicators [33]) model.
4.1 Goal (Intention) Reasoning: A Formal (Axiomatic)
Model
In the BIM, Intention analysis and reasoning are given a prominent role to help
stakeholders in the de?nition of their intentions and relationships among them, such
as con?icts and negative or positive contributions.
As we described in Section 2.1, the Intention primitive type can be used to de?ne
the hierarchy of the Vision, Goals, and Objectives of an organization in which nodes
can be connected by in?uence links.
In this section, we want to provide an underneath algorithm which enables the
reasoning on the Intentions belonging to such a hierarchy.
At this aim, we project the BIM toward the goal reasoning model described in [14].
In this work, the authors adopts a formal goal model to make the goal analysis process
concrete through the use of forward and backward reasoning. Notice that, the model
is used in the context of the Tropos methodology [2] which adopts the i
?
[41] modelling
framework for analyzing requirements (Early Requirements and Late Requirements
1
).
In particular, the formal model goals is used by the software engineer to cope with
qualitative relationships and inconsistencies among goals during the early requirements
phase.
1
The former is concerned with understanding the organizational context within which the system-
to-be will eventually function; the latter, on the other hand, is concerned with a de?nition of the
functional and non-functional requirements of the system-to-be.
34
The formal model description resides in the de?nition of the notions of goal graphs
and the axiomatic representation of goal relations. The goal graphs is de?ned trough a
set of goal nodes G
i
and of relations (G
1
, ..., G
n
)
r
? G over them, including the (n+1)-
any relations and, or and the binary relations +S, -S, +D, -D, ++S, - -S, +,-++, - -.
For a in depth description we remand to [14] while here we brie?y recall the intuitive
meaning of such relationships.
For and and or we have that:
• (G
1
, ..., G
n
)
and
? G means that G is satis?ed (resp. denied) if all G
1
, ..., G
n
are
satis?ed (resp. if at least one G
i
is denied);
• (G
1
, ..., G
n
)
or
? G means that G is denied (resp. satis?ed) if all G
1
, ..., G
n
are
denied (resp. if at least one G
i
is satis?ed);
For the other binary relationships, an example is provided by: G
2
+S
? G
1
(resp.
G
2
++S
? G
1
) means that if G
2
is satis?ed, then there is some (resp. a full) evidence
that G
1
is satis?ed, but if G
2
is denied, then nothing is said about the denial of G
1
.
To generalize the previous G
2
+S
? G
1
relationship, we said that, the “S” (resp. “D”)
symbol denotes the fact that the satis?ability (resp. deniability) value of the source
goal, e.g., G
2
, is propagated; the “+” (resp. “-”) symbol denotes the fact that the
propagation is positive (resp. negative), in the sense that satis?ability propagates
to satis?ability (resp. deniability) and deniability propagates to deniability (resp.
satis?ability).
Finally, the relations +, -, ++, - - are de?ned such that G
2
r
? G
1
is a shorthand for
the combination of the two corresponding relationships G
2
rS
? G
1
and G
2
rD
? G
1
, e.g.,
G
2
+
? G
1
is a shorthand for the combination of G
2
+S
? G
1
and G
2
+D
? G
1
. The ?rst
kind of relationships are called symmetric and the latter two asymmetric.
Now, a set of four distinct predicates over goals are introduced to be used with
ground axioms in order to reasoning on the goal model. They are: FS(G), FD(G)
and PS(G), PD(G); which mean, respectively, that there is (at least) full evidence
that goal G is satis?ed and that G is denied, and that there is at least partial evidence
that G is satis?ed and that G is denied. In their work, the authors provide a set of
ground axioms for the propagation rules which are soundness and completeness. An
example of of relation axiom is: G
2
+S
? G
1
: PS(G
2
) ? PS(G
1
).
Given a goal graph and an initial values assignment to some goals, the underlying
algorithm exploits the ground axioms for forward and backward reasoning tasks. In
particular, for the forward reasoning the assigned goals are called input goals (typically
the leaf goals) while for the backward reasoning the assigned goals are called target
goals (typically root goals).
The aim of the forward reasoning is the propagation of initial values (i.e., the input
goals) to all other goals of the graph; the user can look the ?nal values of the goals of
interest (i.e., the target goals).
Instead, the aim of the backward reasoning is the backward search of the possible
input values (i.e., the input goals) leading to some desired ?nal value (i.e.the target
values), under desired constrains, e.g., avoiding con?icts among goals.
