Back-propagation neural network based importance–performance analysis for determining critical service attributes

https://doi.org/10.1016/j.eswa.2006.12.016Get rights and content

Abstract

Importance–performance analysis (IPA) is a simple but effective means of assisting practitioners in prioritizing service attributes when attempting to enhance service quality and customer satisfaction. As numerous studies have demonstrated, attribute performance and overall satisfaction have a non-linear relationship, attribute importance and attribute performance have a causal relationship and the customer’s self-stated importance is not the actual importance of service attribute. These findings raise questions regarding the applicability of conventional IPA. Therefore, this study presents a revised IPA which integrates back-propagation neural network and three-factor theory to effectively assist practitioners in determining critical service attributes. Finally, a customer satisfaction improvement case is presented to demonstrate the implementation of the proposed Back-Propagation Neural Network based Importance–Performance Analysis (BPNN-IPA) approach.

Introduction

Delivering superior customer value and satisfaction are crucial to the competitive edge of a firm (Armstrong and Kotler, 2000, Weitz and Jap, 1995). Undoubtedly, service quality and customer satisfaction are principal drivers of financial performance. Matzler, Bailom, Hinterhuber, Renzl, and Pichler (2004) contended that customer satisfaction increases customer loyalty, reduces price sensitivity, increases cross-buying and increases positive word of mouth. Hansemark and Albinsson (2004) also noted that customer satisfaction directly influences customer retention and firm market share. Numerous empirical studies have confirmed the positive correlation between customer satisfaction and profitability (Anderson et al., 1994, Eklof et al., 1999, Hallowell, 1996, Johnson et al., 1996, Zeithaml, 2000). Therefore, improving customer satisfaction is a critical issue for business managers in today’s competitive global marketplace. With this goal in mind, numerous business managers are continually attempting to identify critical service attributes that generate customer satisfaction and loyalty in order to stay abreast of competitors.

Numerous practitioners and researchers have applied Importance–Performance Analysis (IPA) to identify the critical performance attributes in customer satisfaction survey data for products and services (Chu and Choi, 2000, Enright and Newton, 2004, Hawes and Rao, 1985, Huan et al., 2002, O’Neill and Palmer, 2004, Tikkanen et al., 2000, Yavas and Shemwell, 1997, Zhang and Chow, 2004). Hansen and Bush (1999) pointed out that IPA is a simple and effective technique that can assist practitioners in identifying improvement priorities for service attributes and direct quality-based marketing strategies. Practitioners apply IPA to analyze two dimensions of service attributes: performance level (satisfaction); and, importance to customers. Analyses of these dimension attributes are then integrated into a matrix that helps a firm identify primary drivers of customer satisfaction and, based on these findings, set improvement priorities (Matzler, Bailom, et al., 2004). Hence, following a customer satisfaction survey and IPA, business managers can make rational decisions about how to best deploy scarce resources to attain the highest degree of customer satisfaction.

Although IPA is an extremely valuable method, previous studies have several important shortcomings. For example, Matzler, Bailom, et al. (2004) noted the original IPA has two implicit assumptions: (1) attribute performance and attribute importance are independent variables; and, (2) the relationship between attribute performance and overall performance is linear and symmetrical. These assumptions are erroneous in the real world, the relationship between attribute-level performance and overall customer satisfaction is asymmetrical (Kano et al., 1984, Matzler and Sauerwein, 2002, Matzler, Bailom, et al., 2004, Matzler et al., 2003, Ting and Chen, 2002) and the relationship between attribute performance and attribute importance is causal (Matzler, Bailom, et al., 2004, Oh, 2001, Ryan and Huyton, 2002, Sampson and Showalter, 1999).

Berman (2005) noted that customer delight is not same as customer satisfaction. Customer delight requires that customer receive a positive surprise that exceeds their expectations. Berman also mentioned that the must-be, satisfier, and delight attribute categorization system developed by Kano et al. (1984) is a popular approach for better understanding customer delight. However, other studies of customer satisfaction have indicated that satisfaction attributes can be understood using three categories: basic factors, performance factors, and excitement factors (Anderson and Mittal, 2000, Brandt, 1988, Johnston, 1995, Matzler and Hinterhuber, 1998, Matzler and Sauerwein, 2002, Matzler et al., 1996, Oliver, 1997). The impact of satisfaction attribute performance on overall customer satisfaction differs according to category. For example, if delight (excitement) attributes are not met, customers do not feel dissatisfied. However, if delight (excitement) attributes are met, the result is customer delight. Therefore, practitioners of IPA must consider three-factor theory to determine critical service attributes that capable of generating customer satisfaction, delight and loyalty.

The relative importance of attribute is the input variable of IPA method. Generally, IPA practitioners use customers’ self-stated importance. However, numerous studies had revealed that the relationship between attribute performance and attribute importance is causal (Matzler, Bailom, et al., 2004, Oh, 2001, Ryan and Huyton, 2002, Sampson and Showalter, 1999). In other words, when attribute performance changes, so does relative attribute importance (Matzler & Sauerwein, 2002). Consequently, the customers’ self-stated importance does not adequately measure the actual relative importance of attributes. To overcome this problem, practitioners of IPA frequently use statistically inferred importance ratings. Multiple regression and structural equation modeling are conventional statistical methods for obtaining the implicitly derived importance of attributes (Garver, 2003). However, these statistical methods assume that (1) data are relatively normal, (2) the relationships between independent and dependent variables are linear and (3) multicollinearity between independent variables is relatively low (Taylor, 1997). However, in customer satisfaction research, such assumptions are almost always violated (Garver, 2002). Thus, the implicitly derived importance of attributes via statistical methods can be biased and misleading.

