CHARPTER 5 The Unequal Divided Scale
Section 2 Research Methodology
This study aims to modify the scale value of options on 5-points Likert scale to duly illustrate the nonlinear relationship between quality performance and patient satisfaction. Several scholars (e.g. Anderson & Sullivan, 1993; Oliva, Oliver, &
Bearden, 1995) and practitioners (e.g. Coyne, 1989) have agreed the nonlinear relationship of Customer Satisfaction and Quality Performance. Ting and Chen (2002) and Matzler, et al. (2004) further pointed out that the relationship is not only nonlinear but also asymmetric. Tan and Shen (2000) proposed a relationship expressed as s = f(u,p), where s, represents the customer satisfaction, p, represents the product or service performance and u is the adjustment factor for each Kano category. They argue the attractive attributes fulfill customer satisfaction much more effectively than must-be attributes do. That is to say, the input of attractive attributes would lead to more sharply increase of customer satisfaction comparing to the improvement level of the service performance. Therefore, to input attractive attributes would produce the outcome of △s/s > △p/p; the definition of △s and △p would be the small shifts of s and p respectively. It would get similar outcome applied to one-dimensional attributes as △s/s = △p/p and to must-be attributes as
△s/s < △p/p. In the simple way, it assumes that the relationship between △s/s and
△p/p would be linear. Consequently, there would have an equation, △s/s = u(△p/p), while using a parameter u to show the above three relationship formulae. To input attractive attributes will result in u > 1; to input one-dimensional attributes will result in u = 1; and to input must-be attributes will result in 0 < u < 1. It finally proposed the following equation to evaluate customer satisfaction:
s = cpu (10) where c is a constant for different categories.
1 The Customer Satisfaction Degree of Kano’s Model
Basing on Kano’s two dimensional model (see Figure 1), if the element determined as must-be attribute, it means that, at right area of Y-axis, the condition exceeds customer satisfaction level which will be scored as 0 point. At the left area of Y-axis, the condition does not fulfill customer satisfaction level which will be
scored as -1 point. In addition, if the element determined as one-dimensional attribute, it means that, at the right area of Y-axis, the condition exceeds customer satisfaction level which will be scored as 1 point. At the left area of Y-axis, the condition does not fulfill customer satisfaction level which will be scored as -1 point.
2 The Unequal Divided Scale Value Model: The Modification for Likert-type Scale
According to original 5 points Likert scale for questionnaire, equal divided scale value would be applied to calculating. For example, the questionnaire contains five options and the scale value is 1, 2, 3, 4 and 5 for each option respectively. These values will be applied to further calculating. However, based on Kano’s model, the form of questionnaire to evaluate the service quality will depend on the question itself categorized in which attribute. That is to say, it would lead to various proportions on calculating interviewees’ evaluations. It discloses that there exist satisfaction variances among each attribute. Some attributes reflect linearity whereas some reflect nonlinearity. In order to let evaluations’ calculation conform to Kano’s model, this study modifies the scale value of options.
In accordance with Tan and Shen (2000), they adjusted the value of u in equation (10) for two nonlinear attributes, must-be and attractive, by 0.5 and 2 respectively. And u value for linear attribute, one-dimensional, is 1. Besides, according to Shahin (2004), u value was suggested to be 0 for indifferent attributes.
In this study, we adopt the both ideas of Tan and Shen (2000) and Shahin (2004). It depends on experts’ experience and comprehension of the relationships to select u value. Nevertheless, no matter does the selected numerical value for u, but how it can appropriately reflect the real relationship.
As above, this study defines the X-axis value as the gap between customer expectation and perception instead of quality performance. Therefore, the p value of equation (10) represents the gap level between customer expectation and perception.