35
In general, the forward reasoning is used for evaluating the impact of the adoption of
the di?erent alternatives with respect to the root goals; while, the backward reasoning,
is used to analyze goal models and ?nd the set of goals at the minimum costs that if
achieved can guarantee the achievement of the desired top goals and softgoals.
The algorithm of the formal goal model can be used within the BIM to allow such
reasoning. In fact, the goal relationships are accounted for within BIM through the
In?uenceClass metaclass illustrated in Figure 2.2.
Notice how the qualitativeStrength and the quantitativeStrengh allow, respectively,
to record the qualitative (e.g., + or - -) or quantitative (e.g., 0.7 or -0.3) strength of an
in?uence. The type attribute allows to specify whether the satis?ability or deniability is
propagated, i.e., S or D. The StateClass, which is inherited from the ObjectClass,
is used to record the four states associated to FS(G), FD(G), PS(G) and PD(G)
predicates.
Finally, the ResourceClass, can help in the backward “search“ when we desire
to ?nd the set of Intentions at the minimum cost that, if achieved, can guarantee
the achievement of the desired top Intentions. Indeed, the ResourceClass can
represent the monetary resource required for the achievement (through a Process) of
an Intention which is used in the minimum cost analysis
2
.
Figure 4.1 shows an example of Intentions reasoning with respect to the example
described in Figure 2.3.
Figure 4.1: An example of Intentions reasoning with BIM.
Notice how the formal model is used for both Situations and Intentions. In the
?gure, the semantic of the in?uence relationships is the following:
• the satis?ability of “Outsourcing advertising company hired” Situation is prop-
agated negatively (-S) to the “Cost decreased” Intention; this means that if the
former holds the latter is partial denied; nothing is said about the denial of the
“Outsourcing advertising company hired”;
2
Alternatively, a redundant attribute called cost can be added in the de?nition of the IntentionClass
metaclass.
36
• the satis?ability (resp. deniability) of “Best customers attracted and retained”
Intention is propagated positively (++) to the “Outsourcing advertising com-
pany hired” Situation; this means that if the former is satis?ed (resp. denied)
the latter holds (resp. does not hold);
• the satis?ability (resp. deniability) of “Best customers attracted and retained”
Intention is propagated positively (+) to the “More products sold” Intention;
this means that if the former is satis?ed (resp. denied) the latter is partial
satis?ed (resp. denied);
• the satis?ability of “Sta? need training” Situation is propagated negatively (-
S) to the “More products sold” Intention; this means that if the former holds
the latter is partial denied; nothing is said about the denial of the “Sta? need
training” Situation;
• the satis?ability of “Christmas season” Situation is propagated positively (++S)
to the “More products sold” Intention; this means that if the former holds the
latter is (at least partial) satis?ed; nothing is said about the denial of the “Christ-
mas season” Situation;
In order to show an example of forward reasoning on the model de?ned in Figure
4.1
3
, we input such a model in the same tool used in [14]. The result is as shown in
Figure 4.2.
Table 4.1 shows the results obtained by applying forward reasoning. The ?rst three
rows correspond to Situations, followed by three rows for the top Intentions and
three rows for the bottom Intentions with respect to the Intentions hierarchy. In
the table, three experiments are described through initial values (Init ) and ?nal values
(Fin) for satis?ability (S) and deniability (D) of Situations/Intentions. In partic-
ular, these values can be: full (F); partial (P); an empty cell when the corresponding
element is not involved in the reasoning; or, a question mark symbol (?) when a result
cannot be calculated.
A brief description of the experiments is the following:
• Exp 1 : The “Christmas season” Situation is satis?ed (see the F value for the
As result, we have that the “Shareholder value increased” is partial denied due
to the partial denying of the “Cost decreased” Intention.
• Exp 2 : The “Christmas season”, the “Sta? need training”, and “Focused on
career and skills development” initial values are set to full satis?ed (F). As result,
we have a full “Revenue increased” satisfaction (see below for the semantic of the
con?icts) but no information for the “Shareholder value increased” Intention
(represented by the question mark). This result is due to the fact that the ground
3
Notice that: i) we need to add an extra node (namely, “-”) to simulate the feature model approach
for the decomposition; ii) we need to select at least one of the OR sub-Intentions to properly use
the tool while preserving the semantic of our model.
