Service Development using Fuzzy QFD in the banking industry

2.1 Quality Function Deployment

Fig. 1 shows the House of Quality (HOQ), the basis for the QFD concept. The customer requirements (CRs) or “whats” are listed in the left wing of the HOQ. Each of these requirements has an importance value elicited from customers usually through surveys, interviews, or focus groups (Sireli et al. 2007). The importance values are located in the third column of Fig. 1. The central element of the HOQ in Fig. 1 is the relationship matrix. The HOQ lists service characteristics (SCs) or “hows” horizontally along the top of the relationship matrix. The relationship matrix describes the strength of the relationship between each pair of variables CR and SC. The “roof” represents the correlation between each pair of SCs, while the far right wing of the HOQ shows the competitive evaluation of competing alternatives.

Figure. 1.

The house of quality

At the bottom of the HOQ are the importance weights of SCs. The importance weight of each SC for customer requirements is quantitatively computed by matrix row and column operations. Importance weight IW_j for SC_j, j = 1, 2, 3, …, m, can be calculated by

where W_i = the degree of importance for CR_i, i = 1, 2, ···, 1,

l = the number of CRs, and

r_ij = the numerical value representing the strength of the relationship between CR_i and SC_i.

The QFD evaluates the impact values of service characteristics (SCs) on meeting customer requirements (CRs) by prioritizing the SCs based on their importance weights (Sireli et al. 2007).

The construction of the HOQ starts with customer requirements. The SERVQUAL is introduced to identify key customer requirements in the banking industry. The SERVQUAL is also used to measure customer satisfaction of bank services.

2.2 SERVQUAL

Service quality has become an important research topic, because of its relationship with costs, profitability, and customer satisfaction and retention (Buttle 1996). However, it is often challenging to measure service quality. The SERVQUAL has been used as a survey instrument to measure service quality (Sahney et al. 2004). The SERVQUAL has the following five dimensions (Buttle 1996, Park et al. 2012):

The SERVQUAL consists of 22 statements in the five dimensions. Four to five statements are used for each dimension. The survey results are collected into the five dimensions.

When the SERVQUAL is used for a specific problem, however, it should be modified based on the expectations of customers (Paryani et al. 2010). An adaptation of their scale may be desirable when a particular service is dealt with (Parasuraman et al. 1988).

2.3 Fuzzy Theory

Let the universe of discourse X be the subset of real numbers R. Fuzzy set A=xμAxx∈X,μAx∈0,1 is a set of ordered pairs, in which μ_A(X) is called a membership function. The membership function can take a value from the closed interval [0,1]. The greater μ_A(X) is, the stronger the truth of the statement that element x belongs to set A is. The function μ_A(X) is defined by equation (2) with a triplet (c₁,c₂,c₃):

Each linguistic term can be characterized by a triangular fuzzy number that represents its approximate value range between 0 and 7, and denoted as(c₁,c₂,c₃), where ≤c₁≤c₂≤c₃≤7 (Verkuilen 2005; Benitez et al. 2007; Erginel 2010). c₁ and c₃ are the lower and upper bounds, and c₂ is the most likely value of the linguistic term, respectively.

Table 1 shows the default values of the linguistic terms used in this paper (Verkuilen 2005; Erginel 2010). The membership functions can be calculated by equation (2). Linguistic terms such as satisfaction and importance degrees can be fuzzified by triangular fuzzy numbers. For each survey statement, group opinions of n respondents are aggregated as the average fuzzy number of n triangular numbers by equation (3):

Table 1.

Triangular fuzzy numbers

where • is the multiplication of a scalar and a fuzzy number, A_k = (c₁^(k), c₂^(k), c₃^(k)) k = 1, 2, 3..., n, and ⊕ is the add operation of fuzzy numbers (Bandemer and Gottwald 1995). A is the average of n responses for a survey statement. Equation (3) shows that the overall average can be represented by a new triangular fuzzy number (Buckley 1985).

The last step is to defuzzy the information obtained by equation (3). A fuzzy number is the result of fuzzy synthetic decision of each alternative (Tsaur et al. 2002). Thus, defuzzification is required to convert the fuzzy number into a crisp real number. The procedure of defuzzification is to locate the Best Nonfuzzy Performance (BNP) value. This research uses equation (4) for defuzzification (Chen 1996):

Z_A is used to construct the HOQ in this paper. That is, Z_A is used to represent the levels of importance and satisfaction for each CR in the HOQ.

4.1 Customer Requirements (CRs)

The SERVQUAL (Parasuraman et al. 1988) was adopted and modified to construct part of the HOQ for the banking industry based on a literature survey (Ko and Lee 2000; Shih et al. 2011; Tsai et al. 2011). The modified SERVQUAL was used as customer requirements (CRs) for the HOQ. Table 2 shows the details (Yu and Kwak 2015).

Table 2.

