Expert Systems with Applications - پارس پروژه · Strategy maps and the balanced scorecard ... amongst measures of different perspectives and objectives, ... show how value

Expert Systems with Applications 40 (2013) 1–14

Contents lists available at SciVerse ScienceDirect

Expert Systems with Applications

journal homepage: www.elsevier .com/locate /eswa

Fuzzy cognitive strategic maps in business process performance measurement

M. Glykas ⇑Department of Financial and Management Engineering, University of the Aegean, GreeceAegean Technopolis, The Technology Park of the Aegean Region, Chios, Greece

a r t i c l e i n f o

Keywords:Strategic simulationFuzzy Cognitive MappingBusiness metricsManagement tools

0957-4174/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.eswa.2012.01.078

⇑ Address: Department of Financial and Managemethe Aegean, Greece.

E-mail address: [email protected]

a b s t r a c t

This paper elaborates on the application of Fuzzy Cognitive Maps (FCMs) in strategy maps (SMs). Thelimitations of the Balanced Scorecards (BSCs) and SMs are first discussed and analyzed. The need forsimulated scenario based SMs is discussed and the use of FCMs as one of the best alternatives is pre-sented. A software tool for the development, simulation and analysis of FCM based SMs is also presented.The effectiveness of the resulting software tool and FCM theory in SMs is experimented in two case stud-ies in Banking.

� 2012 Elsevier Ltd. All rights reserved.

1. Strategy maps and the balanced scorecard

Organizations are operating in a continuously changing environ-ment. Market competition requires from management to continu-ously adapt their business objectives and revise strategic plans.Organizational performance measurement systems provide thelinkage between strategic goals and daily operations. Traditionalsolely financial based performance measure systems cannot longermeet management expectations. For the last decade managers andacademic researchers are focusing on frameworks, methodologiesand tools that provide integrated performance measurementsystems (PMSs) that analyze organizations from both financialand non-financial perspectives. The most notable example of thistype of PMSs, is the Balanced Score Card (BSC) (Kaplan & Norton,2004). It consists of four perspectives, financial perspective,customer perspective, internal process, and learning and innova-tion. Usually, 20 to 25 key performance indicators are allocated toeach of perspective.

The aim of the BSC is to link business objectives with operationalobjectives in a balanced way. The first version of BSC or also calledFirst Generation of BSC has many limitations: for example it con-tains a too simplistic unidirectional causality mechanism, it neglectsthe notion of cause and effect relationships in time; and it presentshigh level of vagueness in linking strategic and operational goals.

A big evolution for the BSC was the introduction of strategymaps (SMs) (Eccles & Pyburn, 1992). SMs focus on the causal-effectrelationships even amongst measures of different perspectives andobjectives, and the alignment of intangible assets. Strategy maps(SMs) represent visually relationships among the key components

ll rights reserved.

nt Engineering, University of

of an organization’s strategy (Eccles & Pyburn, 1992). We couldargue that SMs describe strategy in a picture; they are powerfultools which show how value is created through cause and effectrelationships. Kaplan and Norton argue that they create ‘‘the miss-ing link between strategy formulation and strategy execution’’(Kaplan & Norton, 2004).

Strategy maps are particularly helpful for:

� Promoting understanding and clarity of strategy.� Encouraging greater engagement and commitment to strategy.� Ensuring alignment of resources.� Identifying gaps or blind spots.� Making more effective and efficient use of resources.� Aligning remuneration with strategy – particularly in the soft

areas and where objectives have a duration >12 months.

A strategy map describes how an organization creates value byconnecting strategic objectives in explicit cause and effect relation-ships. They provide an excellent snapshot of strategy and aresupported by measurable objectives and initiatives.

Strategy maps enable organizations to (Lawson & Desroches,2007):

� describe strategy in a single picture.� Clarify strategies and communicate them to employees.� Identify the key internal processes which drive success.� Align investments in people, technology and organizational,

capital for maximum impact.� Expose gaps in strategies so that early corrective action can be

taken.� Identify explicit customer value propositions.� Map the critical internal processes for creating and delivering.

the value proposition.

http://dx.doi.org/10.1016/j.eswa.2012.01.078

mailto:[email protected]

http://dx.doi.org/10.1016/j.eswa.2012.01.078

http://www.sciencedirect.com/science/journal/09574174

http://www.elsevier.com/locate/eswa

2 M. Glykas / Expert Systems with Applications 40 (2013) 1–14

� Align human resources, information technology and organiza-tion culture to internal processes.

Strategy maps can be used for developing and reviewing strat-egy at organizational, departmental and even project level.

The strategy map (Kaplan & Norton, 2000, 2004, 2008) evolvedfrom the four-perspective model of the balanced scorecard, addinga visual dimension which improves clarity and focus.

There are five main principles behind strategy maps:

� strategy balances contradictory forces.� Strategy is based on a differentiated customer value

proposition.� Value is created through internal business processes.� Strategy consists of simultaneous complementary themes.� Strategic alignment determines the value of intangible assets.

Strategy maps are used in many frameworks as part of theirstrategy and change management offerings an example of a strat-egy map can be seen in the picture below.

The balanced scorecard (Kaplan & Norton, 2000, 2004, 2008) is aperformance management system that enables organizations toimplement a business vision and strategy.

2. Shortcomings of strategy maps

Although there may be benefits related to the design and use ofstrategy maps, a number of authors have highlighted possibleshortcomings (e.g. (Ahn, 2001; Buytendijk, 2008; Norreklit, 2000)).

2.1. Feedback loops

The development of strategy maps could be criticized as toomuch of an inward-looking exercise. Also, the cause-and-effectrelationships depict a one-way, linear approach often starting withthe ‘learning and growth’ perspective and culminating in financialresults instead of depicting non-linear, two-way linkages. Howeversince the Balanced Scorecard perspectives are not independent,feedback loops should be included in the maps (Franco & Bourne,2005).

2.2. Need for fuzziness in causal relationships

Predictions about the future state of a market and values thatbusiness goals and objectives can reach always contain the issueof uncertainty. Also the influence values of cause and effect rela-tionships in strategy maps contain by themselves the issue ofvagueness or fuzziness as more than one cause node can be linkedto the same effect node with different levels of influence. There is aneed therefore for a theory that will accommodate this fuzziness incausal relationships.

2.3. The missing element of time

Norreklit (2003) argues that strategy maps do not discriminateamongst logical and causal links. Typically, in many organizations,there are inconsistencies in the frequency of gathered values andthe range in which the values vary over a period of time.

Othman (2007) argues that a very serious drawback of strategymaps is the lack of representing the time evolution element in stra-tegic plans. This missing time element also influences the ability tomodel performance indicators in SMs.

2.4. Need for dynamic-flexible SMs

According to Buytendijk (2008) relying on a static SM over themid and long term, is equivalent to assuming not only that theorganization and its strategy will stay the same, but also that com-petitors will continue to behave in the same way. Furthermore, ifstrategy maps are supposed to have predictive abilities, one couldquestion the validity of analyzing past data to predict future states.

2.5. Need for tools with simulation capabilities

Currently there are no tools in the literature that provide simu-lation capabilities of composed-decomposed and linked strategicmaps. There are only tools that allow the composition of perfor-mance calculation of performance measures that are based on val-ues of other performance measures the values of which need to becalculated beforehand.

3. Strategic scenario simulations in SMs

A model is essentially an imitation of something real and simu-lation is a process of using a model to imitate the behavior ofsomething real. A model is only an imitation of reality and is notin itself reality. This is both an advantage, in that it is usually mucheasier to manipulate and understand the behavior of a model thanreality, and a disadvantage in that one must be careful in general-izing from conclusions drawn from examining the behavior of amodel.

Ackoff (1962), Ackoff (1979a) and Ackoff (1979b) suggest thatmodels fall into three categories: iconic, analogue and symbolic;he includes mathematical models under the symbolic category.

Based on the solution approach: analytical or simulation. Manymathematical problems cannot be solved analytically and simula-tion offers a means of ’solving’ such problems.

Simulation is a process of ’driving a model of a system with suit-able inputs and observing the corresponding outputs’ (Paul, Fox, &Schrage, 1987); in effect simulation is a means of experimentingwith a model of reality. Nelson and Winter (1982) see a computerprogram as ’a type of formal theoretical statement’ and simulationas ’a technique of theoretical explanation’.

Simulation is particularly useful where it is impossible, danger-ous or inordinately expensive to experiment with reality. This isgenerally speaking the case with real-life business firms or econo-mies and hence simulation offers the business researcher a meansof examining and experimenting with economic and businesssystems.

