Syst Pract Action Res (2011) 24:325–354
DOI 10.1007/s11213-011-9190-z
ORIGINAL PAPER
The Application of Fuzzy Cognitive Map
in Soft System Methodology
Payam Hanafizadeh • Rojin Aliehyaei
Published online: 26 January 2011
Ó Springer Science+Business Media, LLC 2011
Abstract Facing the issues of structural complexity, on which stakeholders have different views, has increasingly led to the use of Soft Systems Methodology (SSM) in
solving managerial problems. Moreover, the weaknesses of this methodology in considering all point of views and ensuring the effectiveness of the proposed changes have
provided the motivation for applying Fuzzy Cognitive Map (FCM) in SSM. Using FCM as
a modeling tool makes it possible to combine the views of different experts and form group
FCM (GFCM). GFCM has the potential to be applied as a useful decision support tool in
the stage of offering recommendations and changes. The methodology proposed in this
article is applied to ticketing system of Raja passenger train company. This system,
influenced by various policies and views, is analyzed with the recommended methodology
and then the solutions for developing the system are suggested in a prioritized manner.
Keywords Soft system methodology Fuzzy cognitive map Root definition
Action research
Introduction
One of the most important challenges of organizations, nowadays, is making decisions to
effectively solve soft problems, and managers encounter many conflicts in dealing with
such issues (Montazemi and Conrath 1986). In such a situation, the use of SSM as a
framework for solving ill-structured problems is increasingly growing among analysts
(Brocklesby 1995; Ingram 2000; Rose 2002; Shapiro and Shapiro 2003; Mirijamdotter and
Bergvall-Kårebirn 2006).
P. Hanafizadeh (&) R. Aliehyaei (&)
Department of Industrial Management, Allameh Tabataba’i University,
P.O. Box 14155-6476, Tehran, Iran
e-mail: hanafizadeh@gmail.com
URL: www.hanafizadeh.com
R. Aliehyaei
e-mail: rojin.aei@gmail.com
123
326
Syst Pract Action Res (2011) 24:325–354
Peter Checkland’ SSM, suitable for dealing with highly complex situations, is the lesson
learned from many action research projects (Wilson 1993). Checkland believes that the
traditional engineering approaches are appropriate for structured issues where the problem
situation is well described (hard problems). However, with large and ill-structured problems about whose definition divergent views exist (soft problems), the role of traditional
engineering approach is weakened and that of soft systems approach is emphasized
(Checkland 1978).
This methodology is a 7-stage process using the conceptual model to show relevant
activities in human activity system. SSM begins when crisis happens in the current circumstances of a particular company and forces it to search for a solution. It is then the time
for root definitions and models to be formed based on a distinct ‘worldview’, i.e., that view
of the world which enables an analyst to attribute meaning to what is observed. These
models are compared with what is understood in reality which results in clarification of
necessary actions for reaching the desirable situation (Wilson 1993; Mirijamdotter and
Bergvall-Kårebirn 2006). While the importance of this methodology has been realized in
recent decades, the limitations of its use have also been taken into consideration by many
researchers (Mingers 1984; Flood and Jackson 1991; Lane and Olivia 1998; Jackson 2003;
Rodr’iguez-Ulloa and Paucar-Caceres 2005; Yinghong 2007). In its fourth stage, this
methodology has no precise modeling tool as well as a definite technique to compare
recommended solutions in the real world. Furthermore, the root definitions of relevant
systems and the models obtained from them are meaningful only under special worldview
and analysts propose the final recommendations based on the selected view. Besides, the
effectiveness of system thinking in this methodology increasingly depends upon the
knowledge and experience of the participants (Checkland and Scholes 1990; Lane and
Oliva 1998; Avison and Fitzgerald 2003).
Another limitation of this methodology is that it does not revise the consistency and
contrast among various solutions in practice. For example, imposing some changes
simultaneously may cause conflicting results (Lane and Oliva 1998; Rodr’iguez-Ulloa and
Paucar-Caceres 2005; Yinghong 2007). Thus, utilizing a new model as an aiding tool in
SSM and creating combined methodologies including some methods of similar or different
thinking paradigms have became one of the relatively new debates in system thinking
(Mingers 1984; Munro and Mingers 2002). For instance, Soft System Dynamics Methodology (SSDM) is a result of an action research project conducted by Rodr’iguez-Ulloa
with an integrated framework of 10 steps in 1999. In the proposed methodology, SSM and
systems dynamics have been combined to create a synergetic tool for solving soft problems
(Rodr’iguez-Ulloa and Paucar-Caceres 2005).
The FCM, as a model showing the mind map of decision makers and representing
causal relations among various factors of the issue, attracts the attention of some
researchers to have it combined with the SSM (Hjortsø et al. 2005; Siau and Tan 2005). In
this study, FCM is combined with SSM so that in addition to defining the issue more
precisely, the analysis of different concepts and their effects on the goals of the system can
be carried out.
Considering the definition of methodology, the authors assert that the phrase ‘‘combined
methodology’’ has been employed in the sense that soft systems methodology is used from
the viewpoint of systems development, and FCM is used from the modeling perspective.
Generally, FCM is considered as a modeling tool embedded in SSM in this study.
Regarding the constraints of SSM, especially lack of a precise modeling tool and
regarding the capabilities of FCM in modeling stockholders’ perceptions, applying FCM in
SSM is proposed. FCM is used as a tool to represent various viewpoints of system
123
Syst Pract Action Res (2011) 24:325–354
327
beneficiaries through a visualized model enabling them to communicate with each other on
system requirements. In addition, the efficiency of models in SSM is highly affected by the
perceptions, knowledge, and expertise of their developers (Fiol and Huff 1992; Siau and
Tan 2005). Using combined techniques to aggregate FCMs in this methodology facilitates
the application of group knowledge and reduction of the probability of individual’s errors
(Kosko 1986; Khan and Quaddus 2004).
The graphic structure of FCM allows systematically indicating causal relations, particularly backward or forward chaining by which the analyst can determine the strength of
a concept’s influence on the goals of system (Kosko 1986). Furthermore, causal relations
among concepts clarify the contrast and consistency of various solutions. Thus, it can be
concluded that using this model in the final stage of SSM can facilitate choosing comprehensive solution to improve the current condition of problems. Table 1 summarizes the
Table 1 How FCM can address SSM’s limitation
Limitations of soft systems methodology
How FCM can address these limitations
SSM does not possess an accurate tool for
changing root definitions into model.
Fuzzy cognitive map helps to produce a chain of
concepts which define the problem in an integrated
form and represents the perceptions of
stakeholders. Also, it is used as an empowering
tool to increase the effectiveness of their decision
(Siau and Tan 2005; Hjortsø et al. 2005; Lin and
Wu 2008).
Fuzzy cognitive map can attract interviewees’
attentions and activate their minds in order to
structure the problem (Fiol and Huff 1992; Siau
and Tan 2005)
When the manger faces a massive amount of
information, cognitive map can magnify the
preferences and key factors (Siau and Tan 2005)
Fuzzy cognitive map specifies the information
defects, weak reasoning, and helps to find out the
areas needing information gathering (Siau and Tan,
2005)
Meaningful definitions and models proposed
under a particular view (Lane and Oliva 1998;
Rodr’iguez-Ulloa and Paucar-Caceres 2005)
Fuzzy cognitive map can show the expert’s
knowledge in group decision making (it has the
potential to be used as a group tool) (Khan and
Quaddus 2004)
The effectiveness of systems thinking in SSM
depends on individual’s knowledge, wisdom,
expertise, and attitude (Yinghong 2007)
Fuzzy cognitive map makes it possible to combine
the experts’ views.So, it can give more accurate
results and reduce the ambiguity (Kosko 1987,
1992; Banini and Bearman 1998; Kardaras and
Karakostas 1999; Khan and Quaddus 2004;
Hossain and Brooks 2008)
Not revising the coordination and conflicts
between the proposed changes (Checkland
and Scholes 1990; Lane and Oliva 1998)
Fuzzy cognitive map provides the feasibility for
static and dynamic analysis of different scenarios
(Khan et al. 2001; Irani et al. 2002)
Lack of precise and normative tools for ensuring
the effectiveness of changes and conformity
among them (Avison and Fitzgerald 2003)
Not prioritizing and offering optimum solution
for improving system (Flood and Jackson, 1991;
Lane and Oliva 1998)
123
328
Syst Pract Action Res (2011) 24:325–354
limitations of SSM and advantages of using FCM for improving the recommended
methodology.
