Causal analysis of customer needs in the banking system by applying fuzzy group decision making

Effective customer satisfaction investigation is a very important precondition in the banking system. It is needed to get the knowledge of how to use advanced methods to identify customer preferences, classification and prioritization of banks and how to use the evaluation result to improve their quality services. Iranian Bank also plays a leading role in providing financial facilities in which satisfaction of customers is top priority. The general contribution of this article proposes concepts, methods and models to conceive the important criteria affecting the customers’ satisfaction in banking systems based on Delphi method. Classifying these criteria based on Kano Model, the Kano Model organized important criteria on how they are perceived by customers that are categorized by basic, performance and excitement needs. This article tries to introduce a group decisionmaking method, named DEMATEL in fuzzy environment, to determine both direct and indirect relationships between criteria and prioritization of banks. At the end, the most important criteria in each group are determined.


INTRODUCTION
Banking sectors almost offers financial services to their customer.For the last two decades, due to an increasingly competitive, saturated and dynamic business environment, banks in many countries have adopted customer-driven philosophies to address the rapid and changing needs of their customers (Walker et al., 2008;Al-Eisa et al., 2009).In fact the notation of customer need, satisfaction and loyalty in modern banking industries as well as other industries has emerged as a key factor for survival and growth.Analysis of customer needs is an important task with focus on the interpretation of the voice of customer and subsequently derivation of the explicit requirements that can be understood by marketing and engineering (Jiao and Chen, 2006).In general, customer needs analysis involve three major issues, namely; (1) Understanding of customer preferences.
(3) Requirements prioritization.Among many approaches that address customer needs analysis the Kano model has been widely practiced in industries as an effective tool of understanding of customer preferences owning to its convenience in classifying customer needs base on survey data (Kano et al., 1984).Nevertheless, Kano models are not equipped with quantitative assessment for requirements prioritization (Xu et al., 2008).Customer needs can be evaluated according to the different criteria, the multi-criteria decision making (MCDM) approach is suitable for evaluating customer expectation.Based on the various points of view or the suitable measuring method, the criteria can be categorized into distinct aspects (Tseng, 2008).So in this study we apply DEMATEL method for requirements prioritization.DEMATEL method, as a sort of structural modeling approach, can separate the involved criteria of a system into the cause group and effect group (Lin et al., 2004).
The DEMATEL method is based on digraphs, which can separate involved factors into cause group and 1) Identifying Preferences effect group.Directed graphs, known as digraphs, are more useful than directionless graphs because digraphs can demonstrate the directed relationships of subsystems (Wu and Lee, 2007).The characteristic of this method is to show the relationship with certain scores by using matrix operation.The relationship is cause-effect relationship.Another characteristic is to grasp not only direct effect but also indirect effect.Hence, the DEMATEL method has been successfully applied in many fields.But crisp values are an inadequate reflection of the vagueness of the real world (Bellman and Zadeh, 1970;Zadeh, 1965).So we apply the concept of fuzzy theory to the DEMATEL method for solving MCDM problems and expand this method by using fuzzy approaches, fuzzy DEMATEL approach is the evaluation of ranking problem in real world systems that are very often uncertain or lack of Information (Tseng, 2009).Knowing this, a framework which considers both factors and applies the graph theory based on fuzzy DEMATEL method combined with Kano models is provided.So banking industries would be prevented from spending on the requirements which have not any effect on customer satisfaction.In this way it would be recognized which are the most effective ones in achieving more efficiency and effectiveness.The conceptual model of the technique has been illustrated in Figure 1.

CUSTOMER SATISFACTION AND KANO MODEL
Customer satisfaction is an essential factor in acquiring loyal customers who would also recommend their regular establishment to other customers.What is critical to customer satisfaction is that someone listens to the customer's complaint and suitable responses to this voice.The customer needs to be heard and responses should be given.The review of the research recognized that there is connection between service quality, customer needs, satisfaction and loyalty in banking industry.