Phelan and Wigan (1995) point out that it is particularly diffi-cult for researchers in the strategy field to successfully carry outexperiments in order to determine the ‘laws’ of successful strategicmanagement. They suggest that three distinct difficulties arise:that of observation, manipulation, and replication; they go on tosuggest that simulation may assist strategy researchers in over-coming some of these difficulties.

The tools and technique of simulation can be useful to practis-ing managers in two ways: for training in strategy-making and foruse in actual strategy-making. Simulation is used as an aid fortraining managers in the art and science of strategy-making byuniversities and by management training consultants. Within aca-demia, simulations are increasingly becoming an integral part ofstrategy textbooks; It is interesting to note that the simulationsbeing used in universities have tracked the movement in theoriesunderpinning strategic thinking from industrial organization eco-nomics theory to the resource-based view to systems and com-plexity theories.

M. Glykas / Expert Systems with Applications 40 (2013) 1–14 3

Secondly, simulation can be used in the actual process of strat-egy-making. When examining the external environment duringthe analysis phase of strategy-making, simulation could be used todevelop and test various business scenarios (Johnson & Scholes,1999).

Simulation could also be used in the choice or matching phaseof strategy-making. This phase is rather underdeveloped in thestrategic management literature and relatively poorly supportedby formal techniques (Moroney, 1999). Simulation could assist ingenerating and evaluating strategic options. Other techniques fromthe field of management science such as the multi-criteria decisioncould also be applied to the choice-matching phase of the strategy-making process.

This use of management science techniques by strategists couldalso have the effect of improving the profile of management scienceand operations research. These functions are often seen by manag-ers suited only to operations analysis and are pushed ’further andfurther down in their organizations’ (Ackoff, 1962; Ackoff, 1979a).

Simulation can be carried out using several different mecha-nisms. Straightforward spreadsheet models may be used to carryout ‘what-if’ type analyses. Specialised simulation packages can beused to develop more sophisticated models. Customised modelscan also be developed using standard programming languages, thisis rare in business but common in engineering eg. custom develop-ment of software to simulate the behavior of networks (eg. gas,water, sewerage), structures (eg. buildings, bridges), mechanismsor vehicles.

Finally, simulation can be used as a research tool by businessacademics in the field of business and corporate strategy. Simula-tion offers a means of experimenting with virtual business organi-zations and economies, something that is not possible with realorganizations and economies.

There are six major tasks that comprise the strategic planningprocess (Ackoff, 1962): (1) goal formulation, (2) environmentalanalysis, (3) strategy formulation, (4) strategy evaluation, (5) strat-egy implementation, and (6) strategic control.

Goal formulation is concerned with the formulation of a set ofgoals for the organization. Environmental analysis includes identi-fication of environmental factors which seem relevant to some tar-get goals. Strategy formulation refers to a sequence of steps takenduring the period of time when resources, the operating environ-ment, and a set of goals are preliminarily identified. Strategy eval-uation is concerned with the suitability of strategy formulatedabove for present or future environments (situations). Strategyimplementation is essentially an administrative task and inher-ently behavioral in nature. Strategic control checks to see whetherthe strategy is being implemented as planned and the results pro-duced by the strategy are those intended. If deviation occurs, thenfeedback takes place and the strategic planning process recyclesthrough tasks (1) and (6) as indicated above.

3.1. Scenario analysis in SMs

In order to resolve the limitations of SMs Buytendijk (2008)argues that scenario analysis (Miller & Waller, 2003; Schoemaker,1995; Schwartz, 1991; Van Der Heijden, 2005; Wack, 1985) couldplay an important role in the design of strategy maps, as it is aneffective method to look at the future.

They believe that strategy maps should not be closed, static rep-resentations of strategy. Organizations and the environment theyoperate change continuously based on ‘PESTEL factors’ (Political,Economical, Social, Technological, Environmental and Legal). Inmany situations past performance can not be a source of dataanalysts can rely on.

According to Wack Scenario planning, assists managers andstrategists to explore different alternatives in present, intermedi-

ate and future desired states and even conclude in states thatwould seem unthinkable (Wack, 1985).

Strategy in fact is based on scenarios (Fink, Marr, Siebe, & Kuhle,2005), which in contrast to forecasting techniques, which aim toprovide answers about only the final future states, encourage peo-ple to pose questions about different pathways that can be fol-lowed (Van Der Heijden, 2005).

Scenarios can be used for a number of purposes (Schoemaker,1995):

1. identify early warning signals;2. assess the robustness of the organization’s core competencies;3. generate better strategic options;4. evaluate the risk/return of each option in view of the

uncertainties.

The combination of strategy maps and scenario analysis has anumber of advantages (building on Miller & Waller (2003), andKaplan & Norton (2008)):

� Strategy maps and scenarios are effective means to communi-cate the present and future strategy of an organization.� Both tools are built on a holistic view of the organization and its

environment, and on how key activities and processes areinterrelated.� The internal focus of strategy maps is complemented by focus of

scenario analysis on environmental factors.� Through strategy maps and scenarios, both qualitative and

quantitative aspects can be taken into account.� Both tools require the participation of several stakeholder

groups; this could increase the validity and robustness of theorganization’s strategy.� The development of strategy maps and scenarios also imply the

comparison of mental models (Senge, 1990) and the achieve-ment of intersubjective agreement between participants.� Finally, contingencies, uncertainties, trends, and opportunities,

which are seldom anticipated, could be identified and evaluatedthrough scenario analysis, incorporated in the strategy maps,and thus acted upon Miller and Waller (2003).� Although the joint development of scenarios and strategy maps

could result in major benefits, we are not suggesting that thecomplete process of developing a strategy map should be drivensolely by external influences as identified in various scenarios.

Even though the literature on strategy mapping is fairly vast,few authors have suggested the combination of strategy mapsand scenario analysis (Fink et al., 2005; Othman, 2007) and noneof them has described the actual design process and nobody usingany tools with simulation capabilities.

To address this issue, Buytendijk (2008) propose a series ofsteps to create a scenario-based strategy map:

1. Consider the strategy map and identify the strategic objectivesthat describe the assumptions for the business model. Forinstance, ‘cost leadership’ for a budget airline, or ‘ultimatesafety’ for a car manufacturer, or ‘superior service’ in a hotelchain.

2. Create different scenarios, for example using PESTEL analysis.Identify the new orð

3. unchangedÞ critical success factors in each of those scenarios.4. Create a strategy map with objectives for each of those scenar-

ios, based on the specifics of that scenario.5. Establish the commonality of objectives across the various sce-

narios. The more an objective is present across scenarios, themore ‘future-proof’2 such an objective will be, and the higherthe probability that these goals could be reached in a changing


environment. In order for this commonality analysis to work,objectives will have to be specific; this implies that predomi-nantly high-level objectives such as ‘‘maintain profitability’’and ‘‘seek growth’’ do not provide practical guidance and willmost likely only change in the gravest of discontinuities.

In the next sections we will present a tool that assists in thecreation, monitoring and simulation of strategic maps based onthe theory of Fuzzy Cognitive Maps. With the use of this tool strat-egists and managers can explore different strategic scenarios andsee the impact of their thinking and the evolution of these scenar-ios over time in a step by step simulation.

4. Fuzzy Cognitive Maps

Fuzzy Cognitive Maps is a modeling methodology for complexdecision systems, which originated from the combination of FuzzyLogic (Zadeh, 1965) and Neural Networks. An FCM describes thebehavior of a system in terms of concepts; each concept representsan entity, a state, a variable, or a characteristic of the system (Dick-erson & Kosko, 1997).

Kosko in Kosko (1986b) defined a concept Ci that constitutescausal relationships in FCM as

Ci ¼ ðQ i[ � Q iÞ \Mi

where Qi is a quantity fuzzy set and �Qi is a dis-quantity fuzzy set.�Qi is the negation of Qi. Each Qi and �Qi partitions the whole set Ci.Double negation �� Qi equals to Qi, implying that �Qi correspondsto Qc

i , the complement of Qi. However, negation does not meanantonym. Therefore, if a dis-quantity fuzzy set �Qi does not corre-spond to the complement of Qi, we will call it as anti-quantity fuzzyset to clarify the subtle meaning in the dis-quantity fuzzy set, asproposed by Kim and Lee (1998). Mi is a modifier fuzzy set thatmodifies Qi or �Qi concretely. The modifier fuzzy set fuzzily inter-sects the fuzzy union of a quantity fuzzy set and a dis-quantity fuz-zy set.