This article is organized in seven sections; after the introduction, the second section
briefly addresses SSM methodology and the stages where FCM is applied in it. The third
section reviews literature on FCM and the analytical and combinational techniques of this
model. In the forth section, the improved stages in SSM are presented, followed by a case
study on the application of the proposed framework in section five. Finally, discussion and
conclusions about the results of applying methodology in the real world are presented and
suggestions for further studies are summarized.
Revision of SSM for Using FCM
Figure 1 indicates a general view of the methodology, and the stages in which FCM has
been employed are shown in grey. As it can be seen in Fig. 1, SSM involves seven stages.
The first and second stages are stages of finding out about the situation, the output of which
is a rich picture depicting real world concerns. The rich picture includes symbols and
words that show an image of the concerns and expectations present in the system and
create a basis for forming root definitions (RD) based on the worldviews in the system.
‘‘W’’ is in fact the short form for worldview (i.e., Weltanschauung). Worldviews (Ws)
indicate the different perceptions of different individuals from the same event.
The aim of the third stage in the methodology is to derive root definitions (RDs) from
the rich picture. RDs incorporate the point of views which make the activities and performance of the system meaningful. Producing several root definitions help to avoid any
hoped-for-utopian analysis (Wilson 1993). At this stage, CATWOE analysis can be used
for accurate formulation of RDs. This method is utilized as a checklist for ensuring the
Fig. 1 A general view of the methodology
123
Syst Pract Action Res (2011) 24:325–354
329
completeness of the root definitions. Two important components of this method are
worldview and transformation process, both of which must exist in root definitions.
Transformation process shows the basic core activities of the system including interconnected set of actions necessary to transform some input(s) into some output(s).
At the next stage, i.e., the fourth stage of SSM, the selected root definition(s) are
changed into a model. Then, at the fifth stage, the model obtained in the systemic world is
compared to the real world and finally, at stages 6 and 7, activities are suggested for
making changes in the real world and improving the present system (Checkland and
Scholes 1990).
In Fig. 1, the stages of applying FCM in SSM are shown in grey. First, FCM is used as a
modeling tool, and combinational FCM techniques are utilized for collecting the views of
experts and covering various viewpoints in stage 4 of SSM. Then, static analysis of FCM is
employed for improving the system and measuring the effect of each change on the goals
of system.
Fuzzy Cognitive Map
FCMs are the extension of cognitive maps (Axelrod 1976) used for representing causal
reasoning. Cognitive maps are collections of nodes connected by edges or links. The nodes
represent variable concepts from a domain and the edges represent causal relationships. A
positive edge from Ci (node i) to Cj (node j) means Cj increases as Ci increases and Cj
decreases as Ci decreases. A negative edge from Ci to Cj means Cj increases as Ci
decreases and Cj decreases as Ci increases (Kosko 1986).
A FCM is a cognitive map, except that a numerical value is associated with causal links
for showing the degree of relationship between two concepts. The directed edges take
values form the interval [-1, 1]. Fuzzy linguistic terms such as {very weak, weak,
medium, strong, very strong} can be used instead of numerical values (Kosko 1986). An
example of FCM is shown in Fig. 2.
A FCM
can be described by a connection matrix such as F. The matrix F is defined by
F ¼ e ij ; where eij is the weighting value of the directed edge from Ci to Cj. The matrices
associated with a FCM are always square matrices with diagonal entries as zero.
Fig. 2 An example of FCM (Khan and Quaddus 2004)
123
330
Syst Pract Action Res (2011) 24:325–354
The element in the ith row and jth column of matrix F, eij , would represent the
weighting value of the link directed out of node Ci into node Cj.
FCMs have been extended through time. Introducing indices such as centrality and
hierarchal degree (kosko 1986; Özesmi and Özesmi 2004) as well as static and dynamic
analysis to forecast the system behaviour (Vasantha and Ram Kishore 1999; Meghabghab
2002; Lee and Chung 2006), incorporating time factor in FCM (Park and Kim 1995),
applying combined technique of FCM (Kosko 1988; Taber 1991; Hossain and Brooks
2008) and using FCM in group decisions (Khan and Quaddus 2004) are but some of the
advances made in the field of FCM.
FCMs are used in various areas of application such as information retrieval,
(Montazemi and Conrath 1986), medical research (Vasantha and Ram Kishore 1999),
information systems development (Kardaras and Karakostas 1999), software engineering
(Hossain and Brooks 2008), web design (Lee and Chung 2006), and forecasting staff
behaviour on web (Meghabghab 2002), as well.
Static Analysis of FCM
A FCM can be applied for a static analysis of the domain to (a) discover the relative
importance of concepts in model; and (b) obtain the indirect and total causal effect between
two concepts (Khan and Quaddus 2004). The application of each of these analyses is
discussed in the next section.
Centrality Degree
The centrality of concept Ci, C(Ci), is an index for determining the relative importance of
this concept. As it is observed in Eq. 1, C(Ci), is obtained from sum of idðCi Þ and odðCi Þ
CðCi Þ ¼ odðCi Þ þ idðCi Þ
ð1Þ
id(Ci) is sum of absolute weighting values of causal links constituting all paths connecting
node Cj to Ci, where, i 6¼ j, (i.e., the column sum of the absolute values of node Ci in the
connection matrix), and od(Ci) is sum of absolute weighting values of causal links constituting all paths connecting node Ci to all nodes Cj, where, i 6¼ j, (i.e., the row sum of the
absolute weighting values of node Ci in the connection matrix). High centrality degree of a
concept not only shows the number of repetitions of the given concept, but also its
importance in the entire model (Kosko 1986).
The Indirect and Total Causal Effect Between Two Concepts
In order to find out the total causal effect from concept Ci to conceptCj, the indirect effects
over all the paths must be obtained.
A causal path from concept Ci to concept Ci Ci ! CKr2 ! ! CKrz ! Cj can be
denoted with ordered indices as ðKr1 ; Kr2 ; . . .; Krz So the indirect effect of concept Ci on
concept Cj over path r, Ir Ci ; Cj , is obtained by Eq. 2.
Ir ðCi ; Cj Þ ¼ minf e Cp ; Cpþ1 : ðp; p þ 1Þ 2 ðl; kr1 ; kr2 ; . . .; krz ; jÞ
ð2Þ
According to Eq. 2, the indirect effect of concept Ci on concept Cj through path r is
defined by the minimum operator. The e(Cp, Cp?1) is the weighting value of causal
relations between node p and node p ? 1.
123
Syst Pract Action Res (2011) 24:325–354
331
where, p and p ? 1 are contiguous left to right path indices.
Then, according to Eq. 3, T(Ci, Cj) is defined using the maximum operator where m is
the number of all indirect effects between Ci and Cj.
T Ci ; Cj ¼ max Ir Ci ; Cj : 1 r m
ð3Þ
If the number of relations with negative edge is zero or even, then the indirect effect of
the path is positive. If the number of negative relations in the path is odd, then the indirect
effect of the path is negative (Kosko 1986).
Combining Fuzzy Cognitive Maps
Matrix representation of FCMs makes it possible to combine different FCMs obtained from
different experts. For combining FCM, first the augmented FCM matrices are summed
according to Eq. 4. Fa is the FCM matrix of expert a and n is the number of experts. Wa is
equal to the credibility weight of expert a, and in FCM literature, it is common to use
Wa = 1 for all experts (Taber and Siegel 1987).
