A novel framework for customer-driven service strategies: A case study of a restaurant chain

Tourism Management 41 (2014) 119e128

Contents lists available at ScienceDirect

Tourism Management journal homepage: www.elsevier.com/locate/tourman

A novel framework for customer-driven service strategies: A case study of a restaurant chain Li-Fei Chen* Department of Business Administration, Fu Jen Catholic University, No. 510, Zhongzheng Rd., Xinzhung Dist., New Taipei City 24205, Taiwan, ROC

h i g h l i g h t s
We
We
We
We

propose a qualityeperformance analysis for quality improvement strategies. develop the strategic positioning portfolio for service activity design. propose a signal-to-noise method for classifying Kano’s quality attributes. use a real case study to demonstrate the effectiveness of the proposed approach.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 18 April 2013 Accepted 4 September 2013

Importanceeperformance analysis (IPA) is a popular customer-driven tool that enables companies to understand market competition and identify improvement priorities for various attributes of products and services. Despite the widespread use of IPA, previous studies have identified specific deficiencies. For example, the managerial improvement directions derived from IPA are potentially misleading because they ignore the asymmetric and nonlinear relationships between attribute performance (AP) and customer satisfaction (CS). Furthermore, the relationship between AP and importance is erroneously assumed to be independent. By contrast, the Kano model offers useful insight into quality attributes based on the asymmetric and nonlinear relations between AP and CS. In this study, a customer-driven framework is proposed, integrating the advantages of traditional IPA and the Kano model to elucidate the market competition position of each service and product attribute, providing strategic improvement guidelines for managers to design service activities. By conducting a case study of a restaurant chain, we demonstrate the effectiveness of the proposed approach. Ó 2013 Elsevier Ltd. All rights reserved.

Keywords: Customer satisfaction Service quality Importanceeperformance analysis The Kano model Signal-to-noise ratio

1. Introduction The importanceeperformance analysis (IPA) was introduced by Martilla and James (1977) and has been a popular customer-driven tool among researchers and practitioners, elucidating the market competition of companies and facilitating the identification of improvement opportunities and strategic planning (Azzopardi & Nash 2013; Garver, 2003; Oh, 2001). Typically, IPA can be implemented by scoring the importance and performance of specific product or service attributes based on the voice of customers. These data were plotted on a matrix comprising four quadrants (Fig. 1). According to their positions on the matrix, the following improvement strategies can be recommended: (a) keep up the good work; (b) concentrate here; (c) low priority; and (d) possible

* Tel.: þ886 2 2905 2966; fax: þ886 2 2905 2753. E-mail addresses: [email protected], [email protected]. 0261-5177/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tourman.2013.09.003

overkill. IPA is an appealing tool because it is simple and easy to use, allowing the managerial implications of IPA to be intuitively interpreted (Arbore & Busacca, 2011). Thus, IPA has been applied in numerous industries such as tourism and hospitality (Chang, Chen, & Hsu 2012; Deng, 2007), health care (Yavas & Shemwell, 2001), education (O’Neill & Palmer, 2004), and banking (Matzler, Sauerwein, & Heischmidt, 2003). Despite its widespread use, the specific limitations of IPA have been criticized in extant literature. For example, various methods of calculating importance or performance may lead to different interpretations and subsequent means of correcting perceived problems (Garver, 2003; Oh, 2001). In addition, a slight difference in the position of an attribute could cause its inferred priority to change dramatically (Bacon, 2003). Another critical problem of IPA is that ignoring nonlinear and asymmetric relations between attribute performance (AP) and customer satisfaction (CS), and erroneously assuming that the relationship between AP and importance is independent, could cause the improper commitment

120

L.-F. Chen / Tourism Management 41 (2014) 119e128

Quadrant II

Importance

High

Low

Quadrant I

Concentrate here

Keep up the good work

Quadrant III

Quadrant IV

Low priority

Possible overkill

Low

High Performance

Fig. 1. Traditional IPA matrix.

of scarce resources to misguided improvement efforts (Bacon, 2003; Mikuli c & Prebe zac, 2008; Oh, 2001). Since its introduction in the 1980s, the Kano model has become a popular model for evaluating quality attributes, and has been applied numerous industries. The Kano model facilitates exploring the nonlinear and asymmetric relations between AP and CS, classifying quality attributes into the following categories: (a) must-be; (b) one-dimensional; (c) attractive; and (d) indifferent (Kano, Seraku, Takahashi, & Tsuji, 1984). The performance level of different quality attributes results in varying effects on the perception of CS and customer dissatisfaction (CD). When the CS is proportional to the level of performance, it is considered a onedimensional factor. The increasing level of performance of a must-be factor does not increase the CS, but any decrease in this factor causes CD. Conversely, an increase in the level of performance of an attractive attribute enhances CS, but a low level of performance does not specifically cause CD. Regardless of the level of performance of an attribute, if it results in neither CS nor CD, an indifferent factor is attained (Chen, 2012). To avoid misinterpretations when using IPA, it is crucial to consider the Kano’s quality categories (Arbore & Busacca, 2011; Mikuli c & Prebe zac, 2008; Tontini & Picolo, 2010). For example, when customers rate a must-be factor as highly important, then its corresponding improvement strategy is either “keep up the good work” or “concentrate here.” However, managers should consider the possibility that further improvement might be unnecessary if an increase of this attribute would not create a significant improvement in CS. By contrast, when customers rate an attractive factor as unimportant, then its corresponding improvement strategy could be “low priority” or “possible overkill.” However, because an attractive factor can generate substantial customer delight, enlarging differentiation, a company can lose competitive opportunities by overlooking that item. Nevertheless, the Kano model possesses certain deficiencies that must be addressed. For example, it cannot identify relative importance of attributes in the same category, e.g., onedimensional attributes (Bi, 2012). Therefore, quantitative measures must be developed to evaluate the asymmetric impacts on CS/CD. Furthermore, without emphasizing the current performance levels of product and service attributes, the Kano model is limited in identifying improvement opportunities (Tontini & Silveira, 2007). Despite the debate in the extant literature regarding IPA and the Kano model, scant studies have attempted to address these problems by integrating both models (Tontini & Silveira, 2007).