Kosko in Kosko (1986b) also formally defined the positive andnegative fuzzy causal relationships (or fuzzy causality) as follows.

Definition 1. Ci causes Cj iff (Qi \Mi) � (Qj \Mj) and (�Qi \Mi) �(�Qj \Mj)

Fig. 1. Simple FCM.

Definition 2. Ci causally decreases Cj iff (Qi \Mi) � (�Qj \Mj) and(�Qi \Mi) � (Qj \Mj)

Here ‘‘�’’ stands for fuzzy set inclusion (logical implication).A more insightful and practical definition of FCMs follows. FCM

nodes are named by concepts forming the set of conceptsC = {C1,C2, . . . ,Cn}. Arcs (Cj,Ci) are oriented and represent causallinks between concepts; that is how concept Cj causes concept Ci.Arcs are elements of the set A = {(Cj,Ci)ji � C � C. Weights of arcsare associated with a weight value matrix Wn�n, where each ele-ment of the matrix wji 2 [�1, . . . ,1] � R such that if (Cj,Ci) R A thenwji = 0 else excitation (respectively inhibition) causal link fromconcept Cj to concept Ci gives wji > 0 (respectively wji < 0). The pro-posed methodology framework assumes that [�1, . . .,1] is a fuzzybipolar interval, bipolarity being used as a means of representinga positive or negative relationship between two concepts.

In practice, the graphical illustration of an FCM is a signed graphwith feedback, consisting of nodes and weighted interconnections

(e.g. !Weight). Signed and weighted arcs (elements of the set A) con-

nect various nodes (elements of the set C) representing the causalrelationships that exist among concepts. This graphical representa-tion (e.g.Fig. 1) illustrates different aspects in the behavior ofthe system, showing its dynamics (Kosko, 1986b) and allowing

systematic causal propagation (e.g. forward and backward chain-ing). Positive or negative sign and fuzzy weights model the expertknowledge of the causal relationships (Kosko, 1991). Concept Cj

causally increases Ci if the weight value wji > 0 and causally de-creases Ci if wji < 0. When wji = 0, concept Cj has no causal effecton Ci. The sign of wji indicates whether the relationship betweenconcepts is positive (Cj !

Wj;iCi) or negative (Cj !

Wj;i � Ci), while the va-lue of wji indicates how strongly concept Cj influences concept Ci.The forward or backward direction of causality indicates whetherconcept Cj causes concept Ci or vice versa.

Simple variations of FCMs mostly used in business decision-making applications may take trivalent weight values [�1,0,1].This paper allows FMCs to utilize fuzzy word weights like strong,medium, or weak, each of these words being a fuzzy set to providecomplicated FCMs. In contrast, (Kwahk & Kim, 1999) adopted onlya simple relative weight representation in the interval [�1, . . . ,1].To this extend, research Kwahk and Kim (1999) offered reducedfunctionality since it did not allow fuzzy weight definitions.

Generally speaking FCM concept activations take their value inan activation value set V = {0,1} or {�1,0,1} if in crisp mode or[�d,1] with d = 0 or 1 if in fuzzy mode. The proposed methodologyframework assumes fuzzy mode with d = 1. At step t 2 N, each con-cept Cj is associated with an inner activation value at

j 2 V , and anexternal activation value et

aj2 R. FCM is a dynamic system.

Initialization is a0j ¼ 0. The dynamic obeys a general recurrent

relation atþ1 ¼ f ðgðeta;W

T atÞÞ, "t P 0, involving weight matrixproduct with inner activation, fuzzy logical operators (g) betweenthis result and external forced activation and finally normalization(f). However, this paper assumes no external activation (hence nofuzzy logical operators), resulting to the following typical formulafor calculating the values of concepts of FCM:

atþ1i ¼ f

Xn

j¼1;j – i

wjiatj

!ð1Þ

where atþ1i is the value of concept Ci at step t+1, at

j the value of theinterconnected concept Cj at step t, wji is the weighted arc from Cj toCi and f: R ? V is a threshold function, which normalizes activations.Two threshold functions are usually used. The unipolar sigmoidfunction where k > 0 determines the steepness of the continuousfunction f ðxÞ ¼ 1

1þe�kx. When concepts can be negative (d < 0), func-tion f(x) = tanh(x) is used.

To understand better the analogy between the sign of theweight and the positive/negative relationship, it may be necessaryto revisit the characteristics of fuzzy relation (Kaufmann, 1975;


Lee, Kim, Chung, & Kwon, 2002). A fuzzy relation from a set A to aset B or (A,B) represents its degree of membership in the unit inter-val [0,1]. Generally speaking, sets A and B can be fuzzy sets. Thecorresponding fuzzy membership function is lf : AxB ? [0,1].Therefore, lf(x,y) is interpreted as the ‘‘strength’’ of the fuzzymembership of the fuzzy relation (x,y) where x 2 A and y 2 B. Thenthis fuzzy relation concept can be denoted equivalently as x!lf

yand applied to interpret the causality value of FCM, since wji (thecausality value of the arc from nodes Cj to Ci) is interpreted asthe degree of fuzzy relationship between two nodes Cj and Ci.Hence, wji in FCMs is the fuzzy membership value lf(Cj,Ci) andcan be denoted as Cj!

wj;iCi.

However, we understand that the fuzzy relation (weight) be-tween concept nodes is more general than the original fuzzy relationconcept. This is because it can include negative (�) fuzzy relations.Fuzzy relations mean fuzzy causality; causality can have a negativesign. In FCMs, the negative fuzzy relation (or causality) between twoconcept nodes is the degree of a relation with a ‘‘negation’’ of a con-cept node. For example, if the negation of a concept node Ci is notedas� Ci, then lf(Cj,Ci) = �0.6 means that lf(Cj,� Ci) = 0.6. Conversely,lf(Cj,Ci) = 0.6 means that lf(Cj,� Ci) = �0.6.

FCMs help to predict the evolution of the system (simulation ofbehavior) and can be equipped with capacities of hebbian learning(Kosko, 1986a, 1998). FCMs are used to represent and to model theknowledge on the examining system. Existing knowledge of thebehavior of the system is stored in the structure of nodes and inter-connections of the map. The fundamental difference between FCMsand a Neural Networks is in the fact that all the nodes of the FCMgraph have a strong semantic defined by the modeling of the con-cept whereas the nor input/nor output nodes of the neural networkhave a weak semantic, only defined by mathematical relations.

4.1. Applications of fuzzy cognitive maps

Over the last years, a variety of FCMs have been used for captur-ing - representing knowledge and intelligent information in engi-neering applications, for instance, GIS (Liu & Satur, 1999) and faultdetection (e.g. Ndouse & Okuda, 1996; Pelaez & Bowles, 1995). FCMshave been used in modeling the supervision of distributed systems(Michael, 2010; Stylios, Georgopoulos, & Groumpos, 1997). Theyhave also been used in operations research (Craiger, Goodman,Weiss, & Butler, 1996), web data mining (Hong & Han, 2002; Leeet al., 2002), as a back end to computer-based models and medicaldiagnosis (e.g. Georgopoulos, Malandraki, & Stylios, 2002).

Several research reports applying basic concepts of FCMs havealso been presented in the field of business (Xirogiannis & Glykas,2004) and other social sciences (Michael, 2010). Research in Axel-rod (1976) and Perusich (1996) have used FCM for representingtacit knowledge in political and social analysis. FCMs have beensuccessfully applied to various fields such as decision making incomplex war games (Klein & Cooper, 1982), strategic planning(Diffenbach, 1982; Ramaprasad & Poon, 1985), information retrie-val (Johnson & Briggs, 1994) and distributed decision process mod-eling (Zhang, Wang, & King, 1994). Research like (Lee & Kim, 1997)has successfully applied FCMs to infer rich implications from stockmarket analysis results. Research like (Lee & Kim, 1998) alsosuggested a new concept of fuzzy causal relations found in FCMsand applied it to analyze and predict stock market trends. Theinference power of FCMs has also been adopted to analyze the com-petition between two companies, which are assumed to use differ-ential games mechanisms to set up their own strategic planning(Lee & Kwon, 1998). FCMs have been integrated with case-basedreasoning technique to build organizational memory in the fieldof knowledge management (Noh, Lee Lee, Kim, Lee, & Kim, 2000).Recent research adopted FCMs to support the core activities ofhighly technical functions like urban design (Xirogiannis,

Stefanou, & Glykas, 2004). Summarizing, FCMs can contribute tothe construction of more intelligent systems, since the more intel-ligent a system becomes, the more symbolic and fuzzy representa-tions it utilizes.