Different versions of a FCM specified to a specific domain will consist of unequal
number of concepts, and as a result, their connection matrices will have different sizes.
Therefore, the matrixes are augmented, so if there are total n nodes, each augmented matrix
Fa has n rows and n columns.
Fs ¼
n
X
Wa Fa
ð4Þ
a¼1
The next step is to determine the mean of the weighting value of causal links obtained
from all experts. For this purpose, according to Eq. 5, all arrays of Fs matrix are divided by
the total credibility weight (Tsadiras et al. 2001). So Ft is the combined FCM matrix
obtained from all experts.
Pn
Fs
a¼1 Wa Fa
P
Ft ¼ n
¼ P
ð5Þ
n
W
a
a¼1
a¼1 Wa
Group Fuzzy Cognitive Map (GFCM)
A group fuzzy cognitive map (GFCM) can be applied as an aiding tool in group decision
making. In group decision, there are individual or sub-groups with different point of views
or concerns about the same issue, so FCM obtained from each subgroup may have different
structure.
In forming GFCM, FCMs obtained from sub-groups and individuals are considered as
the main components of GFCM. First, the FCMs are merged in order to obtain the first
preliminary structure of GFCM.
The group will review all these concepts to identify similar concepts, because some
concepts with similar meaning may be given different names by different experts. Then, in
order to avoid redundant information, the group should make a decision to keep only one
of these concepts and remove others.
There are some rules for making a decision about removing or retaining a concept.
According to these rules, those nodes with fewer outgoing links are more suitable to remove,
as eliminating them will have less impact on other concepts. If the two candidate nodes for
eliminating have equal number of outgoing links, then the node with less causal influence on
123
332
Syst Pract Action Res (2011) 24:325–354
other concepts in the GFCM is considered as the first for removal. The causal influence of a
node is sum of absolute weighting values associated with its entire outgoing links.
By removing redundant concepts, the causal links associated with them will disappear.
Thus, it is necessary to analyze it in order to decide whether new causal links should be
added to the retained equivalent node in GFCM. For example, suppose that graph a and b
in Fig. 3 are merged to form the GFCM in Fig. 4.
Assume node C3 in graph a and node C03 in graph b are similar in meaning. As C3 has
fewer outgoing links, it is removed from GFCM and C03 is remained. Deleting redundant
node ðC3 Þ results in the disappearance of associated causal links. As mentioned previously,
its effect on outgoing links of a deleted node needs to be analyzed. By eliminating C3 , the
causal link between node C3 and node C5 is removed. The influence of the deleted node C3
on node C5 is not reflected by any pathway from its equivalent node. So the removed link
between node (C3) and node C5 is maintained by replacing its old deleted source node (i.e.
ðC3 Þ; by equivalent node. The Weighting value of causal relations in GFCM matrix is the
mean weight of causal links according to Eq. 4 and 5 (Khan and Quaddus 2004). The
preliminary and adjusted GFCM, formed by merging the FCMs in Fig. 3, are shown in
Fig. 4.
Fig. 3 Two different versions of FCM specific to the same issue
Fig. 4 The preliminary and adjusted GFCM formed by merging the FCMs in Fig. 3
123
Syst Pract Action Res (2011) 24:325–354
333
Improved Stages in Soft System Methodlogy
In the next subsections, the way of applying FCM in SSM is explained and stages in the
methodology using FCM or influenced by it are reviewed. As indicated in Fig. 1, the stages
that FCM has been applied in SSM are shown in grey. FCM is first utilised in modelling
stage, i.e., stage 4, and in the next stages of SSM, analytical techniques of FCM are used
for interpreting and decision-making on the way of developing the system.
Stage 4: Developing FCM Model
In SSM, the analyst can use conceptual modeling or other modelling methods for
describing the transformation process of root definitions. As it was mentioned, FCM, due
to its advantages and analytical strength in presenting the relationships among relevant
concepts, has been utilized in this study. The modelling steps of FCM and its application in
SSM are described in next section.
Constructing Preliminary Fuzzy Cognitive Map for Each Root Definition
At this stage, FCM is used as a tool among analysts and system users in order to help
clarifying system requirements from the viewpoints of various stakeholders and
beneficiaries.
To have a FCM equal to each root definition, first the main concepts relevant to each
root definition are recognized by related experts (Step 1 in Fig. 5), and then, a meeting is
held with them for final confirmation of these concepts and determining the causal relations
among them (Step 2 in Fig. 5).
After experts reach a consensus about the structure of FCM, a questionnaire equivalent
to each FCM is developed to assign weighting values to causal relations in FCM. These
values indicate the nature and strength of the relationships. For determining these values,
linguistic fuzzy weights such as ‘little’ or ‘strong’ can be applied instead of numerical
values, because they makes it easier for experts to express their opinions (Kosko 1994).
The linguistic terms used in the questionnaire are equated to the numerical value in interval
[0, 1].
In this way, the FCMs are extracted for all experts in each group (Step 3 in Fig. 5)
(Kosko 1986; Banini and Bearman 1998; Kardaras and Karakostas 1999; Khan et al. 2001;
Tsadiras et al. 2001).
Extraction of Combined FCM Relevant to Each Root Definition
After formation of various FCMs in each group, the combinational techniques of FCM are
used for summing up the views of experts in each group.
For developing the equivalent combined FCM of each root definition, first Fs is calculated through Eq. 4.
Fs ¼
n
X
Wa Fa
ð4Þ
a¼1
where, Fs is sum of FCM matrices obtained from each group of experts.
Fa is the augmented FCM matrix of expert a and n is the number of experts. Wa is also
equal to the credibility weight of expert a. The next step is to determine the mean of the
123
334
Syst Pract Action Res (2011) 24:325–354
Fig. 5 The steps of forming model (stage 4 of SSM)
weights of causal links obtained from all experts. To this aim, according to Eq. 2, all arrays
of Fs matrix are divided by the total credibility weight (Kosko 1992; Tsadiras et al. 2001).
Ft is the combined FCM matrix obtained for each group of experts (Step 4 in Fig. 5).
Pn
Fs
a¼1 Wa Fa
Ft ¼ Pn
¼ P
ð5Þ
n
a¼1 Wa
a¼1 Wa
In this way, by combining FCMs related to various experts and assigning different
weights to each expert, it becomes possible to more accurately and completely change the
root definition and cover the perceptions and expectations of the whole population of the
experts.
Forming Group FCM or GFCM
As mentioned above, one of the limitations of SSM is forming model based on a particular
worldview (W), and analysts cannot consider all views as a whole. At this stage, analysts
form a comprehensive model using group FCM techniques and use it as their main reference for analyzing different views in an integrated pattern.
In order to form a group FCM, the combined FCM of each sub-group with different
worldview (W) is considered as the preliminary data. As there may be different
123
Syst Pract Action Res (2011) 24:325–354
335
interpretations for one concept, the similar concepts in GFCM, named differently by each
sub-group, are identified in order to eliminate the possibility of providing redundant
information. The decision about retaining or removing a concepts and its links is analyzed
according to the rules described in the ‘‘Group Fuzzy Cognitive Map (GFCM)’’ section.
Then, according to Eqs. 4 and 5, the average of corresponding elements of the FCM matrix
is computed for finding the weighting values of causal relations in the GFCM and
equivalent matrix (Khan and Quaddus 2004). This step of forming the model is indicated as
step 5 in Fig. 5.
Since the views of all experts are collected in the model, the obtained model is considered as a comprehensive tool for analyzing and decision-making in the next stages and
is not limited to a specific worldview (W).
Stage 5: Comparing the System World with the Real World
At this stage of methodology, the real world activities are compared with those of system
world using the model obtained from modeling stage (Wilson 1993). Since the number of
activities in the real world is large, the most important activities must be considered. To
this aim, concepts equivalent to the objectives of the system in GFCM, which usually enjoy
high degrees of importance, are first identified. In order to determine these objectives, the
centrality degree of concepts in GFCM model can be employed, because high centrality of
a node indicates its importance in the whole model (Step 1 in Fig. 6.). Thus, the centrality
degrees of all concepts in GFCM are obtained according to Eq. 1.