Customer requirements

In addition to basic personal information such as gender, age, and main bank, the survey for this study had two sections: ‘the level of importance for each CR’ and ‘the level of satisfaction for each CR.’ The levels of importance represent the areas of greatest interest and highest expectations (Evans and Lindsay 2008). On the other hand, the levels of satisfaction are to measure what customers perceived from their banks. In the two sections, a 5-point scale was used for all survey statements: very low, low, moderate, high, and very high. The fuzzy set theory was introduced as an effort to resolve the subjectivity and ambiguity of customer requirements and satisfaction in the survey results. For example, even if several respondents choose the same point ‘high’ on the likert scale for CR1, ‘jobs are accurately done,’ in the importance section, there can be ambiguity in their views meaning different values even in the same choice ‘high.’ In addition, there is ambiguity on the border between two scale points, for instance, ‘high’ and ‘very high (Hedayatpanah 2011).’

When a response to CR1, ‘jobs are accurately done,’ was ‘high’ in the importance section, for example, this can be converted to fuzzy number (4.5, 5.5, 6.5) from Table 1. Next, average A can be obtained by equation (3) from 200 responses for CR1. Finally, A can be defuzzified into Z_A by equation (4). In this way, importance for each customer requirement can be obtained as in Table 3. The defuzzified figures in Table 3 are used as importance for CRs in the HOQ.

Table 3.

Importance for each CR in fuzzy and defuzzified numbers

Table 4 shows the level of satisfaction for each bank and the overall satisfaction level, both for each CR in defuzzified figures. The defuzzified figures in Table 4 are obtained by using the same procedure as in Table 3. The overall satisfaction level is derived not only from the three banks but also from the other 16% of the banks in the survey. The levels of satisfaction for each bank are used for competitive evaluation in the HOQ. In the competitive evaluation, the three banks are relatively compared on the same scale for each CR. Of the three banks considered in this study, Nonghyup turned out to be consistently weak in customer satisfaction despite its many branches as a government-invested bank.

Table 4.

Satisfaction for each CR in defuzzified numbers

For example, the level of satisfaction for Nonghyup is lower than those for the other two banks in CR2, ‘Customers are kept posted on job progress.’ Several banks use text messages for the progress of special tasks such as loan approvals that take several days or weeks. Nonghyup needs to actively consider not only text messages but also smart phone applications with a notice function to provide customers with more detailed information. Nonghyup also has very low levels of satisfaction for CR6, ‘Employees are knowledgeable about their jobs, and CR9, ‘Employees provide customers with confidence,’ compared with the other two banks. Nonghyup needs to reinforce training programs to boost employees’ knowledge and skills. In addition, Nonghyup should establish a new evaluation system for employees and quality control methods for employees’ services.

Many banks actually investigate customer satisfaction for their services. When this does not work very well, Nonghyup needs to go further to control the quality of its services more systematically. That is, Nonghyup should not only evaluate the service quality of the frontline employees who deal directly with customers, but also improve its overall service processes behind the frontline employees.

Compared with Kookmin bank, both Nonghyup and Busan Bank have low levels of satisfaction for CR11, ‘Employees are always ready to respond to customers’ requests.’ Above all, both banks need more training programs to improve employees’ attitudes towards customer service. They also need to reinforce online systems to deal with CR11. Banks have traditionally used web sites and phone services to resolve customer requests online. With smart phone applications that enable one-on-one talks, customers can contact their banks for their requests anytime anywhere. Busan Bank also has the lowest level of satisfaction for CR13, ‘Individual attention is provided to customers,’ compared with the other two banks. Among others, it can consider the option of an enlarged screen of automatic teller machines for elderly people. It can also introduce more facilities for disabled people. For example, web sites of the bank can be designed better for blind or visually impaired people using screen readers.

It can be said from Tables 3 and 4 that a bank with higher levels of satisfaction for important CRs can satisfy customers more efficiently. Important CRs are the ones with higher levels of importance in Table 3. Every bank has limited resources. Thus, it needs to focus on important CRs first. It is efficient to meet more important CRs than less important CRs with the same limited resources for more customer satisfaction. The more satisfied customers are, the bigger customers’ loyalty and intention to continue their business are.

4.2 Relationship Matrix and Correlation Matrix

The next step is to extract service characteristics required to meet customers’ requirements. Several interviews and brainstorming meetings were carried out with two financial experts in the banking industry to generate 21 service characteristics that were directly related to customer requirements. For example, CR2, ‘Customers are kept posted on job progress,’ is related to several SCs including computer systems and SMS notification service, and so on. CR7, ‘Security systems are constructed very well,’ on the other hand, is related to SCs such as modern facilities and the financial status of the bank. Computer systems, internal regulations, and customer call center were related to CR15, ‘Employees are willing to resolve customers’ requests.’ Extracted for CRs related to physical elements were SCs such as parking facilities, branch interior, the size of the bank, automatic teller machines and so forth. Twenty one SCs were finalized in total by the two financial experts as shown in Table 5.

Table 5.