In addition, a few modifications have been proposed. For exam-ple, the research in Silva (1995) proposed new forms of combinedmatrices for FCMs, the research in Hagiwara (1992) extended FCMsby permitting non-linear and time delay on the arcs, the researchin Schneider, Schnaider, Kandel, and Chew (1995) presented amethod for automatically constructing FCMs. More recently, Liuand Satur (1999) have carried extensive research on FCMs investi-gating inference properties of FCMs, proposed contextual FCMsbased on the object-oriented paradigm of decision support andapplied contextual FCMs to geographical information systems(Liu, 2000).

4.2. Updated FCM algorithm

This paper extends the basic FCM algorithm (Kwahk & Kim,1999) by proposing the following new FCM algorithm:

atþ1i ¼ f k1at

i þ k2 �Xn

j¼1;j–i

wjiatj

!ð2Þ

This paper assumes that coefficients k1 and k2 can be fuzzy sets.Coefficient k1 represents the proportion of the contribution of

the value of the concept ai at time t in the computation of the valueof ai at time t + 1. In practice, this is equivalent to assume thatwii = k1. The incorporation of this coefficient results in smoothervariation of concept values during the iterations of the FCM algo-rithm. Coefficient k2 expresses the ‘‘influence’’ of the intercon-nected concepts in the configuration of the value of the conceptai at time t + 1. It is the proposal of this paper that such a coefficientshould be used to align indirectly causal relationships (essentially,the value of concept Ci) with the influence of concept Cj as follows:

� If the set of identified performance concepts Cj, j – i, is incom-plete (e.g. incomplete maps, missing concepts, etc), then theestimation of the value of concept Ci may prove imprecise. Inthis case coefficient k2 may indicate the sufficiency of the setof concepts Cj, j – i, in the calculation of the value of the conceptCi.� If the information necessary to approximate the input values of

concepts Cj, j – i, is incomplete (e.g. incomplete estimation ofbad loans), then the estimation of the value of concept Ci mayalso prove imprecise. In this case coefficient k2 may indicatethe completeness of information utilized in the approximationof the input values of concepts Cj during the calculation of thevalue of the concept Ci.

Ideally, coefficient k2 could break down into two separate coef-ficients (say k2 ¼ x � kx

2 þ y � ky2), where kx

2 aligns indirectly thevalue of concept Ci with the completeness of the set of conceptsCj (e.g. completeness of P&L performance indicators), while ky

2

aligns indirectly the value of concept Ci with the completeness ofavailable information for concepts Cj within the enterprise (e.g lackof information for certain P&L indicators). Parameters x, y couldpresent the relative importance of kx

2 and ky2 in mixed interconnec-

tion problems (e.g. incomplete set of P&L indicators with partialaccounting results). However, preliminary experiments showedthat this separation imposed unnecessary initialization overheadswithout increasing significantly the accuracy of the FCM algorithm.

In contrast to the basic FCM algorithm adopted by mostrelevant research practices the updated one may suit better thefinancial domain because:


� Coefficient k2 ‘‘normalizes’’ the FCM calculations based onincomplete information sets. The updated algorithm supportsbetter the qualitative and trend-based financial planning, pro-viding less conservative results. This normalization provesimportant at business domains mainly because:� it relaxes the need for extra calculations of error margins as a

result of incomplete background information,� it provides reasonable decision modeling approximations with-

out requiring extensive background financial analysis (e.g. com-plete estimation of bad loans, identification of all financialconcept links, etc).� Coefficient k1 results in smoother variation of concept values

during the iterations of the algorithm.

5. FCM-based strategy map scenarios in financial planning

In this section we present the use of the methodology and toolcalled FCM Modeler (Chytas, Glykas, & Valiris, 2010; Glykas, 2004;Glykas, 2010; Glykas & Xirogiannis, 2005; Xirogiannis & Glykas,2004; Xirogiannis & Glykas, 2007, 2004; Xirogiannis et al., 2004;Xirogiannis, Chytas, Glykas, & Valiris, 2008; Xirogiannis, Glykas,& Staikouras, 2010) in SMs in a real case studies in two bankinginstitutions in which the proposed mechanism focuses on supple-menting a typical financial strategy methodology by providing aholistic evaluation framework based on banking Profit & Loss(P&L) performance indicators augmented by external environmentstimuli. In practice the mechanism supplements the recurringfeedback loop between the current financial status, the future stra-tegic objectives and the action plans for improving profitability(the action plan, in turn, affects the future financial status). Thismechanism creates a strategy-level utilization of P&L model basedon a qualitative and trend-based technique. The proposed mecha-nism actually generates two financial assessment flows:

� Flow A: ‘‘Current financial status analysis ? Financial objec-tives’’ to estimate the gap between existing (‘‘as-is’’) and future(‘‘to-be’’) profitability and support the establishment of objec-tives which should bridge this gap.� Flow B: ‘‘Action plans ? Financial objectives’’ to estimate the

impact of strategic change actions to the evolution of financialstatus, assess anticipated financial maturity and align financialobjectives to meet any potential deviations.

During the third phase of this typical financial strategy formu-lation exercise, the top management of the bank sets the overallfinancial performance targets (measured by associated metrics).

Fig. 2. Inherent relationships between fin

These targets are exemplified further to action plan performancetargets (measured by tactical financial metrics) and then to opera-tional financial performance targets (measured by operationalfinancial metrics). All such targets and metrics present inherentrelationships. In practice, overall financial strategy metrics mustcascade to tactical financial metrics to allow the middle manage-ment to comprehend inherent relations among the different man-agerial levels of the bank. Similarly, tactical financial metrics mustpropagate up the overall financial metrics. While P& L inheritanceis usually clear, external stimuli with no apparent relationships tofinancial indicators are not always well defined.

This paper proposes the utilization of such indicators (Fig. 2) todevelop the FCMs and reason about the impact of strategic posi-tioning changes to the desired (‘‘to-be’’) financial models. The pro-posed mechanism utilizes FCMs to interpret:

� financial metrics (P&L indicators, external stimuli, etc) asconcepts (graphically represented as nodes),� decision weights as relationship weights (graphically repre-

sented as arrowhead lines),� decision variables as financial concept values,� hierarchical decomposition (top–down decomposition) of finan-

cial metrics to constituent sub-metrics as a hierarchy of FCMs.This interpretation allows the stakeholders to reason aboutlower level FCMs first (constituent financial metrics indicators)before they reason about higher-level metrics (affected metrics).

The proposed mechanism supports reasoning about the overallor partial financial strategy implementation using indicators fromthe P&L philosophy. In contrast to Kwahk and Kim (1999), the pro-posed mechanism builds on hierarchical metrics interrelationshipsidentified and utilized by the financial strategy formulation meth-odology. The proposed approach does not perform or guide theimplementation of any stage of the strategy formulation methodol-ogy. Also, the approach does not perform or guide the estimation ofthe absolute value of any of the financial metrics and/or the overallP&L performance. It only allows the stakeholders to reason aboutthe qualitative state of financial maturity metrics using fuzzy lin-guistic variables like high–neutral-margin, high-neutral-lowimpact of loans volume to income, etc.

5.1. FCM map modeling in financial scenarios

The following basic steps demonstrate the utilization of FCMs inex-ante financial modeling and planning, that is modeling informa-tion flow ‘‘Action plans ? Financial objectives’’.

ancial strategy and FCM hierarchies.


� Step 1 – Financial planning: Assume that a bank has alreadydeveloped a strategic-level financial plan.� Step 2 – Skeleton FCMs: FCMs are constructed to interpret stra-

tegic objectives and actions to skeleton maps. Skeleton FCMspresent concepts and links with no value assignments, in prac-tice, generic P&L interconnections.� Step 3 – Weight value assignment: Bank experts are asked to

provide linguistic weight variables for the map developedduring the second step. Different weight value assignmentsgenerate different business cases for the same skeleton map.� Step 4 – Quantify potential changes: Stakeholders assign fuzzy

linguistic input values to P&L concepts, to quantify potentialactions.� Step 5 – Simulation: The FCM model simulates the impact of

strategic changes (actions) to the financial objectives.

Similar steps demonstrate flow ‘‘Current financial status analy-sis ? Financial objectives’’.

