A Two-Stage Fuzzy Quality Function Deployment Model for Service Design

Abstract

Over the past decades, a great interest of quality function deployment (QFD) application has shift from industrial product to service sector. However, most previous studies focus on the first-phase of QFD or multi-phase model in service design, which is incomplete and not systematic. Moreover, the inherent vagueness or impreciseness in QFD presents a special challenge to the effective results. Toward this end, this paper tries to propose a two-phase fuzzy QFD model for service industry based on the theory of strategic, tactical and operational (STO). By this, on one hand, our proposed model can deal with the ill-defined nature of the linguistic judgments. On the other hand, it divides the decision and plan process into three correlated level, strategic, tactical and operational. In particular, this paper addresses detailed implementation procedures of the two-phase model, which can translate the customer needs (CNs) to service characteristics (SCs) to service process (SP) elements.

Appendix: Entropy Method

The application of entropy begins with the construction of comparison matrix, step 1-3 is to obtain and analyze the follow comparison matrix X, similar to the step 1-3, step 1-9 is to obtain SCs comparison matrix Y.

$$ {\text{X = }}\begin{array}{*{20}c} {CN_{1} } \\ {CN_{2} } \\ \vdots \\ {CN_{M} } \\ \end{array} \left( {\begin{array}{*{20}c} {C_{1} } & {C_{2} } & \cdots & {C_{\text{p}} } \\ {x_{11} } & {x_{12} } & \cdots & {x_{1P} } \\ {x_{21} } & {x_{22} } & \cdots & {x_{2P} } \\ \vdots & \vdots & \ddots & \vdots \\ {x_{M1} } & {x_{M2} } & \cdots & {x_{MP} } \\ \end{array} } \right){\text{Y = }}\begin{array}{*{20}c} {SC_{1} } \\ {SC_{2} } \\ \vdots \\ {SC_{N} } \\ \end{array} \left( {\begin{array}{*{20}c} {C_{1} } & {C_{2} } & \cdots & {C_{\text{p}} } \\ {{\text{y}}_{11} } & {{\text{y}}_{12} } & \cdots & {{\text{y}}_{1P} } \\ {{\text{y}}_{21} } & {{\text{y}}_{22} } & \cdots & {{\text{y}}_{2P} } \\ \vdots & \vdots & \ddots & \vdots \\ {{\text{y}}_{N1} } & {{\text{y}}_{N2} } & \cdots & {{\text{y}}_{NP} } \\ \end{array} } \right) $$

Take example of comparison X, where ${\text{x}}_{\text{mp}}$ is the performance of company p’s service on customer need ${\text{CN}}_{\text{m}}$. The managerial mean of applied entropy method: If company C₁ performs much better than many companies in terms of a customer need $CN_{m}$, then further improvement may not be urgently needed and thus a lower priority could be assigned to $CN_{m}$. At the other extreme, if C₁ performs much worse than many other companies on CN_m, then it may be difficult for C1 to build a competitive advantage within a short period of time. In both cases, CNs could be assigned a lower priority rating. However, if most companies perform quite similarly on CNs, not too much improvement effort from C₁ may result in a better performance of its service and give C₁ a unique competitive advantage.

Entropy is a measure for the amount of information (or uncertainty, variations) represented by a discrete probability distribution, $q_{1} ,q_{2} , \ldots ,q_{L}$:

$$ E(q_{1} ,q_{2} , \ldots ,q_{p} ) = - \emptyset_{p} \sum\nolimits_{p = 1}^{p} {q_{p} \ln (q_{p} )} $$

(A.1)

where $\phi_{\text{p}} = {1 \mathord{\left/ {\vphantom {1 {\ln ({\text{P}})}}} \right. \kern-0pt} {\ln ({\text{P}})}}$ is normalization constant to guarantee $0 \le E(q_{1} ,q_{2} , \ldots ,q_{p} ) \le 1$. Larger entropy or $E(q_{1} ,q_{2} , \ldots ,q_{p} )$ value implies smaller variations among the Q_p’s and hence less information contained in the distribution.

Take the computation of CNs competitive priority for an example, for the mth row of customer comparison matrix X corresponding to the customer need $CN_{m}$, x_m1, x_m2,…, x_mp, let $x_{m} = \sum\nolimits_{p = 1}^{p} {x_{mp} }$ be the total score with respect to $CN_{m}$, then according to (B.1), the normalized ratings $q_{mp} = x_{mp} /x_{m}$ for p = 1,2,…,P can be viewed as the “probability distribution” of $CN_{m}$ on the P companied with entropy as

$$ E(q_{1} ,q_{2} , \ldots ,q_{p} ) = - \phi_{p} \sum\nolimits_{p = 1}^{p} {q_{p} \ln (q_{p} )} = - \phi_{p} \sum\nolimits_{p = 1}^{p} {\left( {\frac{{x_{mp} }}{{x_{m} }}} \right)ln\left( {\frac{{x_{mp} }}{{x_{m} }}} \right)} $$

(A.2)

It is clear that larger the E($CN_{m}$) value, the less information contained in CNs or smaller variations among the qmp’s (or x _ml ’s). So E($CN_{m}$) can be used to reflect the relative competitive advantage in terms of customer need CN_m. All these E(CNs) value, after normalizations:

$$ E(q_{1} ,q_{2} , \ldots ,q_{p} ) = - \emptyset_{p} \sum\nolimits_{p = 1}^{p} {q_{p} \ln (q_{p} )} $$

(A.3)

where ${\text{e}}_{\text{m}}$ can be considered as the competitive priority rating for company C₁ on the M customer needs. It is similar to compute the competitive priority of SCs.