Fig. 6 The steps of comparing the system world with the real world (stage 5 of SSM)
123
336
Syst Pract Action Res (2011) 24:325–354
idðCi Þ is the column sum of the absolute values of node Ci in the GFCM matrix, and
odðCi Þ is the row sum of the absolute weighting values of node Ci in the GFCM matrix.
Then, according to the causal structure of GFCM, the information about those concepts
directly affecting system goals (i.e., increasing or decreasing goals of system) is gathered
in the comparison table. The table includes the current mechanism of activities equivalent
to concepts directly affecting system goals, the criteria for measuring its function, and the
recommendation for reaching the optimum situation (Wilson 1993; Platt and Warwick
1995). In Fig. 6, this information is shown as steps 2 and 3.
At the end of this stage, a table is obtained which includes suggestions for improving the
present situation on the basis of the investigation of causal relations among those concepts
directly affecting system goals.
Stage 6 and Stage 7: Defining Desirable and Possible Changes and Taking Action
As it was mentioned earlier, the output of the previous stage is a set of comparison tables
which offer a group of recommendations for improving the situation. The recommended
changes should be analyzed carefully, because a particular change may be reasonable to an
analyst, but for the stakeholder who has had a particular experience about that, other
concerns and policies should be simultaneously taken into account (Wilson 1993).
At this stage, the GFCM obtained from the expert groups can also help the analyst.
Using this graph, the paths for reaching certain goal, feasibility of solutions, and the
indirect and total effect of each concept on system goals can be obtained.
In this way, the paths among the concepts obtained from the previous stage and concepts equivalent to the objectives of the system are found (Step 1 in Fig. 7).
The indirect and total causal effects between two concepts in GFCM are calculated
according to Eq. 4 and 5. Therefore, by calculating the total effect of other concepts on the
goals of system, the analyst has a measure to compare the effects of each concept on
system goals, and finally, there will be prioritization for changes. This information is
summarized in Fig. 7 as step 2.
Fig. 7 The steps of defining and selecting the desired actions
123
Syst Pract Action Res (2011) 24:325–354
337
Case Study
In this section, an application of the improved methodology in ticket sales department of
RAJA passenger train company is described.
RAJA commenced its activities in Iran 13 years ago as a state organization aiming at
improving the quality and quantity of facilities for passenger train service. RAJA’s mottos
are optimal use of facilities, constant improvement, increasing efficiency, and offering
desirable services in order to achieve organizational excellence. In this regard, providing
better ticket sales services is one of the key factors in achieving its aims and increasing
customer satisfaction. Since ticket sales system has various stakeholders with different
expectations and needs, it is important to review all aspects of improving ticket sales
system including technical and non-technical issues.
Stage 1 and Stage 2: Finding Out About the Situation
In the first step, the problem situation is explored and explained whose output is a rich
picture (Checkland 1981). For initial identification of the system and exploring the
worldviews, the analysts have used open interviews and documents existing in the company. The interviews were conducted with 5 managers and 15 experts in RAJA with more
than 5 years of experience in ticket sales activities, as well as 2 managers and 5 personnel
of ticket selling agencies. The rich picture of the problem is shown in Fig. 8.
The picture contains a symbol of thinking stick figure indicating someone who is
expressing his/her particular concerns. The picture of computer and network stands for the
concerns of the system managers regarding software and network support of the sales
system. The ‘‘money bag’’ symbolizes the concerns about the amount of allocated budget
to maintenance and development of the sales system. The ‘‘buildings’’ shown at top of the
figure are symbols of the importance of policy-maker organizations and commercial
partners in the rail transportation such as the Ministry of Road and Transportation and train
operating companies. Also, the logo of ‘‘e-commerce’’ at the right bottom corner of the
figure indicates the increasing attention of the company to offering electronic services to
customers. Other pictures and explanations show the factors affecting the development of
the sales system which will be elaborated later.
Stage3. Formulating Root Definitions
After determining different views about the ticketing system, the root definitions are
formed. By conducting many interviews and gathering the required information in RAJA,
three distinctive worldviews are identified which include: privatization, implementing
customer-oriented approach, and e-commerce services with IT infrastructure.
The proponents of privatization believe that diversity of train services and facilities, and
meeting customer demands for trip will be possible through attracting private sector
investors. Some others believe that customer-oriented approach and answerability to
customers’ questions and complaints via supervising the customer services and information
flow are the most important factors for successful ticket selling. Still others believe that
applying information technology infrastructures provide rapid access to ticket services and
enhance customer satisfaction significantly.
On the basis of the three views obtained from the ticketing system, three root definitions
were extracted as followings:
123
338
Syst Pract Action Res (2011) 24:325–354
Fig. 8 The rich picture of ticket selling in RAJA passengers’ train company
RD1: A system owned by RAJA passenger train company and train operating partners
which attempts to provide the passengers with more access to train ticket in the given path
with desirable services. This mission is accomplished through increasing trip offers in
various paths by attracting the private sector investment in the rail transportation. The
limitations of this system relate to the restricting policies on the part of policy-making
organizations in the area of rail transportation and reluctance of private companies for
investment in this area due to its high expenses.
RD2: A system owned by RAJA passenger train company which attempts to provide the
passengers with easy and rapid access to train ticket through the Internet. The limitations of
this system relate to the degree of individuals’ capacity to access the Internet and the level
of society literacy for using electronic services.
RD3: A system owned by RAJA passenger train company which offers the possibility of
reservation of ticket and other rail services to the passengers. Accurate and update
information regarding the services and facilities of trains and answering the questions and
complaints of the customers are among the primary and necessary factors for accomplishing this mission of the system. The limitations of this system relate to the restricting
regulations on the part of policy-making organizations in the area of rail transportation.
123
Syst Pract Action Res (2011) 24:325–354
339
In order to define the necessary characteristics of the system and precisely formulate the
root definition the CATWOE analysis was applied (Smyth and Checkland 1976).
According to the CATWOE analysis for RD1, the system owners are Raja company and
train operating partners. The system customers are those requesting travel with train and
the system actors are the ticket selling agencies and partners. The input of this system is the
request of rail services on the part of the passengers and its output is the ticket for the given
path and train. The worldview (W) dominating the system is increasing the rail transportation capacity by exiting from the exclusive market into the competitive market.
In RD2, the system owner is Raja company. The system customers are those requesting
ticket reservations through the Internet and the main actors of the system are RAJA
company and policy-making organizations. In this system, customers’ demand for preparing train ticket is replied. Limitations of society in access to the Internet and people’s
literacy level in using electronic services must be considered in offering such services.
Finally, according to the CATWOE analysis for RD3, the system owner is Raja company. The system customers are those requesting train tickets. The system actors are the
ticket selling agencies and RAJA company. In this system, the possibility of buying train
ticket is provided for the customers. Offering timely and correct information, and
answering the questions and complaints of the customers, as well as appropriate interaction
with the customers are primary requirement of this system. The limitations of this system
relate to the policies of policy-making organizations in the area of rail transportation. The
results of CATWOE analysis for each root definition are listed in Table 2.
Table 2 The root definitions of ticket selling system of Raja analyzed by CATWOE
Analyzed
elements of
CATWOE
RD1
RD2
RD3
Customers (C)
people who demand rail
travel
Customers who purchase
ticket
Customers who demand rail
travel
Actors (A)
RAJA Co., Train operating RAJA Co., agencies
partners, Ministry of Road
and Transportation,
Islamic Republic of Iran
Railway (RAI)
Transform (T)
Providing access to train
ticket in the given path
and desirable services
Providing access to train
Providing access to train
ticket through the Internet ticket and answering their
questions and complaints
Weltanschauung
or Worldview
(W)
Changing the exclusive
market of railway
transportation into a
competitive market by
privatization
Utilization of information
technology services to
promote e- commerce
services
Developing Customers
Relationship Management
(CRM)policies
System owner
(O)
RAJA Co and Train
operating partners
RAJA Co.