37
Figure 4.2: An example of Intention reasoning using the tool described in [14].
axioms, in this case (G
1
, ..., G
n
)
and
? G, are not able to work with uncertainties
(see Subsection 4.4 for how to address such issues). In fact, we have no informa-
tion (see the question mark symbol ?) for the “Cost decreased” Intention.
• Exp 3 : The “Christmas season”, the “Sta? need training”, “Best customers
attracted and retained” and “Focused on career and skills development” initial
values are set to full satis?ed (F). As result, we have that the “Shareholder value
increased” is partial denied due to the partial denying of the “Cost decreased”
Intention and a con?ict (i.e., full satis?ed and partial denied) on the “Revenue
increased” Intention.
Therefore, in the three experiments we use di?erent strategies to satisfy the top
“Shareholder value increased” Intention which lead to di?erent results.
To conclude this section a ?nal observation regarding the in?uence from intentions
towards Situations must be made. In fact, an Intention can lead to (++) or
avoid/mitigate (- -) a Situation.
A clear example is shown in the analysis performed in Table 4.1 where the “Out-
sourcing advertising company hired” can hold as the result of the satisfaction of the
“Best customers attracted and retained” Intention; vice-versa, we have also that
“Sta? need training” Situation is avoided or mitigated by the satisfaction of “Fo-
cused on career and skills development”.
The latter is the semantic associated to the con?icting values (S=F and D=F) for
38
Situation / Intention Exp 1 Exp 2 Exp 3
Init Fin Init Fin Init Fin
S D S D S D S D S D S D
Outsourcing advertising
company hired
F F
Christmas season F F F F F F
Sta? need training F F F F F F
Shareholder value
increased
P ? P
Cost decreased P ? P
Revenue increased F F P F P
Best customers attracted
and retained
F F F F
Focused on career and
skills development
F F F F
More products sold F F P F
Table 4.1: A formal forward reasoning example.
the “Sta? need training” Situation which is propagated towards the“More products
sold” Intention. Notice also that, the formal model is not able to deal with with
uncertainty when some Intentions have not an initial value since the ground axioms
require a complete information for the reasoning algorithm.
Finally, a similar analysis for the backward reasoning can be performed using similar
experiments as shown in [14] both considering or not a cost criteria.
As a summary, we can said that the Intention reasoning model enables to:
• perform forward reasoning, in order to evaluate di?erent strategies for the satis-
faction of top Intentions elements;
• perform backward reasoning (considering also cost constrains), in order to eval-
uate the optimal input values leading to some desired ?nal value;
• perform analysis on Intention inconsistencies and con?icts in the Intention
hierarchy.
4.2 The SWOT Analysis with the BIM
The SWOT analysis [9] is a strategic planning method which is used to evaluate the
Strengths, the Weaknesses, the Opportunities, and Threats which are involved in
a business environment. The purpose of the analysis is to specify the goals of the
39
organization, business venture or project and identifying those internal and external
factors that are favorable and unfavorable to achieve these goals.
Since a scan of the internal and external environment covers an fundamental role in
the strategic planning process, the SWOT analysis can be considered as the ?rst stage
of such a process in which an organization is helped to focus on key issues.
Therefore, a SWOT analysis starts with the de?nition of a desired state of the world
in terms of a set of strategic goals. Than, the identi?cation of SWOTs with respect
to the these strategic goals is performed. The result is an essential information which
helps the decision makers in understanding the attainability of the selected strategic
goals given such SWOTs. If the goals are not attainable di?erent objectives must be
selected and the process repeated.
In detail, the description of SWOT factors is:
1. Strengths are resources and capabilities of an organization which can be used
as a basis for developing a competitive advantage since they are are helpful to
achieve the strategic goals;
2. Weaknesses are absence of (certain) strengths as resources and capabilities which
may be viewed as a weakness since they are are harmful to achieve strategic goals;
3. Opportunities are external conditions which can be helpful to achieve the strate-
gic goals since represent favorable circumstances for pro?t and growth;
4. Threats are external conditions, usually due to changes in the external environ-
ment, which can be harmful to the strategic goals.
The results of a SWOT analysis are often presented in the form of a matrix as
illustrated in Figure 4.3.
Notice how, some factors may be viewed as strengths/opportunities or weaknesses/threats
depending upon their impact on the organization’s goals, e.g, the opportunity or threat
“changing of customer tastes”.