Customer satisfaction
Satisfaction is a person's feeling of pleasure or disappointment resulting from comparing products/services perceived performance interrelation to his/her expectation.As this definition makes it clear, satisfaction is a function of perceived performance and expectation (Kotler, 2001).It is well established that satisfied customers are key to long term business success (McColl-Kennedy and Schneider, 2000).Companies that have a more satisfied customer base also experience higher economic returns (Yeung et al., 2002).High consumer satisfaction leads to greater customer loyalty, which in turn, leads to future revenue (Bolton, 1998).Organizations having superior service quality have been found to be market leaders in terms of sales and long-term customer loyalty and retention (Eklo et al., 2002).Because of this, organizations competing in similar market niches are compelled to assess the quality of the services they provide in order to attract and retain their customers (Gilbert et al., 2006).Costs of attract a new consumer five times more than to retain an old one (Vukmir, 2006).Banking is one of the many service industries, characterized by high customer contact with individually customized service solutions, where customer satisfaction has been an increasing focus of research.There has been extensive support for the general proposition that customer satisfaction is a significant variable for evaluation in bank management (Moutinho and Smith, 2000).Dove and Robinson's (2002) research indicated that banking customers have much higher satisfaction levels when they believe their problems with the bank have been resolved.The International Journal of Retail and Distribution Management (1995) reported that banking customers who have complained and are satisfied with the service recovery efforts of the bank are three times more likely to recommend the bank to someone else and to do increased business with the bank.Levesque and McDougall (1996) point out that customer satisfaction and retention are critical for retail banks.They investigate the major determinants of customer satisfaction (service quality, service features, customer complaint handling and situational factors), and future intentions in the retail bank sector (Molina et al., 2007).

Delphi method
The Delphi method has been variously described as "a technique designed to elicit opinions from a group with the aim of generating a group response" (Brown et al., 1969); as a "technique that elicits, refines, and draws upon the collective opinion and expertise of a panel of experts" (Gupta and Clarke, 1996); and as a method for "structuring a group communication process so that the process is effective in allowing a group of individuals, as a whole, to deal with a complex problem" (Linstone and Turoff, 1975).This method can be used when there is incomplete Knowledge about a problem or phenomena (Adler and Ziglio, 1996).The original Delphi method was developed by Dalkey of the Rand Corporation in the 1950's for a U.S. sponsored military project (Skulmoski et al., 2007) (2008).
Essentially this method structures and facilitates group communication that focus, upon a complex problem so that over a series of iterations, a group consensus can be achieved about some future direction (Loo, 2002).The Delphi method has five major characteristics identified by Dalkey included: The use of multiple experts; the application of a dialectic reflective process; the anonymity and independence of researchers; the focus is on outliers rather than on commonalities; and finally the process is iterative (Gatfield et al., 1999).The essential elements of the Delphi process, as it impacts on the research include: problem definition and study question, panel selection, panel size and conducting Delphi rounds (Loo, 2002).The study question must be carefully defined, so that the information gathered will be useful to the researchers and will answer the questions that the researchers really wanted to be answered.
In the traditional Delphi method, the first questionnaire takes the form of one or two open-ended questions that allow study participants to elaborate freely on their response.The researcher then captures the information from these free-form responses and compiles the results into the second iteration.
A Delphi study requires significant effort on the part of study participants, and panelists should be chosen for their interest in the problem that is being investigated and their expertise in the field of study.The panel size can vary, but should consist of a sufficient number of participants willing to complete the entire study and to provide enough data for analysis (Howze et al., 1999).Martino suggest that 15 to 30 experts could be used for a heterogeneous population and as 5 to 10 for a homogeneous population (Martino, 1972).The panel size should also take into account the complexity of the problem being studied, the range of expertise required to address the problem, and the purpose of the study (Loo, 2002).In relation to this matter some Delphi studies have used large panel numbering over 100 (Chaney, 1987) or even 345 members (Lecklitner, 1984).And finally When discussing the number of iterations, Linstone and Turoff (1975) state that the process might take five iterations but, with suitable methodological improvements, this is usually reduced to three (de meyrick, 2003).The three round Delphi method is shown in Figure 2.