RAJA Co.
RAJA Co., agencies
Considering people access Limitation resulted from
Limitation resulted from
System
external organization
to computer and internet
external organization
environment
(Ministry of Road and
society (the percentage of
(Ministry of Road and
and the
people who have access to Transportation, Islamic
Transportation, Islamic
resulting
Republic of Railway)
the Internet)
Republic of Iran Railway
restrictions (E)
and train operating
partners)
123
340
Syst Pract Action Res (2011) 24:325–354
At this stage, the model of each root definition is formed in order to have a better
perception of the issue.
Forming the Initial Structure of FCM
After exploring different worldviews on the ticket selling system, the relevant experts of
each worldview were attributed to distinct groups (i.e., group A, group B, and group C).
Expert group A consisted of those experts who believe in privatization. These experts
have master’s degree or higher education levels and minimum of 10 years experience in
railway transportation and privatization. Group B consisted of specialists with 8 years
background in information technology and e-commerce. Group C involved those who
believed in implementing customer-oriented approach in Raja; they had graduate degree
and minimum of 10 years experience in customer-oriented activities. All experts in 3
groups cooperated in at least 2 projects related to ticket selling. As the numbers of experts
with such characteristic were limited, no sampling was conducted and all experts participated in the study. The number of experts from whom the questions were asked was 9 in
group A, 6 in group B, and 8 in group C.
Using the information gathered from the previous stage, the relevant concepts of each
worldview were derived separately, and delivered to related experts (Step 1 in Fig. 5).
After finalizing the list of concepts and determining their relationships in different meetings with groups A, B and C, the preliminary FCM of each worldview was formed (Step 2
in Fig. 5).
The perceptions and expectations of the experts of each group from the system were
made clear by holding meetings with them.
The next stage includes assigning weighting values to causal relationships by the
experts. Therefore, an equivalent questionnaire for each FCM was prepared. The questions
are related to weighting of each causal relation (Step 3 in Fig. 5). A sample of questions
related to group C is presented in the Appendix 1. Experts’ response to questions needed to
be selected from the alternatives of ‘none, very weak, weak, strong, and very strong (i.e.,
based on likert spectrum); the linguistic terms were attributed to numerical values in
interval [0, 1]. These values are {0, 0. 25, 0. 5, 0. 75, 1} (Taber 1991; Hossain and Brooks
2008).
The prepared questionnaires were administered to 5 university professors and experts to
estimate their face and content validity and remove possible defects. Then, data of questionnaire were analyzed using SPSS software and Cronbach’s alpha coefficient (a) was
computed as a measure of internal consistency in the questions. For the questionnaire of
group A, Cronbach’s alpha was 0.713; for group B Cronbach’s alpha was 0.749; and for
group C it was 0.718. Since all these values are more than 0.7, it can be argued that all
questionnaires have acceptable reliability. Then, the questionnaires of three groups were
distributed among the related experts and after completing forms, the connection matrices
of FCMs were separately obtained for all experts.
Deriving Combined FCM for Each Root Definition
At this stage, FCMs of each group of the experts are combined to obtain a combinational
FCM as a final model of each RD (Step 4 in Fig. 5).
In order to derive the combinational FCM, the combinational matrix of each worldview
was calculated using the Eq. 4 and 5, where Fr is the augmented matrix for expert
r obtained from the previous stage. As it is common to use Wr = 1 for all experts in the
123
Syst Pract Action Res (2011) 24:325–354
341
literature on FCM (Taber and Siegel 1987), and since there was no remarkable difference
among the participants in the study, weight 1 is assigned to all the experts.
For instance, the effect of the node ‘‘quality level of software support’’ on ‘‘quality level
of services offered by agencies’’, as one item of the combinational matrix FT, is calculated
using Eq. 1 and 2. The weights assigned by 8 experts of group C are 0.75, 0.8, 0.25, 1,
0.75, 0.5, 0.75, and 0.75 whose mean is the equivalent to the item of the combinational
matrix related to the relationship between these two concepts (Fig. 9).
These calculations were done for all causal relations, and the combinational matrices
were obtained for groups A, B, C. Combinational FCM graph based on RD1 (experts of
group A) is shown in the Appendix 2.
Forming the Group FCM (GFCM)
In order to form the group FCM, the combined FCM obtained from each group of experts
was considered as the preliminary data.
At this stage, 3 combined FCMs obtained from the previous step are merged to form the
GFCM (Step 5 in Fig. 5). The nodes in GFCM were analyzed and the similar concepts in
graphs which may be named differently by each group of experts were found and the
decision about removing and retaining concepts was made according to the rule described
in ‘‘Group Fuzzy Cognitive Map (GFCM)’’ section.
Since there is a large number of concepts and causal relations in the GFCM obtained, by
presenting only a part of the nodes and links of the FCM model related to each worldview
(W) and the final GFCM obtained from them in Figs. 10, 11, 12, and 13, the way of
creating GFCM becomes clear. According to the model obtained from the expert group on
privatization in Fig. 10, the concept of ‘‘congruity of the policies of the Ministry of Road
and Transportation with development of passenger transportation’’ affects ‘‘allocated
budget to RAJA company’’ and finally ‘‘annual budget of sales department for conducting
the present activities and projects’’. Also, ‘‘management’s emphasis on improving ticket
sale’’ is another concept which influences ‘‘annual budget of sales department for conducting the present activities and projects’’.
In Fig. 11, FCM is formed on the basis of RD2 in which the experts have investigated
the effects of concepts such as ‘‘the quality of software support’’ on ‘‘quality of services
offered by the ticket selling agencies’’. The experts of this group believe that ‘‘allocated
budget to sales system’’ and ‘‘management’s emphasis on improving the ticket sales
system’’ affect ‘‘quality level of software support’’ and finally, ‘‘quality of services offered
by the agencies’’.
In the model formed on the basis of implementing customer-oriented view in Fig. 12,
the effect of ‘‘quality level of software support’’ on ‘‘quality of services offered by the
agencies’’ as the organizations having direct contact with the customers, is investigated.
This concept, in turn, is influenced by other concepts such as ‘‘salespersons’ skill level’’.
Fig. 9 Calculation of the weight
of causal relationship between
two nodes of C20 and C29 in the
combinational matrix
123
342
Syst Pract Action Res (2011) 24:325–354
Fig. 10 Part of FCM based on RD1
Fig. 11 Part of FCM based on RD2
As it can be seen, the obtained models possess common concepts like ‘‘quality level of
software support’’, ‘‘quality level of network support’’, and ‘‘quality level of services
offered by the ticket selling agencies’’. In addition, some of these concepts, while having
different names, have similar meanings. For instance, ‘‘annual budget of sales department
for conducting the present activities and projects’’ in Fig. 11 is the same as ‘‘allocated
123
Syst Pract Action Res (2011) 24:325–354
343
Fig. 12 Part of FCM based on
RD3
Fig. 13 GFCM obtained from Fig. 10, 11 and 12
budget to sales system’’ in Fig. 10. According to the rules mentioned in the literature on
GFCM, since the number of outgoing links of the concept ‘‘allocated budget to sales
system’’ in Fig. 11 is more than the number of outgoing links of ‘‘annual budget of sales
department for conducting the present activities and projects’’ in Fig. 10, it was decided to
123
344
Syst Pract Action Res (2011) 24:325–354
remove the latter and retain the former in the final GFCM. Since this node does not have
any outgoing link, there is no need to add a new link in the final GFCM. At the end, the
weights of the causal relations are obtained from the mean of the equivalent causal relations in the present FCMs. In Fig. 13, the final GFCM obtained from Figs. 10, 11, and 12
are presented. For example, the relationship between two concepts of ‘‘quality of software
support services of sale system’’ and ‘‘quality level of services offered by the ticket selling
agencies’’ is obtained from the mean of experts’ views on the weighting value of the causal
relation between these two concepts in groups B and C, i.e., 0.781 and 0.916, which is
0.845.