Another way to use SWOT is for the matching and converting activities. The match-
ing is used to ?nd competitive advantages by “matching” the strengths to opportuni-
ties, while converting is the act of guide strategies in order to convert weaknesses or
threats into strengths or opportunities. Usually, if the threats or weaknesses cannot
be converted an organization should try to minimize or avoid them.
In particular, an organization can use a SWOT analysis to de?ne:
• S-O strategies, which pursue opportunities that ?t good to the organization’s
strengths.
• W-O strategies, which overcome or avoid weaknesses to pursue opportunities.
• S-T strategies, which identify ways to use organization’s strengths to reduce its
vulnerability to external threats.
• W-T strategies, which establish a defensive plan to avoid that organization’s
weaknesses accentuate external threats.
40
Figure 4.3: An example of SWOT matrix.
The BIM provides a formal way to perform the SWOT analysis since: i) allows to link
the SWOT factors directly to the strategic goals they impact upon; ii) allows a formal
reasoning on the set of strategic goals, SWOT factors and in?uences relationships
among them.
The latter can be very useful for the de?nition of S-O, W-O, S-T, W-T strategies
since make feasible the exploration of the di?erent alternatives relying on the forward
reasoning and backward reasoning approaches presented in Subsection 4.1.
As shown in Section 2.1, we use Situation to represent those internal and external
factors which can contribute positively or negatively to the achievement of Intention,
i.e., strategic goals.
Notice that, as described in Subsection 4.1, when a schema is de?ned, some Intention
can be introduced to mitigate or avoid some Situations and some (harmful) Situations
can arise due to the presence, in the schema, of speci?c Intentions.
Moreover, it must be said that, in the BIM, we characterized as strength, weakness,
opportunity or threat the “in?uence” that exist from a Situation to an Intention.
This allows to represent those cases in which the same Situation can represent, for
example, a strength with respect to an Intention while representing a weakness with
respect to another.
Table 4.2 shows how to map the SWOT in?uences to the formal model presented in
Subsection 4.1, while Figure 4.4 illustrates an example.
In the ?gure, the “More products sold” is de?ned to exploit the “Christmas season”
external opportunity. This opportunity is matched by the“E?cient and e?ective dis-
tribution channels” internal strength that allows to deal with the high demand during
41
SWOT In?uence Formal Model In?uence
Strength +S, ++S
Weakness -S, –S
Opportunity +S, ++S
Threat -S, - -S
Table 4.2: SWOT and formal model mapping.
Figure 4.4: A SWOT analysis example with BIM
Christmas.
Moreover, in order to avoid and mitigate weaknesses and threats, two strategic goals
are also de?ned in the schema, namely “Focused on career and skills development” and
“New set of products researched” strategic goals.
The former attempts: i) to mitigate the lack of Sta?’s skills in order to be prepared
for the Christmas; and ii) to reduce the organization vulnerability to the external threat
helping the Sta? to turn the customer’s taste toward the Organization’s products.
The latter, the “New set of products researched”, is de?ned and pursued as a de-
fensive plan to match the new customer’s taste.
4.3 De?ne Strategic Map, Balanced Scorecard and Key
Performance Indicators with BIM
Important instruments for strategic planning are Strategic Maps (SMs) [22] and Bal-
anced SCorecards (BSCs)[21]. The former are visual representation of the strategy of
an organization which shows organization plans used to achieve missions and visions.
In particular, a Strategic map illustrates the cause-and-e?ect relationships between
di?erent strategic goals and the associated measures, the key performance indicators
(KPIs).
These measures are included in the latter, the BSC, which represents a “balanced”
range of metrics against which to measure the Organization’s performance. The mean-
ing of “balance” is provided by the fact that the broader view of leading indicators
42
of performance includes also non-?nancial metrics, such as “learning and growth of
employees”, “customer satisfaction”, etc.
The combination of SMs and BSCs follows the principle of “you cannot measure what
you cannot describe”. In fact, SMs aim to describe the direction of an organization
while BSCs aim to de?ne a comprehensive set of performance measures that provides
the framework for a strategic measurement and a management system.
Both the SMs and BSCs describe and measure organizational performance across
four balanced perspectives: ?nancial, customers, internal business processes, and learn-
ing and growth (for their descriptions and further details see [21] and [22]). In general,
these perspectives, allow to see the organization and the business environment from
di?erent viewpoints and not only from the ?nancial aspects.