Kano model
Kano is a model that provides an effective tool to categorize needs and to understand the nature of them (Matzler and Hinterhuber, 1998).Inspired by Herzberg's Motivator-Hygiene Theory in behavioral science, Kano and his colleagues developed the theory of attractive quality in 1984 (Kano et al., 1984) to categorize the attributes of a product or service, based on how well they are able to satisfy customers' needs.This model has been applied within quality management, product development, strategic thinking, employee management, business planning, and service management (Witell and Lofgren, 2007).The Kano model tries to explain how customer satisfaction will change as customer requirements are met by the organization (Bayraktaroglu and Ozgen, 2008).
Kano model is also dynamic in that once introduced, the exciting feature will soon be imitated by the competition and customers will come to expect it from every other provider (Shahin, 2004).The model allows researchers to gain a deeper understanding of customer preferences by analyzing how they evaluate and perceive product or service attributes (Gruber et al., 2008).Kano groups can be examined mainly in three categories (1) Must-be (basic) needs.Must-be needs can be defined as the basic criteria for service quality in terms of customer satisfaction.Consequently, if these requirements are not fulfilled properly, the customer will be extremely dissatisfied.So, fulfilling the ''must be' requirements will only create a state of ''not dissatisfied'' (Matzler and Hinterhuber, 1998).These needs are so fundamental that they are not expressed by the customer.However, they must be identified since they are very important for the customer (King, 1995).
(2) One-dimensional (performance) needs.One-dimensional criteria result in satisfaction when fulfilled and result in dissatisfaction when not fulfilled (Kano et al., 1984).If these needs are satisfied with improvement in their performance, the customer satisfaction will increase.The better the performance, the happier the customer is.These kinds of needs are generally expressed by the customer (King, 1995).Performance factors are both a necessary and sufficient condition for customer satisfaction (Baki et al., 2009).
(3) Attractive (excitement) needs.These are dreams of customers so they are not expressed by them.The absence of the attribute does not cause dissatisfaction because the customers are unaware of these needs.If these needs are met, the product/service satisfies and delights the customer (King, 1995).Meeting attractive needs will provide competitive advantage for the organization and the organization will find the opportunity to differentiate itself in the competition.Kano refers to this group as "surprising quality".This relationship is shown in Figure 3.
Various approaches have suggested for attributes classification.Almost four of these approaches are applied in classification of attributes.These approaches consist of (Witell and Lofgren, 2007).

Crisp DEMATEL methods
The DEMATE method, developed by the Science and Human Affairs Program of the Battelle Memorial Institute of Geneva

Very satisfied
Attractive Must-be One -dimensional

Degree of achievement
Customer satisfaction  2007), Bayraktaroglu et al. (2008).between 1972 and 1976, was used to research and solve complicated and intertwined problem groups (Fontela and Gabus, 1976).This method was aimed at the fragmented and antagonistic phenomena of world societies and searched for integrated soluions, especially practical and useful for visualizing the structure of complicated causal relationships with matrices or digraphs (Lin and Wu, 2004).DEMATEL was developed in the hope that pioneering the appropriate use of scientific research methods could improve the understanding of a specific problematic, a cluster of intertwined problems, and contribute to the identification of workable solutions through a hierarchical structure.The DEMATEL method is based on graph theory, enabling us to plan and solve problems visually, so that we may divide the relevant factors into cause and effect groups in order to better understand causal relationships.The methodology can confirm interdependence among variables and aid in the development of a directed graph to reflect the interrelationships between variables (Li and Tzeng, 2009).Step of crisp DEMATEL method is described as follows: Suppose that a system contains a set of criteria C = {C1, C2, . ..,Cn}, and there are h experts available to solve a complex problem (Li et al., 2009).
Step 1: Measuring the relationship between criteria requires that the comparison scale be designed as five levels: 0 (no influence), 1 (very low influence), 2 (low influence), 3 (high influence), and 4 (very high influence).Experts generating the direct-relation matrix make sets of the pair wise comparisons in terms of influence and direction between criteria (Tseng, 2009).
Step 2: On the basis of the direct-relation matrix A, the normalized direct-relation matrix X can be obtained through the following formulas (Tseng, 2009): Step 3: The scores given by each expert give us an n×n nonnegative answer matrix Z k , with 1 ≤ k ≤ h.Thus Z 1 , Z 2, …Z k are the answer matrices for each of the h experts, and each element of is an integer denoted by .The diagonal elements of each answer matrix Z k are all set to zero.We can then compute the n×n average matrix A by averaging the h experts' score matrices.The (i,j) element of matrix A is denoted by (Li and Tzeng, 2009). (3) Step 4: Once the normalized direct-relation matrix X is obtained, the total-relation matrix T can be acquired by using Formula (3), in which I is denoted as the identity matrix.