The following generic example demonstrates financial model-ing using FCMs.

� Step 1 – Financial planning: Assume that a bank has alreadydeveloped a simple strategic-level financial plan. Assume thatan objective asks for increased profitability. Also, assume thatthe financial plan proposes the increase of sales volume as thesole action to achieve this objective.� Step 2 – Skeleton FCMs: Utilization of the technique to interpret

strategic objectives and actions to skeleton maps.� Step 3 – Weight value assignment: Experts provide linguistic

weight variables which are interpreted to weight values(Fig. 3, LHS)� Step 4 – Quantify strategic changes: Stakeholders assign input

values to concept ‘‘sales volume’’ (Fig. 3, RHS)� Step 5 – Simulation: The FCM model simulates the impact of

sales volume changes to the financial objectives and outputsthe estimated effect.

Fig. 3 depicts a graphical example with no feedback loops fol-lowed by sample numerical calculations using formula (2). Thisexample assumes that k1 = k2 = 1 and k = 5 as the steepness of thenormalization function. Setting the input variable of ‘‘sales vol-ume‘‘ to ‘‘positively medium’’ (defuzzified to 0,5 or 50%) triggersthe FCM formula (1st case). A zero external concept value indicatesthat the concept remains neutral, waiting for causal relationshipsto modify its current value. A generic interpretation of the first caseindicates that a ‘‘positively medium’’ increase in sales volume in-creases the income 88% and the profitability by 95%. In contrast,

Savol

Inc

Profit

Con

Savol

Inc

Profit

Con

Fig. 3. Sample FCM calculation

a ‘‘positively very low increase’’ in ‘‘sales volume’’ (defuzzified to0,2 or 20%) increases the income and the profitability by 59% and89%, respectively (2nd scenario).

Fig. 4 presents a typical example of a feedback loop. Similarly toFig. 3, changing the external input value of ‘‘sales volume’’ triggersthe FCM formula. However, the feedback loop dictates that calcu-lations stop only when an equilibrium state for all affected con-cepts has been reached, modifying all input values accordingly.

6. Decomposition of business goals and objectives at differentorganizational levels

Every organization has to set its strategy and objectives whichin turn specify the goals to cascaded down the organizational hier-archy for each division, department, organizational unit up to thelevel of the individual manager. At this latter level Critical SuccessFactors for the achievement of corporate objectives and goals playa very important role. These factors are finally represented as mea-surable performance measures in job descriptions. In real life per-formance measurement and monitoring problems that affectorganizational performance may arise. These problems are usedas a feedback loop for readjusting goals and objectives in all orga-nizational problems.

For issues of clarity we provide the following definitions on theconcepts used in the strategic planning process: (Fig. 5).

1. Critical Success Factors (CSFs): the limited numbers of areas inwhich satisfactory results will ensure successful competitiveperformance for the individual, department, or organization.CSFs are the few key areas where ‘‘things must go right’’ for thebusiness to flourish and for the manager’s goals to be attained.

2. Strategy: the pattern of missions, objectives, policies, and sig-nificant resource utilization plans stated in such a way as todefine what business the company is in (or is to be in) andthe kind of company it is or is to be. A complete statement ofstrategy will define the product line, the markets and marketsegments for which products are to be designed, the channelsthrough which these markets will be reached, the means bywhich the operation is to be financed, the profit objectives,the size of the organization, and the ‘‘image’’ which it will pro-ject to employees, suppliers, and customers.

3. Objectives: general statements about the directions in which afirm intends to go, without stating specific targets to be reachedat particular points in time.

4. Goals: specific targets which are intended to be reached at agiven point in time. A goal is thus an operational transformationof one or more objectives.

Change case B

Change case A

Change case B

Change case A

0

0

0,2

0

0

0.5

External input value (t=0)

0,9561

0,8807

0.5

Current Value

0,2les ume

0,5986ome

0,8904ability

cept Change case B

Change case A

Change case B

Change case A

0

0

0,2

0

0

0.5

External input value (t=0)

0,9561

0,8807

0.5

Current Value

0,2les ume

0,5986ome

0,8904ability

cept

s with no feedback loop.

Change case B

Change case A

Change case B

Change case A

0

0

0,2

0

0

0.5

Initial input (t=0)

0,9666

0.9623

0,81

Current Value

0,5128Sales

volume

0,8860Income

0,9569Profitability

Concept Change case B

Change case A

Change case B

Change case A

0

0

0,2

0

0

0.5

Initial input (t=0)

0,9666

0.9623

0,81

Current Value

0,5128Sales

volume

0,8860Income

0,9569Profitability

Concept

Fig. 4. Sample FCM calculations with feedback loop.

Fig. 5. Overview of the strategic planning process.


5. Measures: specific standards which allow the calibration of per-formance for each critical success factor, goal, or objective. Mea-sures can be either ‘‘soft’’ – i.e., subjective and qualitative – or‘‘hard’’ – i.e., objective and quantitative.

6. Problems: specific tasks rising to importance as a result ofunsatisfactory performance or environmental changes. Prob-lems can affect the achievement of goals or performance in aCSF area.

6.1. Decomposition of balanced scorecards

Following the decomposition hierarchy presented in the previ-ous section it is evident that if we use the balance scorecard tech-nique in performance measurement then the same type ofdecomposition should be followed.

The picture below depicts the balanced scorecard decompositionin a real life case study of a retail banking division of a Bank (Fig. 6).

The division contains two departments. The Micro business and theIndividuals. The high level balanced scorecard for the Retail bankingdivision contains measures that are calculated fully or partially atlower level balanced scorecards in the other two divisions.

In the same way the measures and strategy maps for the retailbanking division should be also decomposed in a similar way.

6.2. Decomposed FCMs-SMs

The current implementation of the proposed methodology toolencodes generic maps that can supplement the maturity modelingby storing concepts under different map categories, namely:

� Business category: all concepts relating to core financialactivities.� Social category: all personnel related financial concepts and

external stimuli concepts.� Technical category: all concepts relating to infrastructure and

technology related expenses.� Integrated category: all top-most concepts (e.g. a concept Ci

with no backward causality such that "j:wji = 0), or conceptswhich may fall under more than one main categories.

The dynamic nature of the approach allows easy reconfigura-tion. Further P&L indicators may be added, while concepts maybe decomposed further to comply with specialized analyses ofthe bank. This categorization is compatible with the P&L view ofthe bank to allow greater flexibility in modeling dispersed financialflows. The hierarchical decomposition of concepts generates a setof dynamically interconnected hierarchical maps (Fig. 7: samplemaps and map links).

Currently, the mechanism integrates more than 250 concepts,forming a hierarchy of more than 10 maps. The dynamic interfaceof the mechanism lets its user to utilize a sub-set of these conceptsby setting the value of the redundant ones and/or the value of theirweights to zero. Concepts and weight values have been obtained asfollows: (Fig. 8).

� Strategic-level financial plans of two typical (though major) E.U.commercial banks have been selected and analyzed.� The interpretation of financial plans into FCM hierarchies using

the technique presented has generated the skeleton maps.� FCM hierarchies comply with the typical P&L reports of most

financial sector enterprises.� Financial sector and bank experts have provided weight values

for skeleton FCMs.

Fig. 6. Decomposed balanced scorecards.

Fig. 7. Sample maps and map links.


It should be noted that as a working hypothesis for the FCM-based financial modeling, this paper adopted the operational andfinancial characteristics of typical/average European Union (EU)commercial banks, currently the majority of financial sector enter-prises. Therefore, experts assigned values under the assumptionthat the market for intermediated finance was characterized byrelationship rather than arm’s length lending. Also, experts consid-ered that fact that both banks operated in a bank-oriented (ratherthan a market-oriented) environment, in which banks predomi-nated as financial intermediaries by collecting savings (throughdeposits) and providing the bulk of external funding to the non-financial sector.

6.3. Decomposed scenarios with FCM map linking

A typical example of the interconnection mechanism in nowcommented briefly. In Fig. 8 we present a user defined decomposedmap hierarchy further analyzed in (Figs. 9 and 10).

Consider now maps ‘‘PL analysis’’ and ‘‘Interest income’’. Linkinga concept, which is defined into two maps, generates a hierarchy.Fig. 9 presents the system interface for the generation of the hierar-

chical relationship between maps ‘‘PL analysis’’ and ‘‘Interest in-come’’. Map ‘‘PL analysis’’ decomposes further concept ‘‘netinterest income’’ by using this concept as the link to map ‘‘Interestincome’’.