The purpose of this study is to develop a qualityeperformance analysis (QPA) method that provides a customer-driven framework for identifying strategic service positions and providing quality improvement guidelines. The proposed QPA approach integrates the advantages of the Kano model and IPA, allowing managers to plan service activities. In addition, a signal-to-noise ratio (SNR) approach is designed to measure how AP asymmetrically affects CS and CD. This approach can be used to classify the Kano quality categories and define priorities for improvement. By using a case from the food and beverage industry, we show the effectiveness of the proposed QPA approach, comparing between the proposed QPA and the traditional IPA. Finally, specific methods are selected to compare the power of the SNR approach for classifying the Kano’s quality categories. 2. Literature review 2.1. Importanceeperformance analysis (IPA) The IPA allows companies to identify improvement priorities for various service attributes and elucidating market competition. Typically, the performance aspect of IPA can be measured using CS surveys in which customers rate the level of performance (i.e., satisfaction) of products and services. This is an absolute measure of performance. Relative performance measures, such as gap analyses (Tontini & Picolo, 2010), performance ratios, and comparative scales are also suitable for use in IPA (Garver, 2003). IPA studies have described two types of importance measures: (a) stated importance; and (b) derived importance, both of which demonstrate advantages and limitations. Stated importance can be obtained by asking customers to rate the importance attributes by using Likert-scale ratings (typically ranging from not important to very important). Although this method is commonly applied, it substantially increases survey length, causing poor response rates. In addition, customers may rate all the attributes as important, potentially yielding a low power of discrimination (Garver, 2003). To assess derived importance, customers rate the AP and overall CS for the service being evaluated. The data are subsequently employed to derive the importance of attributes by applying several statistical methods, such as conjoint analysis, correlation analysis, multiple regression, normalized pairwise estimation, partial least squares, and principal components regression (Gustafsson & Johnson, 2004). Using statistically-derived methods for evaluating importance can substantially decrease the survey length and respondent bias. However, because the multicollinearity among service attributes is typically extremely high, any derived importance would be naive, inadequate, uninterpretable, and invalid (Bi, 2012). Therefore, it is critical to select an appropriate method for measuring importance in IPA. Determining the positions of gridlines is another critical factor for IPA. The majority previous IPA studies have reported using scale means (scale-centered approach) or grand means (data-centered approach) to divide the collected data into high- and low-score groups for importance and performance measures (Mikulic & Prebe zac, 2008). In the scale means method, for example, a “3” on a five-point scale would be used to classify the high- and lowperformance groups. However, Peterson and Wilson (1992) showed that “virtually all self-reports of CS possess a distribution in which a majority of the responses indicate that customers are satisfied and the distribution itself is negatively skewed” (p. 62). This characteristic of satisfaction data majorly limits the scale means method when most of the attributes fall into the highperformance group. Compared with the scale means method, the grand means method is more suitable for grouping data. However, the grand means method also exhibits limitations; for example, if

L.-F. Chen / Tourism Management 41 (2014) 119e128

the range of the importance scores of attributes is between 4.4 and 4.8 on a five-point scale, it may be inappropriate to regard attributes that demonstrate low importance scores (e.g., 4.5) as low priority or possible overkill. Previous studies have identified another major shortcoming of IPA; it assumes (a) the relation between AP and importance is independent; and (b) the relation between AP and overall CS is linear. However, growing evidence suggests that the importance of an attribute can change according to its current level of performance (Bi, 2012; Kano et al., 1984), and the relation between AP and overall CS is nonlinear (Lin, Yang, Chan, & Sheu, 2010; Ting & Chen, 2002; Tontini & Silveira, 2007) and asymmetric (Matzler, Bailom, Hinterhuber, Renzl, & Pichler, 2004; Ting & Chen, 2002). Therefore, the managerial improvement directions derived from IPA are potentially misleading. 2.2. The Kano model and classification methods for quality attributes The Kano model demonstrates that an asymmetrical, nonlinear relationship between CS and AP level for various attributes can result in diverse perceptions of CS and CD (Kano et al., 1984). Kano et al. (1984) designed a functional/dysfunctional questionnaire to identify quality attributes. The respondents expressed how they would feel if a specific attribute were present or fulfilled, selecting one of the following answers: 1 ¼ satisfied; 2 ¼ it should be that way; 3 ¼ I am indifferent; 4 ¼ I can live with it; and 5 ¼ dissatisfied. Next, they expressed how they would feel if that attribute were unfulfilled or absent, again selecting one of the five responses. According to Kano’s evaluation table, combining the two responses enables the classification of an attribute. In addition to the functional/dysfunctional questionnaire developed by Kano, various approaches have been recommended for classifying quality attributes. Brandt (1988) proposed a penaltyereward contrast analysis (PRCA) approach to identify quality attributes by employing the following regression equation and dummy variables:

CSi ¼ aj þ b1j D1ij þ b2j D2ij

(1)

where CSi represents the overall satisfaction of the ith customer. The overall satisfaction (CSi) and the performance level of the jth attribute rated by the ith customer (Xij) are measured using a scale rating (e.g., five-point Likert-scale; 1 ¼ extremely low to 5 ¼ extremely high). Two dummy variables are introduced to estimate the effects of AP on CS/CD. D1ij is introduced to estimate the effects of low AP on CD; it is set to 1 if the jth attribute rated by the ith customer is low (e.g., Xij < 3), otherwise it is set to 0. Conversely, D2ij is introduced to estimate the effects of high AP on CS; it is set to 1 if the jth attribute rated by the ith customer is high (Xij > 3) and is otherwise set to 0. The constant aj is the average CS of all reference groups. Penalty (b1j) is expressed as an incremental decrease associated with low AP, whereas reward (b2j) is expressed as an incremental increase associated with high AP. If b1j is nonsignificant and b2j is significantly positive, then the jth attribute is classified as an attractive factor. Next, if b1j is significantly negative and b2j is non-significant, then it is classified as a must-be factor. Finally, if b1j is significantly negative and b2j is significantly positive, then it is classified as a one-dimensional factor (Busacca & Padula, 2005; Chen, 2012; Matzler & Sauerwein, 2002). After the Kano questionnaire, PRCA is the most commonly employed empirical assessment method in relevant studies of the Kano model (Chen, 2012). Various studies have adopted and revised the penalty and reward factors to classify the Kano’s attributes. For example, Füller and Matzler (2008) defined a reward to penalty