T= (4)
Step 5: The sum of rows and the sum of columns are separately denoted as vector D and vector R in the following equations.The horizontal axis vector (D+R) named ''Prominence'' is made by adding D to R, which reveals how much importance the criterion has.Similarly, the vertical axis (D-R) named ''Relation'' is made by subtracting D from R, which may group criteria into a cause group.Or, if the (D-R) is negative, the criterion is grouped into the effect group.Therefore, the causal diagram can be acquired by mapping the dataset of the (D+R, D-R), providing valuable insight for making decisions (Tseng, 2009).

Fuzzy theory
In 1965 Lotfi A. Zadeh introduced the theory of fuzzy sets and fuzzy logic; these terms were coined by him to deal with the phenomenon of vagueness, in the cognition process of the human being.According to Zadeh; the theory of fuzzy sets is a step toward a rapprochement between the precision of classical Mathematics and the pervasive imprecision of representing the uncertainty inherent in real world a rapprochement born of the incessant human quest for a better understanding of mental processes and cognition.Fuzzy database, fuzzy sets and fuzzy logic techniques have been successfully applied in a number of applications: computer vision, decision making, and system design including ANN training (Lakhmi et al., 1998).
Definition: Consider a classical set A of the universe U. A Fuzzy set A is defined by a set or ordered pairs, a binary relation: Where is a function that called membership function.
Definition: In fuzzy sets, each element is mapped to [0, 1] by membership function.This function shows the grade or degree to which any element x in A belongs to the Fuzzy set A (Lee, 2005).
Definition: The α-cut set A α is made up of members whose membership is not less than α .

= {x X | }
Definition: A fuzzy set is called normalized when at least one x A attains the maximum membership grade 1; otherwise the set is called none normalized.Assume the set A is none normalized, then max (Bojadziev, 2007): < 1 Definition: Consider the universe U to be the set of real numbers R. A subset S of R is said to be convex if and only if, for all x1, x2 ϵ S and for every real number satisfying 0 1, we have (Bojadziev, 2007): λ .
Definition: If a fuzzy set is convex and normalized, and its membership function is defined in R and piecewise continuous, it is called "fuzzy number".
Definition: Triangular fuzzy number is a fuzzy number represented with three points such as: A=(a1, a2, a3).This representation is interpreted as membership functions.This summation formula can be extended for n triangular numbers.Also it can be applied for left and right triangular numbers (Bojadziev, 2007).

Theorem (2):
The product of a triangular number A with a real number r is also a triangular number: Theorem (2): The division of a triangular number A with a real number r is defined as multiplication of A by 1/r provided that r≠0 hence gives (Bojadziev, 2007): Theorem (3): The multiplication of two triangular fuzzy numbers A1 = ( ) and A2 = denoted by A1× A2 is defined by the membership function x a a a a a a a a a a a a a a x a a a a a a x a a a a a a x a a a a a a a a a a a a x a a a a x a a a a a a a a a a a a a a x a a a a a a x a a a a a a x a a a a a a a a a a a a x a a a a k=1, 2, …, n.Using addition of triangular numbers and division by a real number the triangular average is defined y (Bojadziev, 2007): Linguistic variables: One of the most important facets of human thinking is the ability to summarize information into labels of fuzzy sets which bear an approximate relation to the primary data.Linguistic descriptions, which are usually summary descriptions of complex situations, are fuzzy in essence (Dubois and Prade, 1980).The concept of linguistic variable introduced by Zadeh in the general framework of fuzziness approaches to serve to quantitative Fuzzy semantics, which has its substantial contributions to approximate reasoning and the analysis of complex systems and decision processes, a linguistic variable is a variable whose values are words or sentences in a natural or artificial language defines that a linguistic variable X is composed of the quintuple (Zadeh et al., 1975;Chen, 1998).(X,T, U, G, M), where T is the set of linguistic terms of X, U is the universe of discourse, G is the set of syntactic rules that generate the element T, and M is the set of semantic rules translated from T that correspond to the fuzzy subset of U (Galindo et al., 2006).