Fig. 9 also presents how algorithm (2) is decomposed tointegrate hierarchical links. The proposed system can portray thefinancial model following either a holistic or a scalable approach.This is analogous to seeing the bank either as a single, ‘‘big bang’’event or as an ongoing activity of setting successive financial tar-gets to selected banking operations. The proposed mechanism canaccommodate both approaches. Essentially, the implementationcan decompose financial concepts to their constituent parts (subconcepts) on demand and let the user reason about lower levelhierarchies of FCM before it passes values to the higher-levelhierarchies. The proposed mechanism also allows the user to spec-ify the degree of FCM decomposition during the map traversal(Fig. 10). Instead of waiting for a lower level FCM to traverse itsnodes and pass its value to higher level map hierarchies, theuser may assign directly an external value to nodes which linkhierarchies. In practice, the simulation is carried out as if thereare no links with other maps.


The following table summarizes the available variations of theproposed FCM algorithm, which encode dynamic map decomposi-tion and user-defined decomposition bound.

Fig. 8. Sample map hierarchy.

Mapdecomposition

FCM algorithm hierarchical calculations

Higher levelhierarchy

Lower levelhierarchy

Unrestricted atþ1i ¼ f ðk1at

i þ k2

�ðwhiath þ

Pnj¼1;j–i;j–hwjiat

j ÞÞat

h ¼ f ðk1at�1h þ k2

�Pn

j¼1;j–hwjhat�1j Þ

User-defined,no externalvalueassigned

atþ1i ¼ f ðk1at

i þ k2

�ðwhiath þ


j ÞÞat

h ¼ f ðk1at�1h Þ

User-defined,externalvalueassigned

atþ1i ¼ f ðk1at

i þ k2

�ðwhiath þ


j ÞÞat

h 2 [�1, . . . ,1]

Fig. 9. Interconnecti

Also the current implementation allows:

� easy customization of the function fand easy re-configuration ofthe formula Atþ1

i to adapt to the specific characteristics of indi-vidual enterprises,� generation of scenarios for the same skeleton FCM,� automatic loop simulation until a user-defined equilibrium

point has been reached. Alternatively, step-by-step simulation(with graphical output of partial results) is also available to pro-vide a justification for the partial results.

The proposed framework exemplifies further financial perfor-mance by decomposing maps into their consistent concepts. Thefollowing sections exhibit sample (though typical) skeleton mapsfor all categories, which provide relevance and research interestto this paper.

7. Financial planning scenarios in two case studies

The proposed tool was used extensively in two Retail Banks.Several planning scenarios were developed and simulated withthe assistance and supervision of knowledge experts.

The majority of the financial concepts of these examples cas-cade to several constituent metrics. This allows the mechanismto express its reasoning capabilities by traversing complicated con-cept interrelations spreading over different maps and hierarchies.As an example, the FCM mechanism was asked to support the fol-lowing decision problem: ‘‘is a certain change in interest rates andadministration expenses and consumer charge, and FUM, etc, goingto have a certain change in the operating income’’. The first exper-iment involves a bank with limited retail market share seeking topenetrate the customer base of its competitors. The second exper-iment involves a bank with an established retail market presenceseeking to enhance further its customer base. The following tablepresents sample actions/objectives, the associated nodes and theirinput/desired values.

For both cases, the Quanta tool iterated a subset of approxi-mately 120 concepts spread over 10 sample hierarchical maps inorder to calculate their equilibrium values.

on mechanism.

Fig. 10. User-defined m

Change action Associatednode

Input value– Bank A

Input value– Bank B

Increase FUM FUM 0.65 0.9Increase deposits Deposits 0.8 0.65Increase

consumerP&L charge

ConsumerP& L charge

0.1 0.35

Increase SBLmargin

SBL margin 0.5 0.68

Increase otheradministrationexpenses

Otheradministrationexpenses

0.68 0.8

Increase interestrates etc

Interest ratesetc

0.58 etc. 0.68 etc.

Objective Associatednode

Desired value– Bank A

Desired value– Bank B

Increaseoperatingincome

Operatingincome

>Positivelyhigh) > 0.65


Increaseprofit fromoperations

Profit fromoperations

>Positivelyverylow) > 0.21



Desiredvalue

FCMestimation

Expertestimation


Operatingincome


0.698 0.780

Increaseprofitfromoperations

Profitfromoperations


0.120 0.100


Desiredvalue

FCMestimation

Expertestimation


operatingincome


0.644 0.730

Increaseprofitfromoperations

Profitfromoperations


0.246 0.216


Fig. 11 compares two decision values for a set of nodes as esti-mated by the FCM mechanism and the team of experts respectivelyfor the first bank.

Given the input values and Fig. 11, the following table presentsthe estimated impact of the change actions. These sample calcula-tions indicate that if the actions are implemented as planned thenthe impact to the first objective will exceed the expectations. Onthe other hand, the impact to the second objective will not meetthe expectations. Fig. 11 also presents the values of other affectednodes for comparison purposes.

Similarly, Fig. 12 compares decision values as estimated by theFCM mechanism and the team of experts respectively for the sec-ond bank.

Given the input values and Fig. 12, the following table presentsthe estimated impact of the change actions. These sample calcula-tions indicate that if the actions are implemented as planned thenthe impact to the first objective will marginally fail the expecta-tions. The impact to the second objective will meet the expecta-tions. Fig. 12 also presents the values of other affected nodes forcomparison purposes.

8. Discussion

8.1. Theoretical & practical value

The proposed modeling mechanism and the developed tool hascontributed on managing performance in the two case studies in

ap decomposition.

0.0

0.2

0.4

0.6

0.8

1.0

oper

atin

g inco

me

prof

it fro

mop

erat

ions

oper

ating

exp

ense

s

core

inco

me

retai

l pro

visio

ns

admin

expe

nses

FUM

mut

ualfu

nds

lend

ing a

ctivit

ies

income fr

om b

onds

SBLvo

lum

e

SMEs l

oan v

olum

e

SBLmor

tgag

e m

argin

FX inco

me

inter

estra

te

SE inde

x

divid

end

incom

e

whole

sale

pro

visio

ns

Values estimated by FCMs

Values estimated by experts

Fig. 11. Financial performance – Bank A.

0.0

0.2

0.4

0.6

0.8

1.0

oper

ating

inco

me

prof

it fro

m o

pera

tions

oper

ating

exp

ense

s

core

inco

me

reta

il pro

visio

ns

adm

in ex

pens

esFUM

mut

ual fu

nds

lend

ing a

ctivit

ies

inco

me fr

om b

onds

SBL vol

ume

SMEs l

oan

volum

e

SBL mor

tgag

e m

argin

FX inco

me

inte

rest

rate

SE inde

x

divid

end

inco

me

whole

sale

prov

ision

s

Values estimated by FCMs

Values estimated by experts

Fig. 12. Financial performance – Bank B.


real life. From a theoretical foundation point of view the theorydeveloped enhances previous attempts as it:

� allows the introduction of fuzziness in SMs with the use ofFuzzy Cognitive Maps,� enhances the theory of Fuzzy Cognitive Maps through the prop-

osition of a new FCM algorithm,� introduces the notion of linked nodes in different SMs which

creates SM hierarchies,� allows the introduction of simulation capabilities in FCM

calculations and thus resulting in simulated scenarios in SMswith the time variant inherited in it,� allows dynamic scenario decomposition and reconfiguration,

and thus dynamic SMs

Overall the proposed theoretical framework has provided solu-tions to most of the SM problems stated in the literature byresearchers and proves that FCMs are one of the most suitable the-ories for Sms

8.2. Added value

Having established the theoretical and practical value of theproposed mechanism, it is useful to discuss also the added valueof incorporating such a mechanism into SMs. It is the belief of thispaper that the resulting tool provides real value to SM projects. Forexample:

� This decision aid mechanism proposes a new approach to sup-plementing SMs. It provides a qualitative though ‘‘intelligent’’support during the business analysis and objectives composi-tion phases of typical strategy formulation projects. It utilizescognitive modeling and offers a strategy-level utilization ofSMs in order to shift focus from quantitative analysis to strat-egy-level impact assessment.� The main purpose of this approach is to drive strategic change

activities for continuous improvement rather than limit itselfto qualitative simulations.