As described in [21], the four perspectives have been found to be robust across a
wide variety of companies and industries but should be considered a template. Indeed,
no mathematical theorem exists to proof that four perspectives are both necessary and
su?cient.
Within each of the four perspectives, the organization must de?ne the following
elements:
1. Strategic goals
4
– strategies which must be achieved in that perspective;
2. Measures – the progresses toward that particular strategic goals;
3. Targets – the target value sought for each measure;
4. Initiatives – what should be done to facilitate the achievement of the target;
5. Cause-e?ect relationships – in?uences among strategic goals (or measures).
Figure 4.5 illustrates an example of such elements in which only Targets are missing.
A typical target can be, for example, a value of $10,000 for the Revenue measure for
satisfying the “Revenue increased” strategic goal.
A common approach to evaluate the performance of an organization and how suc-
cessful it is in achieving short and long-term goals, is the use of KPIs [33]. KPIs are
quanti?able measurements which re?ect the performance of an organization towards
its goals. Therefore, BSCs can express measures and targets through a set of KPIs.
BIM integrates in a single conceptual framework the primitive concepts that charac-
terize SMs, BSCs and KPIs, as well as requirements models in Software Engineering.
Through projection mappings on a global BIM model, it is possible to obtain par-
tial models that can be analyzed through SM, BSC, KPI and formal goal reasoning
techniques [14] as described in previous sections.
Using the fragment in Figure 2.2 and the IndicatorClass described in Figure 2.8(a),
we are able to represent both SMs and BSCs (i.e., a set of KPIs). In particular,
Intentions and Indicators represent strategic goals, their measures and associated
targets. Processes with the type attribute set to Initiative (see Table 32) describe
4
We use the strategic goal term instead of the objective as used in the BSC.
43
initiatives used to reach targets. Instances of the Influence metaclass address cause-
e?ect relationships (both in quantitative and qualitative ways). Finally, the perspective
attribute helps to characterize elements along the four di?erent perspectives.
An example of such mapping is shown in Figure 4.5, corresponding to the model of
Figure 2.1.
Figure 4.5: The BestTech Strategic Map and Balanced Scorecard de?ned with BIM.
Notice how, the BIM model can represent a possible underneath formal schema for
the SM and BSC described in Figure 2.1. Therefore, SM and BSC should be used for
illustration purpose, since familiar to executives, middle managers, etc., while BIM
should be used to formalize such abstracted human-language to a machine-readable
language on which queries, in depth analysis, etc., can be performed.
In conclusion we can a?rm that, as SMs and BSCs do, the BIM is: i) a way of
providing a macro view of an organization’s strategy using the Intention primitive
type to describe strategic and tactical goals; and ii), a way of constructing metrics to
evaluate performance against these strategies using the Indicator primitive type.
However, at the contrary of SMs and BSCs, the BIM allows more in depth analysis
on the schema obtained after the designing activity.
4.4 Probabilistic Graphical Model for Intention
Reasoning
In Section 4.1, we described a solution based on formal logic model to provide a reason-
ing mechanism on Intentions. However, we also highlighted that such kind of model
44
is able to provide only partial results in condition of uncertainty. For example, we
would recall the experiment two in Table 4.1 in which a question mark symbol (?) was
introduced for the “Cost decreased” and “Shareholder value increased” Intentions.
The issue of treat with uncertainty is an inescapable aspect of most real-world
applications; indeed, it is quite common to have not a complete information during
an analysis activity. Future works for BIM, include the investigation of probabilistic
(graphical) models [23], which make the uncertainty explicit and provide models that
are more faithful to reality.
Probabilistic graphical models are approaches model-based which allow interpretable
models to be constructed and then manipulated by reasoning algorithms. These models
can be de?ned by an analyst or can be learned automatically from data in order
to facilitate their construction when a manual design is di?cult or even impossible.
Di?erent Probabilistic graphical models have been de?ned in the scienti?c community,
such as Bayesian networks, undirected Markov networks, In?uence Diagrams, etc. (see
[23] a comprehensive discussion).
One of our goals within BIM, is to adapt such models in order to manage uncer-
tainty to perform causal reasoning and decision making under such circumstances. In
particular, we are concentrating on Bayesian networks and in providing a ?rst step
toward the use of In?uence Diagrams.