121

(RP) ratio as b1j/b2j. A ratio below 1 indicates a must-be factor, a ratio greater than 1 implies an attractive factor, and a ratio of 1 or close to 1 is classified as a one-dimensional factor. Mikuli c and Prebe zac (2008) defined an impact asymmetry index (IAI) as (b2j  b1j)/(b2j þ b1j). A factor is considered attractive (delighter or satisfier), one-dimensional (hybrid), or must-be (dissatisfier or frustrator) if an IAI value is greater than 0.1, between 0.1 and 0.1, or below 0.1, respectively. Tontini and Picolo (2010) employed improvement gap analysis (IGA) to classify the Kano’s categories. Both CS data and the functional/dysfunctional questionnaire are required to perform IGA. The improvement gap (IG) represents the difference between the expected average CS for the functional question and the average current satisfaction for each attribute. Furthermore, the standardized IG and expected average dissatisfaction with the dysfunctional question matrix were applied to identify categories of the Kano model. Among the discussed methods, the functional/dysfunctional questionnaire developed by Kano remains the most popular (Löfgren & Witell, 2008). Mikuli c and Prebe zac (2011) indicated that Kano’s questionnaire provided the most valid and reliable classification method for assessing the Kano model. However, because studies have determined this method to be excessively complex and time consuming, it is impractical for real-world application (Mikulic & Prebe zac, 2011; Witell & Löfgren, 2007). Although specific methods have been proposed to classify the Kano’s categories, validity tests remain lacking that verify the applicability of these methods with the Kano’s questionnaire. Because the Kano model can identify the nonlinear and asymmetric effects of AP on CS, scholars have increasingly focused on integrating the Kano model and IPA (Mikuli c & Prebe zac, 2008; Tontini & Silveira, 2007). However, this additional information is not acquired directly from IPA, but by performing additional matrix analysis. 3. Methodology In this study, we propose a QPA approach that provides a simple but effective framework for collecting CS data, classifying quality attributes, and identifying current service positions, and for developing individual service strategies for each service attribute. Following this approach, the attribute importance of traditional IPA is replaced by evaluating Kano’s quality attributes. In addition, to reduce the effort required for data collection using the functional/ dysfunctional questionnaire developed by Kano, we propose an SNR method to classify Kano’s quality categories. The SNR measure can also be used to define improvement priorities. Furthermore, instead of applying the traditional IPA midpoint method to split attributes into high- and low-performance groups, a three-zone method was used to evaluate AP. The proposed QPA approach involves the following four stages: (1) AP analysis; (2) the Kano’s quality attributes classification; (3) strategic positioning; and (4) quality improvement planning. The concepts of these four stages are detailed as follows. 3.1. Stage 1. Attribute performance (AP) analysis 3.1.1. Collecting satisfaction data and evaluating the performance level of each attribute After consulting with domain experts, the attributes of interest are defined and the satisfaction data for the overall and individual attribute levels are then collected. Let Yi and Xij represent the overall CS level of the ith customer and the CS level of the jth attribute rated by the ith customer, respectively. These satisfaction data are measured by a scale rating such as a 5-point Likert scale. The average CS score of attribute X j is employed to represent its AP

L.-F. Chen / Tourism Management 41 (2014) 119e128

level. To conduct further analyses, certain events are defined. Without loss of generality, let S be a satisfaction event with a value of Yi ¼ {4, 5} and S0 be a dissatisfaction event with a value of Yi ¼ {1, 2, 3}. Similarly, let F be a fulfillment event when the value of Xij ¼ {4, 5}, and F0 be a non-fulfillment event when the value of Xij ¼ {1, 2, 3}. 3.1.2. Defining the zones of high performance, tolerable performance, and low performance CS data are frequently inflated and skewed; thus, rather than applying the midpoint method to allocate attributes into high- and low-performance groups, we classifies the attributes into the following three regions: (1) high-performance; (2) tolerableperformance; and (3) low-performance. In addition, this study differentiates among them by considering the overall CD rate P(S0 ) and overall CS rate P(S) as follows. First, because there is a probability of P(S) that customers are satisfied, we define the highperformance threshold HAP by solving

P X j > HAP




¼ PðSÞ

(2)

In other words, a significant high-performance attribute should demonstrate a performance value greater than the P(S) percentile of the performance values for all attributes. By contrast, because there is a probability of P(S0 ) that customers are not satisfied, we define the significant low-performance threshold LAP by solving

P X j < LAP




¼ PðS0 Þ

(3)

It is reasonable to require that a significant low-performance attribute possess a performance value less than the P(S0 ) percentile for the performance values of all attributes. Finally, any attributes that do not meet these criteria are assigned to the tolerable performance group.

Impact on customer dissatisfaction

122

Quadrant II Must-be factors (Ensure fulfillment)

Quadrant III Indifferent factors (Withdraw or increase efficiency)

3.2.1. Identifying how attribute affects CD by using the proposed SNR measures To identify key dissatisfiers that could cause CD, we calculate Pj(F0 jS0 ) and Pj(F0 jS) to reflect the signal-to-CD and the noise-to-CD of the jth attribute, respectively. Pj(F0 jS0 ) is the conditional probability of failure events among dissatisfied customers, the non-fulfillment of which can influence the overall CD. Conversely, Pj(F0 jS) is the conditional probability of failure events among satisfied customers, the failure of which cannot affect the overall CD. Next, to measure how the jth attribute affected CD, we introduced an SNR for CD, SNRj(S0 ), which is expressed as:

(4)

SNRj(S0 )

Obviously, a high value implies that the non-fulfillment of the jth attribute could substantially affect the overall CD. 3.2.2. Identifying how each attribute affects CS by using the proposed SNR measures When we identify critical satisfiers that could influence CS, we calculate Pj(FjS) and Pj(FjS0 ) to represent the signal-to-CS and the noise-to-CS of the jth attribute, respectively. Pj(FjS) is the conditional probability of fulfilled events among satisfied customers, the fulfillment of which can influence overall CS. Conversely, Pj(FjS0 ) is the conditional probability of fulfilled events among dissatisfied customers, the fulfillment of which cannot affect the overall CS. Next, to measure how the jth attribute affects CS, we introduce SNRj(S) as the SNR of CS, which is expressed as:

One-dimensional factors (Exploit satisfication)

Quadrant IV Attractive factors (Enlarge differentiation)

Low High

Impact on customer satisfaction Fig. 2. Kano’s quality attribute positioning matrix.

SNRj ðSÞ ¼ Pj ðFjSÞ=Pj ðFjS0 Þ

(5)

A high SNRj(S) value implies that the fulfillment of the jth attribute greatly affects the overall CS. 3.2.3. Defining the thresholds to differentiate significant and nonsignificant effect areas of CD and CS Because CS and CD are opposite features of the same concept, we simultaneously consider their thresholds as follows. First, let TS0 and TS be the thresholds for determining the significance of the effect of an attribute on CD and CS, respectively. Because the false alarm rate of dissatisfiers (a), where the true dissatisfiers are incorrectly classified as satisfiers, should not exceed the overall noise-to-CD (i.e., PðF 0 jSÞ), the threshold for determining the significance of the impact on CD (TS0 ) can be obtained by solving the following equation:



a ¼ P SNRj ðS0 Þ < TS0 S0  PðF 0 jSÞ

3.2. Stage 2. The Kano’s quality attributes classification

Quadrant I

Low



SNRj ðS0 Þ ¼ Pj ðF 0 jS0 Þ=Pj ðF 0 jSÞ

High

(6)

An attribute possessing an SNRj ðS0 Þ > TS0 is considered significantly affect CD. Otherwise, the effect on CD is considered nonsignificant. Conversely, the power of the test to satisfiers (1  b), where the true satisfiers are correctly detected, should be greater compared with the overall signal-to-CS (i.e., PðFjSÞ). The threshold for determining the significance of the impact on CS (TS) can be obtained by solving the following equation:

  1  b ¼ P SNRj ðSÞ  TS jS  PðFjSÞ

(7)

As previously stated, an attribute with SNRj(S)  TS is considered to significantly affect CS; otherwise, the effect on CS is considered non-significant. These two thresholds are simultaneously determined because when the equalities hold in Eqs. (6) and (7), it results in a ¼ b. 3.2.4. Identifying the Kano’s quality categories for each attribute The significance of each attribute’s effect on CD and CS is calculated to identify the Kano’s quality categories (see Fig. 2 for the Kano’s quality attribute positioning matrix) as follows: (a) attributes that significantly affect both CS and CD are classified as onedimensional factors, which are strategic items and can be manipulated to exploit CS; (b) those that significantly affect only CS are classified as attractive factors, which are leverage items that can be applied to improve competition differentiation; (c) attributes that significantly affect only CD are classified as must-be factors, which are bottleneck items and an adequate fulfillment should be ensured by managers; and (d) attributes that affect CS and CD

L.-F. Chen / Tourism Management 41 (2014) 119e128

Attribute Performance

High performance

Redundant items

Tolerable performance

Low performance

Major strength

Opportunity

Noncritical items

Minor weakness

Mu s fac t-be tor

On e-d im fac ensio tor nal

Att ra fac ctive tor

Ind iff fac erent to r

Quality categories

Minor strength

Major strength

Risk

Risk

Major weakness

Major weakness

123

because any downgrade in these items could cause substantial t CD. A must-be factor that demonstrates a high performance level can prevent CD; however, it cannot generate CS. Therefore, it is considered a minor strength. Finally, an indifferent factor affects neither CS nor CD. Consequently, the high performance level of this item is unnecessary and the low performance level of this item can be considered noncritical. Shifts in the needs and expectations of customers or the actions of competitors can alter the strategic category of an attribute. Therefore, any qualityeperformance portfolio classification requires regular updates. 3.4. Stage 4. Quality improvement planning According to the locations of the attributes in the qualityeperformance portfolio matrix (Fig. 3), fundamental improvement directions based on the 12 grids (Table 1) are detailed in the following paragraphs.

Fig. 3. Qualityeperformance portfolio matrix.

non-significantly are classified as indifferent factors, which are non-critical; managers should consider withdrawing these items or improving their efficiency. 3.3. Stage 3. Strategic positioning After combining the observations of AP level in Stage 1 and the Kano’s quality categories in Stage 2, each attribute can be placed into 1 of the 12 grids in the qualityeperformance portfolio matrix (Fig. 3). Quality improvement strategies are then developed based on the position of an attribute. First, an attractive or one-dimensional factor that demonstrates a high performance level is a major strength for a company because it can create considerable CS and provide competitive advantages. An attractive factor that demonstrates a tolerable performance level potentially yields competitive advantages. Therefore, it is an opportunity for a company to endeavor for. An attractive factor that demonstrates low performance level is not critically harmful to the competitive position of a company because it has not become a generally expected item for customers; consequently, it presents only a minor weakness. By contrast, a must-be or a one-dimensional factor that demonstrates a low performance level is a major weakness for a company because it can generate considerable CD, causing competitive disadvantages. When must-be or one-dimensional factors demonstrate a tolerable performance level, it generates risk

3.4.1. Attributes that demonstrate low performance levels It is crucial to improve attributes when one-dimensional factors or must-be factors are accompanied by low performance levels because they present major company weaknesses. In addition, they require specific improvement strategies. First, because onedimensional factors significantly affect both CS and CD, they should be the first priority, and aggressive improvement strategies must be implemented immediately. Second, because must-be factors can be regarded as bottleneck items, any subsequent improvement should focus on ensuring adequate fulfillment. However, because it would not produce a significant increase in CS, moderate improvement strategies can be developed to decrease CD. Third, an attractive factor that demonstrates a low performance level is considered a minor weakness. Constructive improvement strategies can be developed when resources are available because fulfilling these items can generate substantial customer delight. Finally, managers need not focus on indifferent factors, even when the performance level is low. 3.4.2. Attributes that demonstrate tolerable performance levels Managers should be vigilant of one-dimensional or must-be factors that are assigned to the tolerable performance group. Although these factors currently might not be crucial problems, any decrease in performance level could easily generate negative effects. In addition, managers should be careful of attractive factor items

Table 1 Segmenting improvement strategy map of qualityeperformance portfolio model. Attribute performance

Quality categories Indifferent factors

Attractive factors

One-dimensional factors

Must-be factors

High performance

Redundant items (S10) A lean strategy is recommended to ensure the efficient allocation of resources.

Major strength (S7) A leverage strategy can be employed to improve competition differentiation.

Major strength (S4) A reasonable aggressive strategy is indicated to exploit CS.

Tolerable performance

Redundant/non-critical items (S11) Be watchful not to allocate unnecessary efforts to improve this item.

Opportunity (S8) Be aware of the opportunity for customer delight if AP can be improved.

Risk (S5) Be vigilant of the risk of potential negative effects due to any downgrade in AP.

Low performance

Non-critical items (S12) Should not pay much attention to this item.

Minor weakness (S9) A constructive improvement strategy can be developed when resources are available.

Major weakness (S6) An aggressive improvement strategy needs to be taken immediately.

Minor strength (S1) An appropriate lean strategy can be applied to ensure an efficient fulfillment. Risk (S2) Be vigilant of the risk of potential negative effects due to any downgrade in AP. Major weakness (S3) A moderate improvement strategy needs to be developed to ensure an adequate fulfillment.

124

L.-F. Chen / Tourism Management 41 (2014) 119e128

Table 2 Statistical results of the proposed SN ratio approach to identify impact level. No.