� The mechanism eases significantly the complexity of derivingexpert decisions concerning strategic planning. Informal exper-iments indicated that the time required by experts to estimatemanually the extensive impact of major strategic changes torealistic SMs could impose considerable overheads. On theother hand the elapsed time for automated estimations usingFCM decision support can be insignificant, onc strategy formu-lation projects should involve continuous argument of strategicchange options (e.g. application of best practices, alternativescenarios, alternative customer focus, etc) until an equilibriumsolution has been agreed by all stakeholders. Informal discus-sions with the principle beneficiaries and stakeholders of thetwo financial planning projects revealed that the proposedFCM decision support can reduce significantly the estimationoverheads of financial maturity, letting the stakeholders focuson the actual planning exercise while exploring in depth allalternatives and controlling effectively major strategic changeinitiatives.� The proposed mechanism can also assist the post strategic and

operational planning evaluation of the enterprise on a regularbasis. FCMs may serve as a back end to performance scorecards(e.g. Bourne, Mills, Wilcox, Neely, & Platts, 2000; Kaplan &Norton, 1996, 2001) to provide holistic strategic performanceevaluation and management. However a detailed analysis ofthis extension falls out of the scope of this paper.

8.3. Preliminary usability evaluation

Senior managers of the two major financial sector enterpriseshave evaluated the usability of the proposed tool and have identi-fied a number of benefits that can be achieved by the utilization ofthe proposed FCM tool as a methodology framework for financialplanning. Detailed presentation of the usability evaluation resultsfall out of the scope of the paper. However, a summary of majorbusiness benefits (as identified by senior managers) is providedto improve the autonomy of this paper:

� Shared Goals– Concept-driven simulated SMs pulls individuals together by

providing a shared direction and determination of strategicchange.

– Shared SMs and performance measurement enables businessunits to realize how they fit into the overall business modelof the bank and what their actual contribution is.

– Senior management receives valuable inputs from the busi-ness units (or the individual employees) who really compre-hend the weaknesses of the current strategic model as wellas the opportunities for performance change.

� Shared Culture– Managers at all business units feel that their individual con-

tribution is taken under consideration and provide valuableinput to the whole change process.

– All business units and individuals feel confident and opti-mistic; they realize that they will be the ultimate beneficia-ries of the planning exercise.

– The information sharing culture supports competitive strat-egy and provides the energy to sustain this by exploitingfully the group and the individual potential.

� Shared Learning– Top management realizes a high return from its commit-

ment to its human resources.– There is a constant stream of improvement within the

company.– The company becomes increasingly receptive to strategic

changes, since the benefit can be easily demonstrated toindividual business units.

� Shared Information– All business units and individuals have the necessary infor-

mation needed to set clearly their objectives and priorities.– Senior management can control effectively all aspects of the

strategic re-design process– The bank reacts rapidly to threats and opportunities.– It reinforces trust and respect throughout the bank.

Summarizing, experimental results showed that FCM-based exante reasoning of the impact of strategic changes (actual or hypo-thetical) to the status of business performance can be effective andrealistic. This is considered to be a major contribution of the pro-posed methodology tool to actual strategic change exercises.

9. Conclusion

This paper presented a supplement to SMs based on FCMs. Thisdecision aid mechanism proposes a new approach to supplement-ing current status analysis and objectives composition phases oftypical strategy formulation projects, by supporting fuzzy cogni-tive modeling and ‘‘intelligent’’ reasoning of the anticipated impactof strategic change initiatives to business performance. The mech-anism utilizes the fuzzy causal characteristics of FCMs as a newmodeling technique to develop a causal representation of dynamicSM principles in order to generate a hierarchical network of linkedperformance indicators based on critical success factors.

This paper discussed the FCM approach in putting realistic andmeasurable objectives in strategic planning projects and presentedsample maps with causal relationships. Preliminary experimentsindicate that the mechanism does not provide fundamentallydifferent estimates than expert decisions. Moreover, the decompo-sition of financial metrics into their constituent parts supportedreasoning of the strategic performance roadmap. The main purposeof the mechanism is to drive strategic change activities rather thanlimit itself to qualitative simulations. Moreover, the proposedmechanism should not be seen as a ‘‘one-off’’ decision aid. It shouldbe a means for setting a course for continuous improvement (Lang-bert & Friedman, 2002).

Future research will focus on conducting further real lifeexperiments to test and promote the usability of the tool, but alsoto identify potential pitfalls. Furthermore, future research will fo-cus on the automatic determination of appropriate fuzzy sets (e.g.utilizing pattern recognition, mass assignments, empirical data,etc) for the representation of linguistic variables to suit each par-ticular project domain. Finally, further research will focus onimplementing backward map traversal, a form of adbuctive rea-soning (Flach & Kakas, 1998). This feature offers the functionalityof determining the condition(s) Cij that should hold in order to in-fer the desired Cj in the causal relationship Cij!

wjkCk. Incorporating

performance integrity constraints reduces the search space andeliminates combinatory search explosion. Backward reasoninghas been tested extensively in other applications and its integra-tion in the proposed methodology framework may provebeneficiary.

References

Ackoff, R. L. (1962). The scientific method: Optimising applied research decisions.Wiley.

Ackoff, Russell L. (1979a). The future of operational research is past. Journal of theOperational Research Society, 30(2), 93–104.

Ackoff, Russell L. (1979b). Resurrecting the future of operational research. Journal ofthe Operational Research Society, 30(3), 189–199.

Ahn, H. (2001). Applying the Balanced Scorecard concept: An experience report.Long Range Planning, 34, 441–461.

Axelrod, R. (1976). Structure of decision: The cognitive maps of political elites.Princeton, NJ: Princeton University Press.


Bourne, M., Mills, J., Wilcox, M., Neely, A., & Platts, K. (2000). Designing,implementing and updating performance measurement systems. InternationalJournal of Operations and Production Management, 20, 754–771.

Buytendijk, F. (2008). Performance leadership. New York: McGraw-Hill.Chytas, P., Glykas, M., & Valiris, G. (2010). Software reliability modelling using fuzzy

cognitive maps. Studies in Fuzziness and Soft Computing, 247, 217–230 [FuzzyCognitive Maps].

Craiger, J. P., Goodman, D. F., Weiss, R. J., & Butler, A. (1996). Modelingorganizational behavior with fuzzy cognitive maps. Journal of ComputationalIntelligence and Organisations, 1, 120–123.

Dickerson, J. A., & Kosko, B. (1997). Virtual worlds as fuzzy cognitive maps. In B.Kosko (Ed.), Fuzzy engineering (pp. 125–141). Upper Saddle River, NJ: Prentice-Hall.

Diffenbach, J. (1982). Influence diagrams for complex strategic issues. StrategicManagement Journal, 3, 133–146.

Eccles, R. G., & Pyburn, P. J. (1992). Creating a comprehensive system to measureperformance. Management Accounting, 74(4), 41–44.

Fink, A., Marr, B., Siebe, A., & Kuhle, J. (2005). The future scorecard: Combiningexternal and internal scenarios to create strategic foresight. ManagementDecision, 43, 360–381.

Flach, P. A., & Kakas, A. (1998). Abduction and induction in AI: Report of the IJCAI97workshop. Logic Journal of the Interest Group on Pure and Applied Logic, 6(4),651–656.

Franco, M., & Bourne, M. (2005). An examination of the literature relating to issuesaffecting how companies manage through measures. Production Planning &Control, 16(2), 114–124.

Georgopoulos, V., Malandraki, G., & Stylios, C. (2002). A fuzzy cognitive mapapproach to differential diagnosis of specific language impairment. Journal ofArtificial Intelligence in Medicine, 679, 1–18.

Glykas, M. (2004). Work flow and process management in printing and publishingfirms. International Journal of Information Management, 24(6), 523–538.

Glykas, Michael (2010). Fuzzy cognitive maps: Advances in theory, methodologies,tools and applications. Studies in Fuzziness and Soft Computing, 247, 23.

Glykas, M., & Xirogiannis, G. (2005). A soft knowledge modeling approach forgeographically dispersed financial organizations. Soft Computing – A Fusion ofFoundations, Methodologies and Applications, 9(8), 579–593.

Hagiwara, M. (1992). Extended fuzzy cognitive maps. In Proceedings of the 1st IEEEinternational conference on fuzzy systems, New York.

Hong, T., & Han, I. (2002). Knowledge based data mining of news information of theinternet using cognitive maps and neural networks. Journal of Expert Systemswith Applications, 23(1), 1–8.

Johnson, R. J., & Briggs, R. O. (1994). A model of cognitive information retrieval forill-structured managerial problems and its benefits for knowledge acquisition.In 27th Annual Hawaii international conference on system sciences, Hawaii.