These steps were taken for all nodes and relations in GFCM model.51 concepts were
extracted and listed in Appendix 2 and the GFCM obtained in this study is shown in
Fig. 14.
Stage 5: Comparing the System World with the Real World
At this stage, using the GFCM model obtained, the system world is compared with the real
world. Since the number of concepts is high in this model and it is not possible to
investigate all of them, the concepts with higher importance, known as the system
development objectives, are examined (Step1 in Fig. 6).
As it was mentioned in ‘‘Stage 5: Comparing the System World with the Real World’’
section, centrality degree is the index showing the importance of concepts in the model.
According to Eqs. 1 and 2, C13 and C14 have the highest centrality in FCM. The results of
computation of centrality degree for the concepts with the maximum values are summarized in Table 3.
Fig. 14 GFCM obtained by merging all FCMs
123
Syst Pract Action Res (2011) 24:325–354
345
Table 3 The results of computation of centrality degree for the concepts with the maximum values
idðCi Þ
Concepts
odðCi Þ
C ðCi Þ ¼ idðCi Þ þ odðCi Þ
C14
Customers’ satisfaction (internet users)
14.102
1.402
15.504
C13
Customers’ satisfaction (Non internet users)
7.800
1.528
9.328
As it can be seen, C13 and C14 have the highest centrality degree and are considered as
the main objectives of system development. Thus, using the GFCM model, the concepts
having direct influence upon these two concepts are identified and after defining activities
equivalent to each concept, the columns of comparative table are filled (Steps 2 and 3 in
Fig. 6).
An example of these comparisons for concept C29 is represented in Table 4 (Wilson
1993; Platt and Warwick 1995).
As it can be observed in GFCM model, node C29 entitled as ‘‘quality of services offered
by the agencies’’ is among the important factors which directly influence C13 and C14. The
criteria for measuring the performance of activity equivalent to this concept, i.e., services
offered by ticket selling agencies, is the customer satisfaction percentage from the services
offered by agencies.
Then, the relations leading to this concept, i.e., C29, in GFCM model is investigated so
that the concepts affecting this specific concept are identified.
The causal relations shown in red in Fig. 14 are as followings:
0:667
0:833
0:69
0:656
0:75
0:75
C27 ! C33 ! C17 ! C40 ! C29
C47 ! C46 ! C29
0:719
C45 ! C29
0:495
C39 ! C29
0:875
0:875
0:849
C7 ! C22 ! C20 ! C29
0:639
0:792
0:849
0:639
0:708
0:75
C2 ! C7 ! C20 ! C29
C2 ! C7 ! C21 ! C29
Table 4 Comparison table of concept C29
Quality of services offered by agencies (C29)
1 Activity
Offering ticket selling services to customers
2 Exist or not
Yes
3 Current mechanism
Selling ticket and helping customers to find and reserved ticket for their
destination considering price and other concerns. The agencies get their
required information by the ticket selling system of RAJA
The percentage of customer satisfaction from services is 60%
4 Criteria for measuring
performance
The percentage of customers’ satisfaction from offering services
5 Proposed change
Improving the network and software support services, promoting
salespersons’ skill. Appropriate information, controlling the distribution of
agencies in the city, investigating violations at agencies
123
346
Syst Pract Action Res (2011) 24:325–354
The chain of causal relations related to concept C29, reveals that the quality of services
offered by agencies (C29) is influenced by C20 (quality level of software support), C21
(quality level of network support), C45 (salespersons’ skill level), C39 (availability of
information and brochures for customers), C47 (number of agencies in the city), and C40
(investigating agencies’ violation). Also C20 and C21 are influenced by C7 (allocated
budget to sales system) and C2 (management ‘emphasis on improving ticket sales system).
Consequently, the mentioned concepts must be taken into consideration for offering better
services by the agencies.
Such tables are created for other concepts which directly affect the Internet and nonInternet customers’ satisfaction.
The methods of getting the favorable situation are identified through comparisons and
analyses. The list of these concepts is presented in the first column of Table 5.
Stages 6 and 7: Defining the Desirable and Possible Changes and Taking Actions
As it is mentioned in the 7-stage methodology of Checkland, the conceptual models
include a collection of activities named as ‘‘what’’ and the analyst observes a collection of
‘‘hows’’ in the real world to achieve each of these whats. Within the framework offered in
this study, the whats are the concepts equivalent to the objectives of the system and the
hows are the paths showing the ways of making desirable changes in the system objectives
(Step 1 in Fig. 7). Hence, after determining the changes which are effective in the present
situation for reaching the desirable situation, the degree of the effect of each change on
system objectives is determined and the decisions as to the conduction of the final activities
in the real world are made.
Using the model, the analyst can calculate the effect of each concept upon the system
development objective over all paths (Step 2 in Fig. 7). For example, in order to calculate
the indirect effect of C7 (allocated budget to sales system) on C13 (customers’ satisfactionnon Internet users), first the paths between the two concepts are specified and then, the
indirect effect from concept C7 to concept C13 is calculated in all paths according to Eq. 2.
The results of these calculations are as following:
0:875
0:875
0:542
0:708
0:75
0:807
0:875
0:667
C7 ! C22 ! C20 ! C24 ! C34 ! C13
I1 ðC7 ; C13 Þ ¼ feðC7 ; C22 Þ; eðC22 ; C20 Þ; eðC20 ; C24 Þ; eðC24 ; C34 Þ; eðC34 ; C13 Þg
I1 ðC7 ; C13 Þ ¼ minf0:875; 0:875; 0:542; 0:875; 0:667g ¼ 0:542
C7 ! C21 ! C29 ! C13
I2 ðC7 ; C13 Þ ¼ feðC7 ; C21 Þ; eðC21 ; C29 Þ; eðC29 ; C13 Þg
I2 ðC7 ; C13 Þ ¼ minf0:708; 0:75; 0:807g ¼ 0:708
Then, the total effect of C7 on C13 is calculated according to Eq. 3.
T ðC7 ; C13 Þ ¼ maxfI1 ðC7 ; C13 Þ; I2 ðC7 ; C13 Þg ¼ 0:708
All these computations are performed for the considered concepts and the results are
shown in Table 5.
The concept C4 has the most effect of the non-Internet customers in comparison to other
concepts, and as it can be seen in the columns 2 and 3 of Table 5, the effect of C4 on each
of the concepts C13 and C14 are 0.867 and 0.639. In the two final columns of the table, the
present situation of activities equivalent to these concepts and solutions for improving
them are offered.