AP

AP region

Impact on dissatisfaction SNj (S0 )

Impact on satisfaction SNj (S)

Quality category

Strategic positioning

X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 AVE STD

3.748 3.771 3.710 3.895 3.638 3.824 3.838 3.767 3.462 3.262 3.500 3.533 3.981 3.719 3.795 4.057 3.686 3.886 3.695 3.876 3.681 3.524 3.829 3.452 3.324 3.605 3.567 3.224 3.673 0.208

Tolerable High Tolerable High Low High High Tolerable Low Low Low Low High Tolerable High High Tolerable High Tolerable High Tolerable Low High Low Low Low Low Low

3.024* 2.467* 1.658 4.163* 2.417* 4.500* 2.896* 2.182* 1.597 1.701 2.057 2.790* 5.760* 2.618* 5.300* 3.038* 4.275* 2.838* 3.060* 9.600* 6.863* 2.786* 7.200* 2.160* 1.550 2.100 2.213* 1.854 3.381 1.959

1.852 1.579 1.347 1.740 1.984 2.167 1.637 1.619 2.028 5.741* 2.515 4.060* 1.615 2.099 2.955 1.377 3.426 1.781 2.431 2.593 4.722* 5.167* 3.538* 4.938* 2.692 2.544 3.021 6.296* 2.838 1.396

M M I M M M M M I A I O M M M M M M M M O O O O I I M A

(S2) Risk (S1) Minor strength (S1) Redundant/non-critical item (S1) Minor strength (S3) Major weakness (S1) Minor strength (S1) Minor strength (S2) Risk (S12) Non-critical item (S9) Minor weakness (S12) Non-critical item (S6) Major weakness (S1) Minor strength (S2) Risk (S1) Minor strength (S1) Minor strength (S2) Risk (S1) Minor strength (S2) Risk (S1) Minor strength (S5)Risk (S6) Major weakness (S4) Major strength (S6) Major weakness (S12) Non-critical item (S12) Non-critical item (S3) Major weakness (S9) Minor weakness

Notes: 1. M: must-be, O: one-dimensional, A: attractive, I: indifferent. 2. “*” represents a significant impact on CD compared with TS0 ¼ 2:155 in non-fulfilled condition or a significant impact on CS compared with TS ¼ 3.436 in fulfilled condition.

that fall below the tolerable performance level, because it could potentially lose opportunities to generate customer delight. Managers must also be aware of opportunities to generate customer delight when the AP of attractive items can be improved. However, efforts to improve tolerable level performance indifferent factors may be unnecessary, causing the inefficient allocation of resources. 3.4.3. Attributes that demonstrate high performance levels When a one-dimensional or attractive factor demonstrates a high performance level, it generates a major strength for a company. Managers should maintain current objective, applying various strategies. First, regarding one-dimensional factors, a reasonably aggressive strategy is indicated fully leverage CS. Second, an attractive factor presents a niche for creating customer delight, and managers should consider employing leverage strategies to improve competition differentiation. Third, must-be factors provide no additional improvement in CS; thus, an appropriate lean strategy can be applied by considering cost controls and simplifying the service processes to ensure an efficient fulfillment. Finally, an indifferent factor that demonstrates a high performance level is possibly overkill, neither increasing CS nor decreasing CD; thus, a lean strategy is recommended to ensure the efficient allocation of resources.

4. Case study 4.1. The case The company in this case is a famous restaurant chain that manages more than 100 branches in Taiwan and was established almost 20 years ago. This case was also studied by Chen (2012) to explore the relationship between CS and AP in the Kano’s model. To identify critical service factors for improving CS, the company manager selected 10 stores in Taipei to conduct a CS survey based

on DINESERVE (Stevens, Knutson, & Patton, 1995). We investigated the overall CS and 28 attributes related to the following institutional dimensions: (a) tangibles; (b) reliability; (c) responsiveness; (d) assurance; (e) and empathy (see Appendix A for a list of all attributes). We applied a convenience sampling method, surveying customers that dined at the case restaurant in 2010 and obtaining 210 usable questionnaires. The CS ratings for the overall CS and each attribute were evaluated using a Likert scale (1 ¼ extremely dissatisfied; 5 ¼ extremely satisfied). The statistical analysis results produced a Cronbach’s a coefficient of 0.936, which is above the benchmark value of 0.70 recommended by Nunnally (1978). Therefore, the CS survey results show good internal consistency and reliability. In addition, to verify the classification power of the proposed SNR approach, the same customers also completed a Kano functional/dysfunctional questionnaire to identify the Kano’s quality categories for each attribute (Cronbach’s a ¼ 0.740). Chen (2012) summarized the distribution of the Kano’s quality categories and the average satisfaction levels for each attribute.

4.2. Implementation The AP can be calculated using the mean CS values for each attribute. We defined the AP segments based on the statistical results in Table 2. The overall CD rate P(S0 ) was 35.7% (75/210), and the overall CS rate P(S) was 64.3% (135/210). Therefore, the threshold for the significantly low-performance level was in the 35.7th percentile of the AP values for these 28 attributes; the highperformance significance level was the 64.3th percentile (i.e., LAP ¼ 3.643, HAP ¼ 3.768). Consequently, attributes that demonstrated an AP  3.643 were categorized as low-performance items and those that demonstrated an AP  3.768 were highperformance; the remaining attributes (i.e., 3.643 < AP < 3.768) were categorized as tolerable performance items.

L.-F. Chen / Tourism Management 41 (2014) 119e128

Next, we applied the proposed SNR method to classify the Kano’s quality category of each attribute by calculating the signal, noise, and SNR values of each attribute for CS and CD. Based on the statistical results shown in Table 2, we defined the following thresholds to evaluate the significance of the effect of each attribute on CS and CD. The overall noise-to-CD (i.e., PðF 0 jSÞ) and overall signal-to-CS (PðFjSÞ) were 0.256 and 0.744, respectively. Therefore, we set the thresholds of significance for CD and CS as TS0 ¼ 2.155 and TS ¼ 3.436, respectively (based on the 25.6th percentile of SNR(S0 ) and 74.4th percentile of SNR(S)). An SNR value that is greater than its corresponding threshold indicates significant effect. For example, in attribute X6, because SNR6(S0 ) ¼ 4.500 (i.e., > TS0 ), X6 significantly affects CD when it is not fulfilled. However, because SNR6(S) ¼ 2.167 (i.e., X13 > X15 > X6 > X4 > X16 > X7 > X18 > X2. Second, we identified seven weakness attributes (X5, X10, X12, X22, X24, X27, and X28) that significantly affected CD and/or CS; however, their AP levels were relatively low. Among these seven items, attributes X5, X12, X22, X24, and X27 were considered major weakness for the case company that require urgent improvement. While aggressive improvement strategies should be adopted immediately for attributes X12, X22, and X24 (X22 > X12 > X24), moderate improvement strategies can be employed to ensure an adequate fulfillment to decrease CD for attributes X5 and X27 (X5 > X27). Attributes X10 and X28 are considered minor weakness; because fulfilling these items can increase customer delight, constructive improvement strategies can be developed when resources become available. Because of the levels of effect on CS (i.e., SNR(S)), the priority should be X28 > X10. Third, six risk attributes (X1, X8, X14, X17, X19, and X21) that demonstrated tolerable performance levels require close scrutiny. Any performance downgrade in these items could negatively affect the CS or CD. However, if unnecessary