It is natural and convenient to assume that the value of p in equation (10) is between 0 and 1. As Table 16 shows, for customer satisfaction, the value of s in equation (10) is between -1 and 1. When input an attractive quality element, the canonical output will be that, the lowest gap level would be 0 with respect to the lowest customer satisfaction as 0. On the other hand, the highest gap level would be 1 with respect to the highest customer satisfaction as 1. Let s(a) and p imply the customer satisfaction
and the gap level of an attractive quality element input, then the canonical output would reflect on well adjusted equation (11) where obtained from equation (10) as follow:
s(a) = p2 (11)
Similarly, when input a one-dimensional quality element, the canonical output will be that, the lowest gap level would be 0 with respect to lowest customer satisfaction as -1. On the other hand, the highest gap level would be 1 with respect to the highest customer satisfaction as 1. Let s(o) and p imply the customer satisfaction and customer satisfaction level of a one-dimensional quality element input. The canonical output for lowest level value would be s(o) = -1 and p = 0, thus the adjustment for equation (10) is s(o) = cp-1. Then input the highest level value s(o) = 1 and p = 1, it would get c = 2. The well adjusted equation (12) as shown as follow:
s(o) = 2p-1 (12)
Thus, let s(m), s(i) and s(r) denote the customer satisfaction of must-be, indifferent and reverse attribute, following similar rules, the canonical output will reflect on equations (13), (14) and (15) respectively as below:
s(m) = p0.5-1 (13) s(i) = Constant (14) s(r) = 1-2p (15)
Since customer satisfaction is influenced by the attribute and quality elements, the attribute of certain quality element should be assessed before analyzing the questionnaire answered. When five-point Likert-type scale is adopted on a questionnaire, customer satisfactions corresponding to the different types of the categories proposed by Kano can be derived from equations (11)-(15). Let input the original equal divided scale value (C1 = 1, C2 =
4
3, C3 = 4
2, C4 = 4
1 and C5 = 0) to
the modified customer satisfaction expectation, it would generate the outcome as shown as Table 15.
Table 15
Modified five-points likert scale list
Judgment Customer Satisfaction
Sm So Sa Sr Si
Strongly agree C1 0 1 1 -1 0
Agree C2 1
4 3
2 1
16 9
2
1 0
Neither agree nor disagree C3 1 4
2 0
16
4 0 0
Disagree C4 1
4 1
2
1
16 1
2
1 0
Strongly disagree C5 -1 -1 0 1 0
3 The Evaluation Way of Customer Satisfaction
Erto and Vanacore (2000) suggest to employee quality experts, who are trained and experienced customer, enables to derive a statistical measurement of the service quality from the impact of intrinsic variation and interaction between the customer and supplier. The quality experts (QEs), indeed, participate in the service processes as customers, but their knowledge and higher experience reduce the variability of their behaviors and expectations. The experts record their experience on some evaluation tables, expressing the degree of their agreement to a specific statement about the effectiveness of each service quality element. All of the experts are knowledgeable about the characteristics of the questionnaires; therefore, a reliable and helpful estimate of the service quality will need very few collected data. After collecting the QEs’ evaluation result, it would quantify the relationships between the fulfillment of each quality element and satisfaction. Assuming that every element has the same impact, the values of means were used to evaluate each quality category. In customer’s point of view, there is no variance between satisfaction and dissatisfaction while providing indifferent attributes. Therefore, the contribution of indifferent attribute would be ignored. The outcomes of contribution of each quality element will apply AHP algorithm to get its weight of contribution. The overall customer satisfaction (CS) will be generated by the product of each contribution multiplying its
weight. For comparing the differences, this study also calculated the CS’ based on a 5-points original Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree) but standardization plus the same AHP weight.
()
100) ) (
(
level scalelower level
scaleupper
W level scalelower Y
CS x
x x
j j j
X (16)
Where x implies must-be, attractive or one-dimension
)
j( x is the number of x quality attribute
)
j( x
w is the weight of the jth quality element of x quality attribute
)
j( x
Y is the average score of the jth quality element of x quality attribute
X
X W
CS
CS
(17)Where WX is the weight of x quality attribute.
The main stages of survey on this study are as shown on Figure 7.