CASE STUDY BASED ON CONCEPTUAL MODEL
Based on the problem and purpose of this paper, the case study based on conceptual model includes the casual analysis of customers' satisfaction in banking system and the measurement of the direct/indirect relationships among customer needs criteria.We use this framework to find the key criteria for illustrating the most effective elements in customers' satisfaction based on fuzzy linguistic variables.

Understanding of customer preferences
According to the proposed model in first step, the Delphi study was used to identify the different customer preferences and combine the views of different experts in this context.Since a Bank has different kind of customers, for understanding customer preferences, we separated the bank customer's base on personality, age, gender, previous records and annual turnover in 16 heterogeneous subgroups, and selected the five participants in expert panel from each subgroup.After the panel size is determined and participants are selected then we find effective elements for achieving high customer's satisfaction by applying 3 round Delphi.This process occur over 6 months.At first we develop the research design and question.Based on question we provide first round questionnaire Eshlaghy et al. 8423 and distributed to members of panel.This questionnaire was unstructured, using broad, open question relating to their opinions upon which elements are affected on customers' satisfaction in Iranian banking system and sent in December 2007 from their Email address.In the end of first round, 69 elements are documented.
Basis on the results of the first round, two rounds of questionnaires, in the form of the checklist, were distributed to the members of the expert panel.Participants were asked to rate each question on a scale of 1 to 5 (1 meaning the element is totally inappropriate and 5 meaning the element is very important).During the second round, the results of the first round were given to all participants and 67% of participants responded to the questionnaires, for each question, rankings 1-5 equaled five cardinal numbers multiplied by 54 participants.The maximum points that could be assigned to an element by the research participant's, when their numerical ratings added together were 270.
Statistical analysis was based on measures of median plus upper and lower quartile.Results of round two was that, of the 69 elements presented in the first iteration, 18 elements received score upper medians plus quartiles (202.5),these means that participants agreeing or strongly agreeing on the importance of elements and were not carried over into the third round.From a total 51 remaining elements, 21 received score lower than quartiles (67.5) and deleted.In the third round, participants responded to the questionnaires with 30 questions for remaining elements (between 67.5 and 202.5) on a scale of 1 to 5 and 13 elements deleted from this round.Finally in the end of Delphi process 35 effective elements on customers' satisfaction was recognized.The customer division has been illustrated in Figure 5.

Requirements classification
In this paper for the classification of requirements the direct questions approach is used.This approach is the one of the easiest approach in classifying attributes.The classification procedure through this approach was initiated by an explanation of the Kano methodology to the each members of the expert panel (respondents).Then each respondent was asked to classify each requirement into a specific dimension.For example: How would you classify the numbers of bank branches?(a) Basic needs?(b) Performance needs?(C) Excitement needs?(d) Other?
In the subsequent steps, first the number of respondents who perceived an attribute as belonging to a certain dimension was counted.Secondly, the statistical mode and a t-test were used to investigate the strength of the classification process.The customer preferences and final result of classification is listed in Table 1.

Requirements prioritization
Step 1: Defining the evaluation criteria According to the purposed model after a set of criteria is developed, for evaluation of requirements, it is necessary to design the fuzzy linguistic scale.To adopt the fuzzy linguistic scale used in the group decision-making proposed by Chen definition, the linguistic variable is expressed: Positive triangular fuzzy numbers (a1, a2, a3) are shown in Figure 6 and Table 2.This table is the linguistic scale and their corresponding fuzzy numbers defined by Wang and Chang (1995) and used by Chen (1998), Lin and Wu (2004), Wu andLee (2007), Yi et al. (2007).