125

resources are allocated to attribute X3 (indifferent factor), it could cause waste. Finally, attributes X9, X11, X25, and X26 do not require the close attention of management. Although their AP levels were low, these attributes affect neither CS nor CD. 5. Comparison and discussion 5.1. Comparison with traditional IPA To compare the proposed QPA approach with traditional IPA, we calculated the standardized regression coefficients by using a multivariate linear regression (MLRs) of the AP ratings over the overall CS rating, measuring the importance of service attributes. We selected MLR as the derived importance measure for comparison because it has been one of the most popular indirect importance measures in the extant IPA literature. Table 4 lists the statistical results. The adjusted R2 was 0.744 and the significance value (p) was 0.000 for the regression model. These approaches yielded various strategy suggestions for 64% (18/28) of the attributes. Regarding the remaining 36% (10/28) of attributes that receive similar strategies recommended by the proposed QPA and traditional IPA; however, the proposed QPA provided the most precise directions for improvement. For example, attributes X2 and X6 were the items that demonstrated high AP levels, for which the traditional IPA model recommended a “keep up the good work” strategy. However, these attributes were must-be factors; thus, it is suggested that managers consider cost control while ensuring adequate fulfillment. Traditional IPA does not encompass the effects of the Kano’s categories, which fail to identify accurate improvement suggestions. In addition, traditional IPA divides AP into two groups by comparing the average AP of all attributes. This approach is inappropriate, particularly for attributes that score extremely close to the threshold. For example, the AP of attributes X14, X17, and X19 in the “keep up the good work” quadrant of traditional IPA (3.719, 3.686, and 3.695, respectively) are proximal to the threshold value (3.673). Because these attributes are must-be factors, managers should be vigilant of any negative effects on CD caused by downgrades in AP. The “keep up the good work” recommendation for these attributes could easily mislead managers. Conversely, the proposed QPA method considers these four attributes to be risk items, warning managers to prevent possible dissatisfaction. Furthermore, when using traditional IPA, certain negative effects that fall below the average importance rating could more significantly affect the overall CS compared with positive effects that are greater than the average importance rating. However, these negative effects are regarded as low-importance and are consequently neglected. For example, attributes X7, X13, X15, X18, X20 and X23 are regarded as “possible overkill” items because they demonstrate a negative importance rating, and are classified as low-importance in the traditional IPA. However, in the proposed method, these factors are must-be or one-dimensional factors that require specific improvement strategies. Consequently, the IPA

Table 3 Segmenting improvement strategies of qualityeperformance portfolio model for the case company. Attribute performance

Kano’s categories Indifferent factors

Attractive factors

One-dimensional factors

Must-be factors

High performance

Redundant items (S10) Redundant/non-critical items (S11) X3 Non-critical items (S12) X9, X11, X25, X26

Major strength (S7) Opportunity (S8) Minor weakness (S9) X28 > X10,

Major strength (S4) X23 Risk (S5) X21 Major weakness (S6) X22 > X12 > X24

Minor strength (S1) X20 > X13 > X15 > X6 > X4 > X16 > X7 > X18 > X2 Risk (S2) X17 > X19 > X1 > X14 > X8 Major weakness (S3) X5 > X27

Tolerable performance Low performance

Notes: 1. Xm > Xn means that the attribute Xm has higher improvement priority than Xn. 2. We don’t have to rank the priority for non-critical attributes.

126

L.-F. Chen / Tourism Management 41 (2014) 119e128

Table 4 The comparison of the proposed QPA approach and traditional IPA. No.

X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 Ave

Traditional IPA

The proposed QPA

Standardized regression coefficients

AP score

Importance level

AP level

Decision

0.034 0.054 0.059 0.040 0.052 0.196*** 0.016 0.012 0.027 0.089* 0.029 0.137** 0.033 0.056 0.082 0.029 0.184** 0.030 0.112** 0.029 0.086 0.018 0.042 0.117** 0.011 0.046 0.169** 0.073 0.048

3.748 3.771 3.710 3.895 3.638 3.824 3.838 3.767 3.462 3.262 3.500 3.533 3.981 3.719 3.795 4.057 3.686 3.886 3.695 3.876 3.681 3.524 3.829 3.452 3.324 3.605 3.567 3.224 3.673

Low High Low Low High High Low Low Low High Low High Low High Low Low High Low High Low High Low Low High Low Low High High

High High High High Low High High High Low Low Low Low High High High High High High High High High Low High Low Low Low Low Low

Possible overkill Keep up the good Possible overkill Possible overkill Concentrate here Keep up the good Possible overkill Possible overkill Low priority Concentrate here Low priority Concentrate here Possible overkill Keep up the good Possible overkill Possible overkill Keep up the good Possible overkill Keep up the good Possible overkill Keep up the good Low priority Possible overkill Concentrate here Low priority Low priority Concentrate here Concentrate here

work

work

work

work work work

Quality categories

AP level

Decision

M M I M M M M M I A I O M M M M M M M M O O O O I I M A

Tolerable High Tolerable High Low High High Tolerable Low Low Low Low High Tolerable High High Tolerable High Tolerable High Tolerable Low High Low Low Low Low Low

(S2) Risk (S1) Minor strength (S11) Redundant/non-critical item (S1) Minor strength (S3) Major weakness (S1) Minor strength (S1) Minor strength (S2) Risk (S12) Non-critical item (S9) Minor weakness (S12) Non-critical item (S6) Major weakness (S1) Minor strength (S2) Risk (S1) Minor strength (S1) Minor strength (S2) Risk (S1) Minor strength (S2) Risk (S1) Minor strength (S5)Risk (S6) Major weakness (S4) Major strength (S6) Major weakness (S12) Non-critical item (S12) Non-critical item (S3) Major weakness (S9) Minor weakness

Notes: 1. *P < 0.1; **P < 0.05; ***P < 0.01. 2. M: must-be, O: one-dimensional, A: attractive, I: indifferent.

method could facilitate inappropriate decisions. Therefore, the proposed QPA approach is more effective compared with the traditional IPA approach, regarding resource allocation suggestions. Finally, only seven standardized regression coefficients demonstrated statistical significant (i.e., >a ¼ 0.1), demonstrating another critical limitation of the traditional IPA approach. It is not practical to ignore the remaining 21 non-significant attributes; furthermore it is theoretically deficient to perform further analyses. However, this issue has seldom been discussed in the extant IPA literature. 5.2. Comparing the classification power of the Kano’s quality categories To verify the classification power of the proposed SNR approach to identify the Kano’s quality categories, the PRCA (Brandt, 1988), IAI (Mikuli c & Prebe zac, 2008), RP ratio (Füller & Matzler, 2008) ,and IGA (Tontini & Picolo, 2010) methods were used for comparison. Subsequently, the results of the Kano’s functional/dysfunctional questionnaire were used as a baseline for comparison; this classification method is highly reliable for assessing the Kano model (Mikuli c& Prebe zac, 2011) and is the most commonly applied instrument for classifying the Kano’s quality categories (Löfgren & Witwell, 2008). Table 5 shows the classification results for these approaches. Table 5 shows that the classification results obtained using the PRCA, IAI, RP ratio, and IGA methods were substantially different compared with those obtained using the Kano’s questionnaire. The overall matching rates of these approaches were less than 8%, demonstrating an extremely poor performance level for the discussed case. While the RP ratio approach tended to classify attributes as attractive factors (26 of 28), the PRCA method classified them as one-dimensional (14 of 28) and attractive factors (13 of 28). The IGA approach involves a focus on collecting Kano functional/ dysfunctional data; however, the results obtained using the IGA