123
Syst Pract Action Res (2011) 24:325–354
347
Table 5 The total effect of concepts on customers’ satisfaction
Concept
The
Current mechanism
The
effect effect
on C13 on C14
Recommendations
1. Congruity of the policies 0.867
of the Ministry of Road
and transportation with
development of passenger
transportation (C4)
0.639
At present, the Ministry of Defining project, contact
with the interested parties
Road and transportation
for continuous relation
greatly stresses the
with policy-making
development of rail
organizations
transportation, but RAJA
Company does not have a
definite mechanism for
interacting with policymaking organizations
2. Government subsidy
regarding rail passengers
(C19)
0.855
At present, the ticket price
subsidies are determined
by the Ministry of Road
and Transportation
3. Appropriate Response to 0.75
customer complaints (C49)
0.75
At present, the complaints of Defining CRM project for
investigating and
the customers are replied
controlling
via different channels, but
communicative channels
there is a need for
integration and
homogenizing
communicative channels
4. Quality level of network
(C21) support
0.75
0.667
At present, many problems Promotion of the services of
exist, the connection of the network support team
agencies with the selling
system is lost, and many
problems arise in offering
services to the customer
5. Management ‘s ideas
regarding privatization
(C6)
0.734
0.734
Conducting privatization
At present, 30% of the
project for explaining the
passenger rail
way of delegating
transportation sector is
delegated to private sector, passenger rail services to
private sector
but its acceleration
requires the integration of
policies adapted by the
managers
6. Salesperson’s skill level
(C45)
0.719
0.61
At present, many clients
complain about the
behavior of the
salespersons
Holding training courses and
periodical tests for sales
personnel
7. Budget allocated to sales 0.708
system(C7)
0.71
At present, a low budget is
allocated to development
of sales system
Allocation of sufficient
financial resources for
maintenance and
promotion of sales system
0.851
Media movements at the
time of approving budgets
by the related organization
(within the framework of
relationships with the
interested parties)
123
348
Syst Pract Action Res (2011) 24:325–354
Table 5 continued
Concept
The
The
Current mechanism
effect effect
on C13 on C14
8. Integration level of
databases (C24)
0.667
0.875
At present, the sales system Acceleration of launching
of the agencies is separate the new sales system
which enjoys from the
from internet selling
necessary integration
system which has brought
about many problems in
offering information to
customers and getting
report from the system
9.Appropriate design of
databases(C27)
0.667
0.75
There is no specific problem In the new project for
at present
developing sales system,
the way of designing data
bases must be considered
as an important factor in
system development
10. Agreement of President 0.639
Deputy Strategic Planning
and Control with the
estimated expenses of
RAJA (C18)
0.639
At present, there is no stable Definition of continuous
interaction with the related
mechanism for the
organization within the
interaction with this
framework of relation with
organization
the interested parties
11. level of programming
for sales system (C23)
0.583
0.869
The sales software
programming language is
very old (programs
running under DOS
operating system) which
has brought about many
problems for system
development
12. Quality level of software 0.542
support (C20)
0.652
At present, the software has Selecting appropriate
contractor for supporting
many errors and has
the sales system software
imposed many expenses
13.Appopriate interaction
with banking system (C36)
0.75
Appropriate and continuous
Lack of appropriate
interaction with contracted
interaction with banks has
banks and identifying
created many problems in
other opportunities for
offering electronic
making more contracts
payment services
and developing internet
electronic payment
services
0.844
At present, many users, due Changing the design of web
pages and offering
to lack of familiarity with
sufficient instructions for
internet, have problems in
completing internet
completing their internet
shopping process
shopping process
14. User friendly online
booking (C51)
_
15. Number of agencies in
the city(C47)
Dual policies regarding
0.719 0.75
increasing the number of
0.656 0.614 agencies in the city
123
Recommendations
Considering an appropriate
programming language for
the new software which is
in line with the other parts
of the system
Estimation of the number of
agencies in each area so
that the distribution of
profit among them
becomes appropriate
Syst Pract Action Res (2011) 24:325–354
349
As it can be seen in Table 5, suggestions were proposed for promoting the present sale
system in 15 areas, and the percentage and way of their effect on the main objectives of the
system were determined. This provides the analysts with an appropriate criterion for
prioritizing activities. For example, ‘‘congruity of the policies of the Ministry of Road and
Transportation with development of passenger transportation’’ has been defined as the most
important factor in increasing ‘‘non-Internet customer satisfaction’’ in which issues like
appropriate interaction with policy-making organizations or media movements for
improvements in decision-making and attention to the rail travel are considered. Among
other suggestions offered were defining customer relationship management (CRM) project
for analysis and control of customer communicative channels as a comprehensive project,
and holding training courses and periodical tests for measuring and promoting the ‘‘skills
of salespersons’’.
Integrating databases, selecting appropriate programming language, and the amount of
government subsidy regarding rail passengers affect Internet-customers’ satisfaction by
0.875, 0.869, and 0.855, respectively.
Another significant result was attention of decision making to increase the number of
sale agencies in the city that, according to Table 6, has dual effects on customer
satisfaction.
On the one hand, by increasing the number of agencies, it is easier for the customers to
access the agencies and this pleases them, and on the other, increasing the number of
agencies results in decreasing the profit of each agency. This can lower the quality of
services delivered by the agencies and finally have negative effect on customer
satisfaction.
This model is considered as an aiding tool for comparing different solutions to reach the
optimum situation in the real world. It can measure the impact of different concepts on
goals of the system.
Thus, it can be claimed that this model helps making effective managerial decisions in
the organizational strategic level and makes it possible to measure the effect of various
changes on system objectives before taking any action in the real world.
Discussion
In this paper, using SSM and FCM model, suggestions were offered for improving the
present situation of ticket selling system in RAJA company. Exploring the real world by
common techniques of SSM and construction of FCM models on the basis of various
worldviews (Ws), provides a great insight into the causes of the crises. For example, using
the rich picture obtained in Fig. 8, a comprehensive image from the problem situation
comes to the mind. Besides, since SSM provides the analyst with the opportunity to
construct separate definitions and models, it becomes possible to investigate various
viewpoints regarding the system. In this study, employing FCM model in the modeling
stage, the needs and expectations of the various stakeholders were identified and an
algorithm was created for modeling the mental perceptions of them in SSM.
In the case study of this research, the final GFCM was obtained by combining three
distinct worldviews on the development of sales system. In this way, system development
was not solely based on the views of a specific group of specific interested parties; rather,
all worldviews (Ws) were covered.
Another advantage of using FCM model in SSM is the possibility of static analysis for
estimation of the effect of changes on a system condition which provides a criterion for
123
350
Syst Pract Action Res (2011) 24:325–354
making changes on the basis of the amount of its effect on system objectives in stages 6
and 7. As it was explained, using these analyses, a list of activities and projects for
improving the system was proposed.
Of course, using FCM models in SSM has some limitations; finding experts and collecting information from them to construct the initial structure of FCM is a time consuming
step. Also computing the weights of the edges between nodes to create group fuzzy
cognitive maps for large systems is complex.
Conclusion
Checkland methodology gained great achievements in previous decades and has had many
applications for solving ill-structured and unstructured problems. However, some limitations of this methodology particularly in the modeling stage have restricted its application.
In this study, by applying FCM as a modeling tool in SSM some improvements were
achieved. Constructing these models, give a complete picture of different stakeholders’
perceptions about the issue and create a tool for combining the experts’ views so that the
analyst is able to consider different attitudes of beneficiaries in the modeling step. Also, at
the stage of comparison of the model with the real world, by analyzing causal relations in
FCM, the way to achieve the optimum status was formed. At the final stage, through
calculating the total effect of concepts on system goals, the normative scales for prioritizing the recommendations were presented.
By increasingly using this methodology in the real world and investigation of its results,
improvements can be achieved in the application of FCM in SSM. In addition, dynamic
analysis of FCM model can be utilized in future studies enabling us to predict the state of
the system through time.
Appendix 1
See Table 6.
Table 6 The sample questions of questionnaire for estimating the severity of relations among factors
Concept name
1
Availability of ticket printing
Machine Affects online
customer’s satisfaction
2
The ticket price affects nononline customers’ satisfaction
3
The ticket price affects online
customers’ satisfaction
4
The Possibility of refunding
E-ticket or canceling trip
affects online customers’
satisfaction
123
Effect
Effect degree
1
Very
strong
2
Strong
Average
Weak
None
Syst Pract Action Res (2011) 24:325–354
351
Appendix 2
.
Appendix 3
See Table 7.