Step 2: Design the fuzzy linguistic scale
To measure the relationship between evaluation factors C= {C1, C2, …, Cn}, it is necessary to ask a group of experts to make assessments in terms of influences and directions between factors.In this step for measuring causal relationships and influence between elements, three questionnaires for each class of requirements are developed and presented to 12 members of expert panel, who complete in terms of fuzzy linguistic scale and return them.In Table 3 the linguistic evaluation of the third expert for the elements of the first group has been shown.As it has been shown in Table 3, the element in second column from the first row influences the "personals ability in consulting the customers into inquiry to the customer's protests, which has been indicated by the linguistic variable high (H), as the high effect rate".

Step 3: Establishing the directed-relation matrix
Once the relationships between those factors were measured by the members of expert panel through the use of the fuzzy linguistic scale, the data could be obtained from each individual assessment.The assessment data of the third expert for the elements of the first group are shown in Table 3.Then, using the triangular fuzzy numbers to aggregate these assessment data, the initial direct-relation matrix was produced.
Where denoted influence requirement i on requirement j in group n for expert k.According to Table 3, variable high (H), is equal to triangular fuzzy number (0.5, 0.75, 1.0) which is shown in the Table 4. Table 4 shows, 36 (3×12) fuzzy matrices each corresponding to an expert for each class of requirements and with triangular fuzzy numbers as its elements, were obtained.These matrices are: Basic needs direct-relation matrices , dimension of this matrices such as Table 3

Step 4: The normalized direct-relation fuzzy matrix
On the Base of initial direct-relation Fuzzy matrix Z, the normalized direct-relation Fuzzy matrix X can be obtained through Formulas (1), (2).Fuzzy shapes of these are presented in Formulas (14 to 16).  ( The calculated values for have been illustrated in Table 5 and the normalized direct-relation fuzzy matrix of the third expert in Group 1 ( ) is shown in Table 6.Then use Theorem (5) to achieve the general consensus in each group, the average normalized direct-relation fuzzy matrix has been calculated.
This average for the first group has been shown in Table 7.In the similar way the average has been calculated for each group.

Step 5: Compute the total-relation fuzzy matrix
According to this step the total-relation matrix is calculated in each group.The elements of fuzzy matrix are also triangular fuzzy numbers.Then the total-relation fuzzy matrix is equal to where is triangular fuzzy numbers that denoted total direct and indirect influence requirement i on requirement j in group n: As a result, the calculation of the matrix is for accessing the possible structure of direct and indirect relations, and then analyzing cause and effect relations.The sum of rows ( ) in fuzzy matrix denoted total influence requirement i on other requirements and the sum of columns ( ) denoted the total influence of other requirements on requirement i.The values of vector D and vector R are calculated by using Formula (5, 7).So the values of ( ), ( ), ( , ( can be acquired which shows in Table 8.
In many practical applications, for Multi Criteria Decision Making should be given a crisp value.Therefore it is needed to defuzzify the result of the fuzzy number.To convert the fuzzy numbers into a crisp value (Formula 13) the defuzzification of ( ), ( ), ( , ( is applied.The crisp value of ( and ( for each groups is shown in the Table 9. Finally for description of the causal relationship between requirements it is necessary to develop a casual diagram.To do this the x axes of each graph is aligned to the and Y axes is aligned to the , thus each requirement illustrates with one point on casual diagram.These diagrams are shown in Figure 4.