method were incompatible with those of the Kano’s questionnaire based on the Kano’s evaluation table. The overall matching rate of the proposed SNR approach was 60.71% (17 of 28), which significantly outperformed the other methods. By using conventional CS survey data, the proposed approach simplifies the data collection process compared with compiling a list of functional and dysfunctional questions. Therefore, the proposed approach is practical to implement and maintains a classification power comparable to that of the Kano’s questionnaire. Consequently, the proposed approach successfully integrates the advantages of the Kano model, which encompasses the asymmetric effects of AP on CS and CD, into IPA analysis, which identifies correct improvement opportunities for strategic planning. 6. Conclusion Organizations must understand customer needs to elucidate their market competition, identify improvement opportunities, and conduct strategic planning. However, the application of traditional IPA for positioning improvement strategies has theoretical limitations that fail to provide acceptable analyses. We propose a customer-driven framework that integrates the advantages of traditional IPA and the Kano model to assist in planning service strategies. The restaurant case study indicates the effectiveness of the proposed approach. Certain aspects of the proposed method can contribute to the field of research. First, a QPA approach was proposed, identifying effective service improvement directions and providing strategic guidelines to assist managers in designing service activities. Next, AP was divided into three zones instead of using the midpoint method applied in traditional IPA; this facilitated the generation of precise improvement strategies. Third, the proposed SNR approach proved suitable for classifying the Kano’s quality categories. The statistical results of the case show strong

Table 5 Statistical results of different approaches in identifying Kano’s quality categories. Attribute no.

PRCA method F-statistics

Penalty indices b1j

47.72** 0.178 0.000 15.02** 0.118 0.083 10.39** 0.082 0.096 26.82** 0.198 0.054 24.16** 0.181 0.159* 38.98** 0.267 0.102 26.61** 0.197 0.136* 26.30** 0.195 0.243** 17.49** 0.136 0.062 47.92** 0.310 0.214** 30.71** 0.221 0.107 57.03** 0.349 0.268** 35.99** 0.251 0.121 26.07** 0.194 0.100 50.10** 0.320 0.088 15.20** 0.120 0.128 53.31** 0.333 0.171** 24.82** 0.186 0.144* 34.79** 0.244 0.154* 56.45** 0.347 0.090 72.25** 0.405 0.142* 47.38** 0.307 0.168** 58.83** 0.356 0.088 53.37** 0.334 0.256** 32.75** 0.233 0.256** 37.85** 0.261 0.158** 51.26** 0.325 0.202** 51.85** 0.327 0.190** with the Kano’s questionnaire

IAI method

RP ratio method

IGA method

Reward indices b2j

Classification results

IAI

Reward to penalty ratio

Classification results

Standardized IG

Standardized AESDQ

Classification results

0.432** 0.362** 0.263** 0.435** 0.349** 0.478** 0.386** 0.297** 0.359** 0.472** 0.435** 0.481** 0.457** 0.410** 0.056** 0.310** 0.513** 0.401** 0.432** 0.555** 0.053** 0.494** 0.567** 0.471** 0.356** 0.460** 0.496** 0.493**

A A A A O A O O A O A O A A A A O O O A O A A O O O O O 3.57% (1/28)

1.000 A 1.595 A 0.465 A 0.779 A 0.374 A 0.648 A 0.479 A 0.100 O 0.705 A 0.376 A 0.605 A 0.284 A 0.581 A 0.608 A 0.222 M 0.416 A 0.500 A 0.472 A 0.474 A 0.721 A 0.456 M 0.492 A 0.731 A 0.296 A 0.163 A 0.489 A 0.421 A 0.444 A 7.14% (2/28)

Infinity 4.361 2.740 8.056 2.195 4.686 2.838 1.222 5.790 2.206 4.065 1.795 3.777 4.100 0.636 2.422 3.000 2.785 2.805 6.167 0.373 2.940 6.443 1.840 1.391 2.911 2.455 2.595 7.14% (2/28)

A A A A A A A A A A A A A A M A A A A A M A A A A A A A

0.147 0.009 0.475 0.233 0.768 0.112 0.078 0.164 1.355 2.529 0.682 0.803 1.407 0.199 0.319 0.578 0.095 1.511 0.233 1.994 0.371 0.285 0.803 0.388 0.785 0.682 0.544 2.356 0.00% (0/28)

0.076 0.169 0.278 1.037 1.192 1.316 0.665 1.148 1.520 1.675 0.853 0.975 0.960 1.006 0.014 0.045 0.681 0.278 0.665 0.107 0.357 0.138 1.161 1.845 2.263 0.481 0.915 0.293

A A I I A A A M Critical item Critical item M A I I A I I I I I M A A Critical item M M M M

Classification results

The proposed SNR method

The Kano’s questionnaire

M M I M M M M M I A I O M M M M M M M M O O O O I I M A 60.71% (17/28)

X(M, I) X(M, I) M O M M M I I I I M M M X(M, I) X(A, O) M M M M I X(M, I) O I I I I X(M, I)

L.-F. Chen / Tourism Management 41 (2014) 119e128

Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10 Q11 Q12 Q13 Q14 Q15 Q16 Q17 Q18 Q19 Q20 Q21 Q22 Q23 Q24 Q25 Q26 Q27 Q28 Matching rate

Adjusted R2

Note: *P < 0.05; **P < 0.01.

127

128

L.-F. Chen / Tourism Management 41 (2014) 119e128

classification power that is similar to that of the functional/ dysfunctional questionnaire developed by Kano, significantly outperforming other classification methods. Fourth, the SNR measures provide the relative importance of the Kano’s quality categories, allowing managers to define improvement priorities accordingly. Furthermore, the proposed QPA approach is easy to implement using the data obtained from a typical CS survey. We recommend that more empirical testing be applied to additional industries to confirm the validity of the proposed approach.

Acknowledgment This work was partially supported by Grants from National Science Council, Taiwan, R.O.C.