45
5 A Case Study
In this section, we sketch a case study for BestTech Inc. for which we constructed
a complete BIM schema. Part of the schema is shown in Figure 5.1. This schema
provides a comprehensive description of the business and its environment, balanced
along the four perspectives discussed earlier. For example, from the Financial Per-
spective, the top-level intention is Shareholder value increased; one of its sub-intention
Cost decreased is further re?ned into Management cost decreased and Supply chain
cost decreased. In general, for each perspective, Intentions have their associated
Indicators, e.g., Market share for Market share increased (from the Customer per-
spective), and they are related to high-level processes (strategies), e.g., Rewards pro-
gram.
The BestTech schema can be queried by the business analyst to answer questions
such as “Which are the in?uencers and sub-intentions for Revenue Increased”, or
“Which are the Intentions whose performance is poor (red zone) and whose deadline
is at the end of the month”. Since data often resides in and scatters across databases,
such queries are translated through schema mappings into database queries, and the
answers are then translated back into business-level concepts. Schema mapping be-
tween a BIM schema and database schema is a ongoing research in our group. More-
over, this schema can be projected along di?erent views. An example is illustrated by
the SM of Figure 4.5 which is a useful view when communicating an organization’s
strategies with the BestTech executives. Moreover, if the need is to perform analysis,
we can project the schema towards a variety of analysis models, as discussed in Sec-
tion 4. With such projections, we can respond to queries such as “Show me all the
Intentions which are in con?ict with at least one other Intention” or “Show me the
impact of denial of the Marketing improved Intention”.
46
Figure 5.1: Part of the BestTech BIM schema.
47
6 Related Work
The use of business-level concepts—such as business objects, rules and processes—has
been researched widely for at least 15 years and is already practiced to some extent
in both Data Engineering and Software Engineering [38, 25, 19]. These e?orts have
more recently resulted in standards, e.g., OMG’s Business Process Modeling Notation
(BPMN) [32]. Such proposals focus on modeling objects and processes, with little
attention paid to objectives.
Enterprise modeling languages (enterprise ontologies, to some) have also been re-
searched for a long time, with the express intention of aligning business and IT con-
cerns. Examples of this line of research include TOVE [10], REA [26] and the Zachman
Framework for Enterprise Architecture [43], as well as TOGAF [39]. Of those, BMM [3]
is closest in spirit to BIM. Our proposal places the BIM concepts we adopted from
BMM on an ontological foundation adopted from DOLCE [11] and also integrates
those with state-of-the-art abstraction mechanisms.
Notably, our concept of Situation is akin to the notions of description and situation
proposed in [12], but the authors there envisioned semantic web applications, rather
than business ones.
The Zachman Framework for Enterprise Architecture is one of the oldest proposals
for enterprise modeling. The framework consists of a table of 5 rows and 6 columns.
The rows de?ne an IT system and its context from di?erent perspectives ranging from
scope (top row), to business model, information system model, technology model and
detailed description (lowest row). Each row of the table uses a di?erent language.
Columns de?ne common questions that need to be answered about each perspective:
what, how, where, who, when and why. The public part of the Zachman framework
consists of this table, with no stand taken on what notation or modeling method to
use. Issues of notation and method to use are addressed in the proprietary part. This
modeling framework has had considerable in?uence on enterprise modeling practice,
including recent work on Service-Oriented Architectures (SOAs). BIM ?ts within the
Zachman framework, focusing on the why column, but o?ers a di?erent set of primitive
concepts for capturing why concerns than other proposals in the literature.
As indicated in the introduction, the other modeling proposals that relate to our
work are i* [42], URN/GRL [18] and KAOS [7, 40], all from the general area of Goal-
Oriented Requirements Engineering. From these we have adopted intentional and
social concepts. These models lack primitive constructs for in?uence relationships,
indicators, and various types of situations integrated in the BIM modeling framework.
Recent proposals extending URN do include indicators [34], but BIM’s indicators are
more general and they can be used to measure any model object, including other
indicators.
From a business perspective, BIM models can capture what is commonly found in
48
Strategic Maps and Balanced Scorecards. They can also be mapped to other languages
that enable goal analysis and SWOT analysis, and we expect other mappings to prob-
abilistic frameworks such as Bayesian networks and Analytics [8] to enable reasoning
under uncertainties.