Johnson, G., & Scholes, K. (1999). Exploring corporate strategy: Text and cases (5thed.). Prentice Hall.

Kaplan, R. S., & Norton, D. P. (1996). Using the balanced scorecard as astrategic management system. in Harvard Business Review, January/February,75–85.

Kaplan, R., & Norton, D. (2000). Having trouble with your strategy? Then map it.Harvard Business Review, 167–176.

Kaplan, R. S., & Norton, D. P. (2001). Leading change with the balanced scorecard.Financial Executive, 17, 64–66.

Kaplan, R. S., & Norton, D. P. (2004). Strategy maps: Converting intangible assets intotangible outcomes. Boston, MA: Harvard Business School Press.

Kaplan, R., & Norton, D. (2008). Execution premium. Boston, MA: Harvard BusinessSchool Press.

Kaufmann, A. (1975). Introduction to the theory of fuzzy subsets. Fundamentaltheoretical elements. New York: Academic Press.

Kim, H. S., & Lee, K. C. (1998). Fuzzy implications of fuzzy cognitive map withemphasis on fuzzy causal relationship and fuzzy partially causal relationship.Fuzzy Sets and Systems, 97, 303–313.

Klein, J. C., & Cooper, D. F. (1982). Cognitive maps of decision makers in a complexgame. Journal of Operation Research Society, 33, 63–71.

Kosko, B. (1986). Differential hebbian learning. In AIP Conference proceedings 151.Kosko, B. (1986b). Fuzzy cognitive maps. Journal of Man-Machine Studies, 24,

65–75.Kosko, B. (1991). Neural networks and fuzzy systems. Englewood Cliffs: Prentice Hall.Kosko, B. (1998). Hidden patterns in combined and adaptive knowledge networks.

International Journal of Approximate Reasoning, 2, 377–393.Kwahk, K. Y., & Kim, Y. G. (1999). Supporting business process redesign using

cognitive maps. Decision Support Systems, 25(2), 155–178.Langbert, M., & Friedman, H. (2002). Continuous improvement in the history

of human resource management. Journal of Management Decision, 40(8),782–787.

Lawson, Hatch, & Desroches (2007). Scorecard best practices: Design, implementation,and evaluation. Wiley & Sons.

Lee, K. C., & Kim, H. S. (1997). A fuzzy cognitive map-based bi-directional inferencemechanism: An application to stock investment analysis. Journal of IntelligentSystems in Accounting Finance & Management, 6(1), 41–57.

Lee, K. C., & Kim, H. S. (1998). Fuzzy implications of fuzzy cognitive map withemphasis on fuzzy causal relationship and fuzzy partially causal relationship.Journal of Fuzzy Sets and Systems, 3, 303–313.

Lee, K. C., Kim, J. S., Chung, N. H., & Kwon, S. J. (2002). Fuzzy Cognitive Map approachto web-mining inference amplification. Journal of Expert Systems withApplications, 22, 197–211.

Lee, K. C., & Kwon, O. B. (1998). A strategic planning simulation based on cognitivemap knowledge and differential game. Journal of Simulation, 7(5), 316–327.

Liu, Z. Q. (2000). Fuzzy cognitive maps: Analysis and extension. Tokyo: Springer.Liu, Z. Q., & Satur, R. (1999). Contexual fuzzy cognitive map for decision support in

geographical information systems. Journal of IEEE Transactions on Fuzzy Systems,7, 495–507.

Michael, G. (2010). Fuzzy cognitive maps: Advances in theory, methodologies, tools andapplications. 978-3-642-03219-6. Berlin: Springer.

Miller, K., & Waller, H. (2003). Scenarios, real options and integrated riskmanagement. Long Range Planning, 36, 93–107.

Moroney, Mike (1999). Strategy evaluation: Towards an updated paradigm. IrishBusiness and Administrative Research, 21(1), 103–129.

Ndouse, T. D., & Okuda, T. (1996). Computational intelligence for distributed faultmanagement in networks using fuzzy cognitive maps. In IEEE internationalconference communication converging technology tomorrow’s application, NewYork.

Nelson, R., & Winter, S. (1982). An evolutionary theory of economic change. BelknapPress.

Noh, J. B., Lee Lee, K. C., Kim, J. K., Lee, J. K., & Kim, S. H. (2000). A case- basedreasoning approach to cognitive map driven –driven tacit knowledgemanagement. Journal of Expert Systems with Applications, 19, 249–259.

Norreklit, H. (2000). The balance on the balanced scorecard – a critical analysis ofsome of itsassumptions. Management Accounting Research, 11(1), 66–88.

Norreklit, H. (2003). The balanced scorecard: What is the score? A rhetoricalanalysis of the balanced scorecard. Accounting, Organizations and Society, 28,591.

Othman, R. (2007). Enhancing the effectiveness of the balanced scorecard withscenario planning. International Journal of Productivity and PerformanceManagement, 57, 259–266.

Paul, B., Fox, B. L., & Schrage, L. E. (1987). A guide to simulation (2nd ed.). Springer-Verlag.

Pelaez, C. E., & Bowles, J. B. (1995). Applying fuzzy cognitive maps knowledgerepresentation to failure modes effect analysis. In IEEE annual reliabilitymaintainability symposium, New York.

Perusichm, K. (1996). Fuzzy cognitive maps for political analysis. In Internationalsymposium on technology and society. Technical Expertise and Public Decisions,New York.

Phelan, S. E., & Wigan, M. R. (1995). Using simulation for theory generation instrategic management. In 2nd Australasian conference in strategic management,20–21 April 1995, La Trobe University, Melbourne.

Ramaprasad, A., & Poon, E. (1985). A computerised interactive technique formapping influense diagrams (MIND). Strategic Management Journal, 6(4),377–392.

Schneider, M., Schnaider, E., Kandel, A., & Chew, G. (1995). Constructing fuzzycognitive maps. In IEEE conference on fuzzy systems, New York.

Schoemaker, P. (1995). Scenario planning: A tool for strategic thinking. SloanManagement Review, 25–40.

Schwartz, P. (1991). Learnings long view. Doubleday.Senge, P. (1990). The fifth discipline: The art and practice of the learning organization.

Doubleday.Silva, P. C. (1995) New forms of combinated matrices of fuzzy cognitive maps. In

IEEE International Conference on Neural Networks, New York.Stylios, C. D., Georgopoulos, V. C., & Groumpos, P. P. (1997). Introducing the theory

of fuzzy cognitive maps in distributed systems. In 12th IEEE internationalsymposium on intelligent control, Istanbul, Turkey.

Van Der Heijden, K. (2005). Scenarios: The art of strategic conversation. Chichester:John Wiley & Sons.

Wack, P. (1985). Scenarios: Shooting the rapids. Harvard Business Review, 63(6),139–150.

Xirogiannis, G., Chytas, P., Glykas, M., & Valiris, G. (2008). Intelligent impactassessment of HRM to the shareholder value. Expert Systems with Applications,35(4), 2017–2031.

Xirogiannis, G., & Glykas, M. (2004). Fuzzy cognitive maps in business analysis andperformance-driven change. IEEE Transactions on Engineering Management,13(17), 111–136.

Xirogiannis, G., & Glykas, M. (2004). Fuzzy causal maps in business modeling andperformance-driven process re-engineering. Lecture Notes in Computer Science,3025, 331–341 [Methods and Applications of Artificial Intelligence].

Xirogiannis, G., & Glykas, M. (2007). Intelligent modeling of e-business maturity.Expert Systems with Applications, 32/2, 687–702.

Xirogiannis, G., Glykas, M., & Staikouras, C. (2010). Fuzzy cognitive maps in bankingbusiness process performance measurement. Studies in Fuzziness and SoftComputing, 247, 161–200 [Fuzzy Cognitive Maps].

Xirogiannis, G., Stefanou, J., & Glykas, M. (2004). A fuzzy cognitive map approach tosupport urban design. Journal of Expert Systems with Applications, 26(2),257–268.

Zadeh, L. A. (1965). Fuzzy Sets. Journal of Information and Control, 8, 338–353.Zhang, W. R., Wang, W., & King, R. S. (1994). A-pool: An agent-oriented open system

for distributed decision process modeling. Journal of Organisational Computing,4(2), 127–154.

Documents

Expert Systems with Applications - پارس پروژه · Strategy maps and the balanced scorecard ... amongst measures of different perspectives and objectives, ... show how value