Table 7 List of variable concepts
Variable
concepts
Description
c1
Congruity of the policies of the of RAI ‘s (Islamic Republic of Iran Railways)with
development of passenger transportation
c2
Management’s emphasis on improving ticket sales system
c3
Allocated budget to RAJA Co.
c4
Congruity of the policies of the Ministry of Road and Transportation with development of
passenger transportation
c5
Percentage of passenger trains assigned to private sector
c6
Management ‘s ideas regarding privatization
c7
Allocated budget to sales system
c8
Facilities given to private sectors
c9
Passengers rail Capacity to transport passenger
c10
Ticket availability (inventory)
c11
Ticket price
c12
RAJA’s income
123
352
Syst Pract Action Res (2011) 24:325–354
Table 7 continued
Variable
concepts
Description
c13
Customers’ satisfaction(Non online users)
c14
Customers’ satisfaction(Online users)
c15
Availability and variety of trains and services
c16
Managers’ satisfaction with sales system
c17
Comprehensiveness of reports on sale
c18
Agreement of President Deputy Strategic Planning and Control with the estimated
expenses of RAJA
c19
Government subsidy regarding rail passengers
c20
Quality level of software support
c21
Quality level of network support
c22
Quality of hardware maintenance
c23
Level of programming for sales system
c24
Integration level of databases
c25
Sales data Security
c26
Hardware Security
c27
Appropriate design of data bases
c28
Web sites Bandwidth
c29
Quality of services offered by agencies
c30
Processing speed of sales system
c31
Easy access to Raja’s website
c32
Appropriate link with banks
c33
System flexibility in scheduling
c34
Update information about ticket inventory
c35
Customer satisfaction(Internet user) about e-payment
c36
Appropriate interactions with different banking systems
c37
Availability of credit cards
c38
Feasibility of developing sales system
c39
Availability of information and brochures
c40
Investigating agencies violation (failure)
c41
Possibility to receive the print of reserved ticket from agencies in stations
c42
Availability of Ticket Printing Machine
c43
Possibility of refunding E-ticket or canceling trip
c44
Possibility of refunding ticket or canceling trip in agencies
c45
Salespersons’ skill level
c46
Profit of agencies
c47
Number of agencies in city
c48
Number of ticket available online
c49
Appropriate Response to customer complaints
c50
Offering presales services
c51
User-friendly online booking
123
Syst Pract Action Res (2011) 24:325–354
353
References
Avison DE, Fitzgerald G (2003) Information systems development: methodology techniques and tools.
McGraw-Hill Education, Oxford
Axelrod R (1976) Structure of decision. Princeton University Press, Princeton, NJ
Banini GA, Bearman RA (1998) Application of FCMs to factors affecting slurry rheology. Int J Min Process
52(4):233–244
Brocklesby J (1995) Using soft systems methodology to identify competence requirements in HRM. Int J
Manpow 16(5):70–84
Checkland PB (1978) The origin and nature of hard system thinking. J Appl Syst Anal 5(6):99–109
Checkland P (1981) Systems thinking, systems practice. Wiley, Chichester, pp 163–164
Checkland PB, Scholes J (1990) Soft systems methodology in action. Wiley, Chichester, UK
Fiol CM, Huff AS (1992) Maps for managers: where are we? Where do we go from here? J Manag Stud
29(3):267–285
Flood RL, Jackson MC (1991) Creative problem solving: total systems intervention. Wiley, Chichester
Hjortsø CN, Christensen SM, Tarp P, Rapid A (2005) Stakeholder and conflict assessment for natural
resource, management using cognitive mapping: the case of Damdoi Forest Enterprise, Vietnam. Agric
Hum Values 22(2):149–167
Hossain S, Brooks L (2008) FCM modeling educational software adoption. Comput Educ 51(4):1569–1588
Ingram H (2000) Using soft systems to manage hotel. Manag Serv Qual 10(1):6–10
Irani Z, Sharif A, Love PED, Kahraman C (2002) Applying concepts of FCM mapping to model: the IT/IS
investment evaluation process. Int J Prod Econ 75(1):199–211
Jackson MC (2003) System thinking: creative holism for managers. Wiley, Chichester
Kardaras D, Karakostas B (1999) The use of FCMs to simulate the information systems strategic planning,
process. Inf Softw Technol 41(4):197–210
Khan M, Quaddus MA (2004) Group decision support using FCMs for causal reasoning. Group Decis Negot
13(5):463–480
Khan MS, Quaddus MA, Intrapairot A (2001) Application of a FCM for analyzing data warehouse diffusion,
applied informatics: artificial intelligence and applications; advances in computer applications In:
Proceedings of IASTED international symposia, pp 32–37
Kosko B (1986) FCMs. IntJ Man-Mach Stud 24(1):65–75
Kosko B (1987) Adaptive inference in fuzzy knowledge networks. In: Proceedings IEEE international
conference on neural network,San Diego, pp 261–268
Kosko B (1988) Hidden patterns in combined and adaptive knowledge networks. Int J Approx Reason Arch
2:377–393
Kosko B (1992) Neural networks and fuzzy systems: a dynamical systems approach to machine intelligence.
Prentice-Hall, New York
Kosko B (1994) The new science of fuzzy logic: fuzzy thinking. HarperCollins, London
Lane DC, Oliva R (1998) The greater whole: towards a synthesis of the system dynamics and soft system
methodology. Eur J Oper Res 107(1998):214–235
Lee K, Chung N (2006) Cognitive map-based web site design: empirical analysis approach. Online Inform
Rev 30(2):139–154
Lin C, Wu W (2008) A causal analytical method for group decision making under fuzzy environment. Exp
Syst Appl 34(1):205–213
Meghabghab G (2002) Fuzzy cognitive state map vs. Markovian modeling of user’s web behavior, invited
paper. Int J Comput Cogn 1(8):51–92
Mingers J (1984) Subjectivism and soft systems methodology—a critique. J Appl Syst Anal 11:85–103
Mirijamdotter A, Bergvall-Kårebirn B (2006) An appreciative critique and refinement of Checkland’s soft
systems methodology. In: Strijbos A (ed) Search of an integrative vision of technology—interdisciplinary studies in information systems SaB. Editor Springer, New York, pp 79–102
Montazemi A, Conrath DW (1986) The use of cognitive mapping for information requirements analysis.
MIS Quart 10(1):45–58
Munro I, Mingers J (2002) The use of multimethodology in practice-results of a survey of practitioners.
J Operat Res Soc 53(4):369–378
Özesmi U, Özesmi S (2004) Ecological models based on people’s knowledge multi-step FCM modeling
approach. Ecol Model 176(1–2):43–64
Park K, Kim S (1995) Fuzzy cognitive maps considering time relationships. Int J Hum Comput Stud
42(2):157–168
Platt A, Warwick S (1995) Review of soft systems methodology. Ind Manage Data Syst 95(4):19–21
123
354
Syst Pract Action Res (2011) 24:325–354
Rodr0 ıguez-Ulloa RA, Paucar-Caceres A (2005) Soft system dynamics methodology (SSDM): combining
soft systems methodology and system dynamics (SD). Syst Pract Act Res J 18(3):303–334
Rose J (2002) Interaction, transformation and information systems—an extended application of soft system
methodology. Inf Technol People 15(3):242–268
Shapiro F, Shapiro R (2003) Towards and improved collaboration model for the national healthcare system
in England and Wales: a critical and constructive approach using operational research. Logist Inf
Manage 16(3/4):246–258
Siau K, Tan X (2005) Improving the quality of conceptual modeling using cognitive mapping techniques.
Data Knowl Eng 55(3):343–365
Smyth DS, Checkland PB (1976) Using a systems approach: the structure of root definitions. J Appl Syst
Anal 5(1):75–83
Taber WR (1991) Knowledge processing with FCMs. Exp Syst Appl 2:83–87
Taber WR, Siegel MA (1987) Estimation of expert weights using FCMs. In: Proceedings of the 1st IEEE
international conference on neural networks, San Diego, pp 319–325
Tsadiras AK, Kouskouvelis I, Margaritis KG (2001) Making political decisions using FCMs: the FYROM
crisis. In: Proceedings of the 8th Panhellenic conference on informatics, Cyprus, pp 501–510
Vasantha WB, Ram Kishore M (1999) Symptom disease model in children. Ultra Sci 11(3):318–324
Wilson B (1993) Systems: concepts, methodologies, and applications. Wiley, UK
Yinghong Z (2007) Soft systems methodology based on decision making knowledge integration. In: Proceedings of IEEE international conference on the wireless communications, networking and mobile
computing conference (WICOM), Shanghai, pp 21–25, Sep 2007
123