49
7 Conclusions
One important problem of Business Intelligence technologies is that information re-
quired and generated by such technologies is rarely explicitly linked to business con-
cepts, decisions and outcomes, and is therefore hard to interpret and use. In this report
we have proposed the Business Intelligence Model (BIM), as ?rst step towards bridg-
ing the gap between the worlds of business and data analytics. The proposed model
extends the notion of conceptual schema to accommodate business concepts such as
strategic objectives, business processes, in?uences, indicators, risks and trends. We
have showed, through examples taken from a case study how a BIM schema can sup-
port governance activities, including monitoring, auditing and analysis at the strategic
level. As mentioned before, for BIM to be useful, we also need technologies for trans-
lating queries speci?ed over a BIM schema into queries over database schemas, also for
translating answers back into business terms. Such work is being carried out within
the context of the strategic network for Business Intelligence, funded by the Natural
Sciences and Engineering Research Council (NSERC) of Canada
1
.
As for future work, along one direction, we are further evaluating and re?ning BIM
with a large scale, real-world case study. Along another, we are extending to cover
the tactical level of business organizations, and along a third, we plan to extend our
model to incorporate uncertainty in strategic modeling and analysis through the use
of Bayesian networks. This will enable BIM to support statistical decision making [23]
and will complement the logic-based analysis techniques currently within BIM’s scope.
Acknowledgments
This work was supported by the Business Intelligence Network (BIN) and NSERC.
We are grateful to G. Mussbacher, G. Richards, E. Yu and many others for useful
discussions.
1http://bin.cs.toronto.edu/home/index.php andhttp://www.nserc-crsng.gc.ca/Partners-Partenaires/Networks-Reseaux/BIN-RVE eng.asp
50
8 Appendix
8.1 BIM Taxonomy
Refer to Table 2.1 and Table 3.2 for the taxonomy’s description.
Figure 8.1: The BIM ’s taxonomy.
51
Bibliography
[1] istarquickguide.http://istar.rwth-aachen.de/tiki-
index.php?page=iStarQuickGuide, 2009.
[2] P. Bresciani, P. Giorgini, F. Giunchiglia, J. Mylopoulos, and A. Perini. Tropos:
An agent-oriented software development methodology. Autonomous Agents and
Multi-Agent Systems, 8(3):203–236, 2004.
[3] Business Rules Group. The business motivation model: Business governance in
a volatile world.http://www.businessrulesgroup.org/bmm.shtml, 2007. Release
1.3.
[4] BusinessDictionary.com. Businessdictionary.http://www.businessdictionary.com,
2009.
[5] P. P.-S. Chen. The entity-relationship model—toward a uni?ed view of data.
ACM Trans. Database Syst., 1(1):9–36, 1976.
[6] A. Colin and K. Thomas. The essence of multilevel metamodeling. In Proc. of
the 4th Int. Conf. on The UML, Model. Lang., Concepts, and Tools, pages 19–33,
London, UK, 2001. Springer-Verlag.
[7] A. Dardenne, A. van Lamsweerde, and S. Fickas. Goal-directed requirements
acquisition. Sci. Comput. Program., 20(1-2):3–50, 1993.
[8] T. H. Davenport and J. G. Harris. Competing on Analytics: The New Science of
Winning. Harvard Business School Press, March 6, 2007.
[9] T. R. Dealtry. Dynamic Swot Analysis. Dynamic Swot Associates, 1994.
[10] M. S. Fox. The tove project towards a common-sense model of the enterprise. In
IEA/AIE ’92: Proceedings of the 5th international conference on Industrial and
engineering applications of arti?cial intelligence and expert systems, pages 25–34,
London, UK, 1992. Springer-Verlag.
[11] A. Gangemi, N. Guarino, C. Masolo, A. Oltramari, and L. Schneider. Sweetening
ontologies with dolce. In Proc. of the 13th Int. Conf. on Knowl. Engineer. and
Knowl. Manage. Ontol. and the Semant. Web, pages 166–181, London, UK, 2002.
Springer-Verlag.
[12] A. Gangemi and P. Mika. Understanding the semantic web through descrip-
tions and situations. In Proceedings of ODBASE03 Conference, pages 689–706.
Springer, 2003.
52
[13] A. Gill. Introduction to the Theory of Finite-State Machines. McGraw Hill, 1962.
[14] P. Giorgini, J. Mylopoulos, and R. Sebastiani. Goal-oriented requirements analysis
and reasoning in the tropos methodology. Eng. App. Artif. Intel., 18:159–171,
2005.
[15] C. Heitmeyer. Software cost reduction. In Encyclopedia of Software Engineering,
Two Volumes. Wiley, January 2002.