Service characteristics

Based on the customer requirements and service characteristics, the relationship and correlation matrixes of the HOQ were constructed with importance weights of the SCs. The opinions of two financial experts in the banking industry were also used in this stage. Fig. 2 shows the correlation matrix that provides the correlation between any pair of SCs. Four symbols are used to represent the correlations. This research uses the symbol ▲ to denote a strong positive correlation, ◇ for a positive correlation, ● to denote a strong negative correlation, and ⧈ for a negative correlation. For example, SC13, automatic teller machines, has a positive correlation with SC12, the size of the bank, because bigger banks have more automatic teller machines. On the other hand, the SC13 has negative correlations with SC9, workloads, and SC7, the number of employees. That is, automatic teller machines lead to reducing the number of employees or workloads.

Figure. 2.

Correlation matrix

Fig. 3 presents the relationship matrix between CRs and SCs. The symbol △ indicates a weak relationship with 1 point, O denotes a medium relationship with 3 points, and ◎ represents a strong relationship with 5 points. For example, CR6, ‘employees are knowledgeable about their jobs,’ has a medium relationship with SC10, internal regulations, with 3 points. The CR6 has strong relationships with SC5, employees training, and SC6, knowledge sharing among employees, all with 5 points.

Figure. 3.

Relationship matrix

The degree of importance for each CR is provided in Table 3. The degrees of importance for CRs are combined with the values of the relationship matrix to calculate importance weights. The importance weight for each SC can be calculated by equation (1). The results obtained by equation (1) are presented at the bottom of Fig. 3 and in Table 6. In addition, normalized importance weights are calculate and added into Fig. 3 and Table 6. The normalized importance weights are calculated by setting the maximum value to 100, the minimum value to 0, and intermediate values to relative values between 0 and 100. SC5, Employee training, received the highest points, 505, followed by SC8, Continuous customer management, with 490 points, SC9, Workloads, with 422 points, and SC4, Customer call center, with 360 points. The importance weights in Table 6 can be used to determine the priority of quality improvements for service characteristics in the process of service development. The normalized importance weights are used to catch the relative differences of the importance weights better.

Table 6.

Importance weights for service characteristics

Fig. 4 shows the constructed HOQ. The HOQ presented in Fig. 4 indicates that employee training and close human relationships with customers should be emphasized, while workloads for each employee need to be reduced to boost service quality for customers. Such a result should be considered for maximum customer service under limited resources.

Figure. 4.

The first house of quality

5. Additional Analysis with the Importance-Satisfaction Model

5.1 Importance Satisf action Model and New HOQ

The importance satisfaction model maps the area for improvement to focus on the key attributes where a company should make its efforts to improve the service quality, as in Fig. 5 (Yang 2003). For example, an item with low importance does not need to be improved earlier than other items with high importance. As a result, time and cost can be efficiently spent to meet customer requirements and boost service quality in new service development.

Figure. 5.

Importance-satisfaction Model

The X-axis indicates importance, and the Y-axis represents satisfaction in the importance satisfaction model as shown in Fig. 5. Starting from the upper right quadrant, the numbering goes counter-clockwise from the first quadrant up to the fourth quadrant. The importance satisfaction model is used to locate the area for improvement that corresponds to the fourth quadrant in Fig. 5. Fig. 5 plots 8 CRs in the area: CR5, CR7, CR10, CR11, CR12, CR13, CR16, and CR17. For example, CR7, ‘Security systems are constructed very well,’ has the highest degree of importance, 5.989, while the CR7 remains at a relatively low level of satisfaction, 4.669. The 8 CRs mean that, although customers consider them very important, their overall service quality remains low. Thus, service quality should be improved urgently in the 8 CRs.

Fig. 6 shows a new HOQ constructed with the 8 CRs from Fig. 5 and all 21 SCs. The correlation matrix of Fig. 6 is the same as that of Fig. 4. The relationship matrix of Fig. 6 is exactly the same as the 8 rows for the 8 CRs of Fig. 4. However, the importance weights for the 21 SCs are newly calculated by equation (1) with the 8 CRs only in Fig. 6.

Figure. 6.

The second house of quality

5.2 Comparison between the f irst and second HOQs

Table 7 presents two lists of the top 5 service characteristics based on the importance weights from the two HOQs of Fig. 4 and Fig. 6. Table 7 shows that SC5, employee training, is most important in both the first and second HOQs. In addition, SC8, continuous customer management, SC9, workloads, and SC4, customer call center, are included in both lists with different rankings. However, SC6, knowledge sharing among employees, in the first list is excluded from the second list, while the second list newly includes SC7, the number of employees, that is in 6th place in the first HOQ.

Table 7.

Top five service characteristics

Table 7 leads to the final list of service characteristics requiring urgent quality improvements through ‘a confluence of results’ from a pair of houses of quality. That is, employee training, continuous customer management, workloads, and customer call center should be considered urgently to improve service quality. Knowledge sharing among employees and the number of employees can be taken into account additionally in the final list, because they are included in either of the two top five lists. The final list can be used as a managerial guideline when managers make their strategic decisions in the process of service development, because the results from a pair of houses of quality provide a bigger support for their decisions than depending solely on the results of a single HOQ. This is an effort to boost customer satisfaction and enhance competitive strength by improving service quality under limited resources.