RESULTS
As shown in the causal diagram which have shown in Figure 7, the evaluation criteria were visually divided into the cause group and effect group The cause group criteria imply meaning of the influencing criteria, whereas the effect group criteria denote the meaning of the influenced criteria (Fontela and Gabus, 1976).In other words, the cause group criteria are difficult to move, while the effect group criteria are easily moved.The cause criteria in to the first group (basic needs) including C1, C2, C3, C4 and C7 while the effect criteria were composed of C5, C6 and C8.among these 8 criteria, "the tendency and consideration employees in answering to the customers" (C3), with the largest value of & has the best effect on the other criteria and the most important factor for satisfy customer basic needs, this means the more employees tendency in answering the less customers dissatisfaction.By contrast, using simple and appropriate application forms (C8), with the most negative value of improved  of the effect group criteria most easily.The cause criteria in to the performance needs group consist of C1, C3, C4, C8, C10 and C11.The performance needs affect criteria group C2, C5, C6, C7, C9 and C12 that can be improved.In to this group "The proficiency and skill of the employees" (C11) with the best positive value of has the more effect on the other criteria and the changes of this criteria is effected to change the other elements.But these changes could not be obtained easily.Similarity description of the first group "Presenting the loans" (C5) is not only with the largest negative value of which most easily improved and cause enhanced the consumer satisfaction but also with the largest value of would be the most important factor for satisfying customer performance needs.In the Group 3 attractive requirements are presented.Based on Kano definition, these requirements can be defined as the criteria which satisfy customers when it is present but do not dissatisfy when it is absent.The cause criteria in to "the attractive" needs group is including of C5, C6, C7, C8, C9, C13 and C15.Between these cause criteria "relational ability of the employees" (C7) with 0.8327 in has the greatest effect on the other criteria.And "accessing the management" (C15) with 3.3556 in is the most important criterion.Because it has the highest intensity relation to other criteria in this group and greatest influence on customer satisfaction.The effect criteria in these needs are C1, C2, C3, C4, C10, C11, C12 and C14.Amongst these "advertising and introducing the services" (C14) with -0.2846 in is a simplest criteria for changing and improving satisfaction.As result, if bank want to obtain high satisfaction in terms of the effect group factors, it would be needed to control and pay attention to the cause group of criteria beforehand.

Conclusions
This study has developed a hybrid comparative method based on Delphi, Kano and DEMATEL techniques for identification, classification, prioritization of the customer needs and recognize causal diagram as playing more significant role for important factors selection to support of customer satisfaction in Iranian banking system.Our proposed method successfully combined three techniques and extends the DEMATEL method by applying linguistic variables in three categories, so that it can deal with vague and fuzzy condition effectively.This comprehensive method is applicable to other industries that are required group decision making in a fuzzy environment, such as manufacturing, tourism, healthcare, cinema and many other industries which face multiple criteria problems.
In order to help a banking system that wants to verify its strategic objective and satisfy customer's needs, we found the effective element on customer's satisfaction and selected the most important of them in three groups based on causal diagrams, these elements have a direct impact on the establishment of strategic objective.According to the achieved results, the most influential costumer needs criteria's in each group are the human factors.
It can show the importance of human factors in service organizations such as a bank.Thus, the management's attention on these elements will result in achieving the observable consensus.This study provides a number of directions for further researches.These can be summarized as follows: (i) There are a number of techniques that can be applied to identify customer needs such as NGT and brain storming, future research may continue by these techniques.
(ii) Future research should focus and continue with a more complete investigation on identifying customer's needs factors that influence on satisfaction in other banking systems.(iii) Further research may represent linguistic variables by other types of fuzzy numbers such as trapezoidal or bell shape numbers or other types of membership functions.
: Let A1 = ( ) and A2 = be two positive triangular fuzzy numbers, A1×A2 approximates a triangular fuzzy number(Laarhoven and Pedrycz, 1983;Lin, 2008): Consider n triangular numbers AK= As in most of Fzzy model, we had to convert the final Fuzzy data into a crisp value.In this study we applied following formula for defuzzification.…, n; and ∆= R-L.Then(Opricovic and Tzeng, 2003):

Figure 5 .
Figure 5.The Classification of customers.
group n for expert k.

Figure 7 .
Figure 7.The causal diagram of needs.

Table 1 .
Customer preferences and classification based on Kano model.

Table 9 .
Values of ,for three groups of requirements.