[16] R. Howard and J. Matheson. In?uence diagrams. Readings on the Principles and
Applications of Decision Analysis, Vol. II, 1984.
[17] R. Hull and R. King. Semantic database modeling: survey, applications, and
research issues. ACM Comput. Surv., 19(3):201–260, 1987.
[18] International Telecommunication Union: Recommendation Z.151 (11/08). User
Requirements Notation (URN) – Language de?nition.http://www.itu.int/rec/T-
REC-Z.151/en.
[19] S. Jablonski. On the complementarity of work?ow management and business
process modeling. SIGOIS Bull., 16(1):33–38, 1995.
[20] K. Kang, S. Cohen, J. Hess, W. Novak, and A. Peterson. Feature-oriented domain
analysis (foda) feasibility study. Technical report, CMU/SEI-90-TR-021, SEI,
Carnegie Mellon University, November 1990.
[21] R. S. Kaplan and D. P. Norton. Balanced Scorecard: Translating Strategy into
Action. Harvard Business School Press, 1996.
[22] R. S. Kaplan and D. P. Norton. Strategy maps: Converting intangible assets into
tangible outcomes. Harvard Business School Press, 2004.
[23] D. Koller and N. Friedman. Probabilistic Graphical Models: Principles and Tech-
niques. The MIT Press, August 2009.
[24] N.-S. Koutsoukis and G. Mitra. Decision Modelling and Information Sys-
tems: The Information Value Chain, volume 26. Springer: Operations Re-
search/Computer Science Interfaces Series, 2003.
[25] P. Loucopoulos and E. Katsouli. Modelling business rules in an o?ce environment.
SIGOIS Bull., 13(2):28–37, 1992.
[26] W. E. McCarthy. The rea accounting model: A generalized framework for ac-
counting systems in a shared data environment. The Accounting Review, 1982.
[27] Metaphysics Research Lab. Stanford Encyclopedia of Philosophy.http://plato.stanford.edu/, 2010.
[28] H. Morris. Analytic applications and decision-centric bi. Information Management
Magazine, June 1 2004.
53
[29] J. Mylopoulos, A. Borgida, M. Jarke, and M. Koubarakis. Telos: representing
knowledge about information systems. ACM Trans. Inf. Syst., 8(4):325–362, 1990.
[30] Object Management Group. Meta-object facility (MOF) 2.0 speci?cation.http://www.omg.org/spec/MOF/2.0/, 2006.
[31] Object Management Group. Object constraint language.http://www.omg.org/spec/OCL/2.0/, May 2006.
[32] Object Management Group. Business process modeling notation (bpmn).http://www.omg.org/spec/BPMN/1.2/, January 2009. version 1.2.
[33] D. Parmenter. Key Performance Indicators. John Wiley & Sons, 2007.
[34] A. Pourshahid, D. Amyot, L. Peyton, S. Ghanavati, P. Chen, M. Weiss, and
A. J. Forster. Business process management with the user requirements notation.
Electronic Commerce Research, 9(4):269–316, 2009.
[35] R. L. Sallam and K. Schlegel. Overcoming the gap between Business Intelligence
and Decision Support.http://www.gartner.com/it/page.jsp?id=927913, 2009.
[36] P. K. Saxena. Principles of Management: A Modern Approach. Global India
Publications Pvt Ltd., 2009.
[37] P.-Y. Schobbens, P. Heymans, J.-C. Trigaux, and Y. Bontemps. Generic semantics
of feature diagrams. Comput. Netw., 51(2):456–479, 2007.
[38] J. Sutherland. Business objects in corporate information systems. ACM Comput.
Surv., 27(2):274–276, 1995.
[39] The Open Group. TOGAF 9 - The Open Group Architecture Framework Version
9, 2009.
[40] A. van Lamsweerde. Requirements engineering: From System Goals to UML
Models to Software Speci?cations. John Wiley & Sons, 2009.
[41] E. Yu. Modelling strategic relationships for process reengineering. PhD thesis,
University of Toronto, Dept. of Computer Science, University of Toronto., 1995.
[42] E. Yu. Towards modelling and reasoning support for early-phase requirements
engineering. In Proc. 3rd IEEE Int. Symp. on Requirements Engineering, IEEE
CS, Washington, USA, 1997.
[43] J. A. Zachman. A framework for information system architecture. IBM Systems
Journal, 26(3):277–293, 1987.
54
doc_835622236.pdf