Appendix A

Table A1. Restaurant service attributes. Items

Service attribute descriptions

X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17

Has staff members who are clean, neat and appropriately dressed. Has a décor in keeping with its image and price range. Has a menu that is easily readable. Has a dining area that is comfortable and easy to move around in. Has restrooms that are thoroughly clean. Has dining areas that are thoroughly clean. Has a visually attractive dining area. Has visually attractive building exteriors. Provides internet service. Provides special discounts. Has visually indication that is clearly. Quickly corrects anything that is wrong. Provides an accurate guest check. Serves your food exactly as you ordered it. Serves you in the time promised. Gives extra effort to handle your special requests. Has employees who can answer your questions completely. Makes you feel comfortable and confident in your dealings with them. Has personnel who are both able and willing to give you information about menu items, their ingredients, and methods of preparation. Makes you feel personally safe. Has personnel who seem well-trained, competent, and experienced. Seems to give employees support so that they can do their job well. Seems to have the customers’ best interests at heart. Makes you feel special. Provides new meals periodically. Has employees who are sensitive to your individual needs and wants, rather than always relying on policies and procedures. Anticipates your individual needs and wants. Has employees who are sympathetic and reassuring if something is wrong.

X18 X19

X20 X21 X22 X23 X24 X25 X26

X27 X28

References Arbore, A., & Busacca, B. (2011). Rejuvenating importance-performance analysis. Journal of Service Management, 22(3), 617e627. Azzopardi, E., & Nash, R. (2013). A critical evaluation of importance performance analysis. Tourism Management, 35, 222e233. Bacon, D. R. (2003). A comparison of approaches to importance-performance analysis. International Journal of Market Research, 45(1), 55e71. Bi, J. (2012). A review of statistical methods for determination of relative importance of correlated predictors and identification of drivers of consumer liking. Journal of Sensory Studies, 27, 87e101. Brandt, R. D. (1988). How service marketers can identify value-enhancing service element. Journal of Services Marketing, 2(3), 35e41.

Busacca, B., & Padula, G. (2005). Understanding the relationship between attribute performance and overall satisfaction: theory, measurement and implications. Marketing Intelligence & Planning, 23(6), 543e561. Chang, K. C., Chen, M. C., & Hsu, C. L. (2012). Identifying critical brand contact elements of a tourist destination: applications of Kano’s model and the importancesatisfaction model. International Journal of Tourism Research, 14(3), 205e221. Chen, L. F. (2012). A novel approach to regression analysis for the classification of quality attributes in the Kano model: an empirical test in the food and beverage industry. Omega-International Journal of Management Science, 40(5), 651e659. Deng, W. (2007). Using a revised importance-performance analysis approach: the case of Taiwanese hot springs tourism. Tourism Management, 28, 1274e1284. Füller, J., & Matzler, K. (2008). Customer delight and market segmentation: an application of the three-factor theory of customer satisfaction on life style groups. Tourism Management, 29, 116e126. Garver, M. S. (2003). Best practices in identifying customer-driven improvement opportunities. Industrial Marketing Management, 32, 455e466. Gustafsson, A., & Johnson, M. D. (2004). Determining attribute importance in a service satisfaction model. Journal of Service Research, 7(2), 124e141. Kano, N., Seraku, N., Takahashi, F., & Tsuji, S. (1984). Attractive quality and must-be quality. The Journal of the Japanese Society for Quality Control, 14, 39e48. Lin, S. P., Yang, C. L., Chan, Y. H., & Sheu, C. (2010). Refining Kano’s quality attributessatisfaction model: a moderated regression approach. International Journal of Production Economics, 126, 255e263. Löfgren, M., & Witell, L. (2008). Two decades of using Kano’s theory of attractive quality: a literature review. Quality Management Journal, 15(1), 59e75. Martilla, J. A., & James, J. C. (1977). Importance-performance analysis. Journal of Marketing, 41, 77e79. Matzler, K., Bailom, F., Hinterhuber, H. H., Renzl, B., & Pichler, J. (2004). The asymmetric relationship between attribute-level performance and overall customer satisfaction: a reconsideration of the importance-performance analysis. Industrial Marketing Management, 33, 271e277. Matzler, K., & Sauerwein, E. (2002). The factor structure of customer satisfaction: an empirical test of the importance grid and the penalty-reward-contrast analysis. International Journal of Service Industry Management, 13(4), 314e332. Matzler, K., Sauerwein, L., & Heischmidt, K. A. (2003). Importance-performance analysis revisited: the role of the factor structure of customer satisfaction. The Service Industries Journal, 23(2), 112e129. Mikuli c, J., & Prebe zac, D. (2008). Prioritizing improvement of service attributes using impact range-performance analysis and impact-asymmetric analysis. Managing Service Quality, 18(6), 559e576. Mikuli c, J., & Prebe zac, D. (2011). A critical review of techniques for classifying quality attributes in the Kano model. Managing Service Quality, 21(1), 46e66. Nunnally, J. C. (1978). Psychometric theory. New York: McGraw-Hill. Oh, H. (2001). Revisiting importance-performance analysis. Tourism Management, 22, 617e627. O’neill, M. A., & Palmer, A. (2004). Importance-performance analysis: a useful tool for directing continuous quality improvement in higher education. Quality Assurance in Education, 12(1), 39e52. Stevens, P., Knutson, B., & Patton, M. (1995). DINESERV: a tool for measuring service quality in restaurants. Cornell Hotel and Restaurant Administration Quarterly, 36(2), 56e60. Ting, S. C., & Chen, C. N. (2002). The asymmetrical and nonlinear effects of store quality attributes on customer satisfaction. Total Quality Management, 13(4), 547e569. Tontini, G., & Silverira, A. (2007). Identification of satisfaction attributes using competitive analysis of the improvement gap. International Journal of Operations and Production Management, 27(5), 482e500. Tontini, G., & Picolo, D. J. (2010). Improvement gap analysis. Managing Service Quality, 20(6), 566e584. Peterson, R. A., & Wilson, W. R. (1992). Measuring customer satisfaction: fact and artifact. Journal of the Academy of Marketing Science, 20(1), 61e71. Witell, L., & Löfgren, M. (2007). Classification of quality attributes. Managing Service Quality, 17(1), 54e73. Yavas, U., & Shemwell, D. J. (2001). Modified importance-performance analysis: an application to hospitals. International Journal of Health Care Quality Assurance, 14(3), 104e110.

Dr. Li-Fei Chen is an Associate Professor of Department of Business Administration at Fu Jen University, New Taipei City, Taiwan. Dr. Chen received her Ph.D. degree in industrial engineering and engineering management from the National Tsing Hua University, Hsinchu, Taiwan. Her current research interests include customer satisfaction, service quality, operation management, and data mining and its applications. The results of her research have been published in numerous academic journals including Omega e The International Journal of Management Science, IEEE Transactions on Semiconductor Manufacturing, IEEE Transactions on Electronics Packaging Manufacturing, Expert Systems with Applications, International Journal of Production Research, Intelligent Data Analysis, Neural Computing and Applications, and Total Quality Management & Business Excellence.