Artificial neural networks (ANNs) are analytic techniques modeled on the learning processes of the human cognitive system and the neurological functions of the brain. Numerous studies applied ANNs for prediction and classification research in the sciences and social sciences. Garver (2002) noted that ANNs overcome the limitations in conventional statistical applications. Bishop (1994) has demonstrated, in situations where non-normal data, multicollinearity, and non-linear relationships are present, how neural networks outperform multiple regression models. Burger, Dohnal, Kathrada, and Law (2001) indicated that ANNs generally have high degrees of freedom, and, thus, they can model the non-linearity of a process under study significantly better than regression approaches. Furthermore, Tsaur, Chiu, and Huang (2002) applied ANNs for measuring importance scores for service aspects at nine international hotels and compared these scores with those derived using a logistic regression model. Therefore, IPA practitioners can utilize ANNs to measure the actual importance of attributes prior to determining actual critical service attributes.

This study presents a revised IPA approach that comprises a back-propagation neural network (BPNN) and three-factor theory. The proposed Back-Propagation Neural Network Based Importance–Performance Analysis (BPNN-IPA) avoids two shortcomings common to conventional IPA when dealing with non-normal data—multicollinearity and non-linear circumstances—and uses the actual importance of service attributes, thereby assisting business managers in determining the critical service attributes for improving service quality or customer satisfaction and achieving competitive advantage.

The remainder of this paper is organized as follows. Section 2 reviews the relevant literature particularly that about IPA, three-factor theory of customer satisfaction and BPNN. To elucidate the actual importance of attributes, Section 3 introduces a BPNN-IPA approach. Next, Section 4 demonstrates the implementation of the proposed BPNN-IPA approach to determine critical service attributes and enhance customer satisfaction at a Taiwanese hot spring hotel. Finally, Section 5 draws conclusions.

Section snippets

Importance–performance analysis

Importance–performance analysis has been applied as an effective means of evaluating a firm’s competitive position in the market, identifying improvement opportunities, and guiding strategic planning efforts (Hawes and Rao, 1985, Martilla and James, 1977, Myers, 1999). Importance–performance analysis, first introduced by Martilla and James (1977), identifies which product or service attributes a firm should focus on to enhance customer satisfaction (Matzler, Bailom, et al., 2004). Typically,

The method for designing a BPNN model

A BPNN model is one of the most widely used ANN models. Therefore, general commercial ANN software packages (e.g., NeuroShell 2; NeuroSolutions 5; NeuFrame; etc.) can be utilized by practitioners when building a BPNN model. The BPNN architecture comprises one input layer, hidden layers and one output layer. The BPNN parameters include a number of hidden layers, a number of hidden neurons, an activation function, learning rate, momentum, etc. All of these parameters have significant impact on

Case study

In this section, an example case is presented to demonstrate the application of the BPNN-IPA for service quality or customer satisfaction improvement stage. The example case is a case study of customer satisfaction improvement for hot spring hotel in Taiwan. For the purpose of demonstration, a standardized questionnaire using closed-response questions and a five-point Likert scale (1 = completely unsatisfied to 5 = completely satisfied) was developed based on a review of hotel customer satisfaction

Conclusion

Conventional IPA was developed as a tool to facilitate prioritization of improvements and resource allocation. The three-factor theory of customer satisfaction indicates the existence of a non-linear relationship between attribute performance (satisfaction) and importance; however, this theory creates questions regarding the applicability of IPA and resulting managerial recommendations. Managers must be aware that changes to attribute performance (satisfaction) are associated with changes to

References (80)

  • I. Kaastra et al.

    Designing a neural network for forecasting financial and economic time series

    Neurocomputing

    (1996)
  • J.F.C. Khaw et al.

    Optimal design of neural network using the Taguchi method

    Neurocomputing

    (1995)
  • R. Law

    Back-propagation learning in improving the accuracy of neural network-based tourism demand forecasting

    Tourism Management

    (2000)
  • K. Matzler et al.

    The asymmetric relationship between attribute-level performance and overall customer satisfaction: a reconsideration of the importance–performance analysis

    Industrial Marketing Management

    (2004)
  • K. Matzler et al.

    How to make product development projects more successful by integrating Kano’s model of customer satisfaction into quality function deployment

    Technovation

    (1998)
  • H. Oh

    Revisiting importance–performance analysis

    Tourism Management

    (2001)
  • C. Ryan et al.

    Tourists and aboriginal people

    Annals of Tourism Research

    (2002)
  • M. Schumacher et al.

    Neural networks and logistics regression: Part I

    Computational Statistics & Data Analysis

    (1996)
  • S.A. Taylor

    Assessing regression-based importance weights for quality perceptions and satisfaction judgments in the presence of higher order and/or interaction effects

    Journal of Retailing

    (1997)
  • H. Tikkanen et al.

    The concept of satisfaction in industrial markets: a contextual perspective and a case study from the software industry

    Industrial Marketing Management

    (2000)
  • S.-H. Tsaur et al.

    Determinants of guest loyalty to international tourist hotels—a neural network approach

    Tourism Management

    (2002)
  • J.V. Tu

    Advantages and disadvantages of using artificial neural networks versus logistic regression for predicting medical outcomes

    Journal of Clinical Epidemiology

    (1996)
  • W. Vach et al.

    Neural networks and logistics regression: Part II

    Computational Statistics & Data Analysis

    (1996)
  • T.-Y. Wang et al.

    Forecasting innovation performance via neural networks—a case of Taiwanese manufacturing industry

    Technovation

    (2006)
  • D. Wu et al.

    Using DEA-neural network approach to evaluate branch efficiency of a large Canadian bank

    Expert Systems with Applications

    (2006)
  • H.Q. Zhang et al.

    Application of importance–performance model in tour guides’ performance: evidence from mainland Chinese outbound visitors in Hong Kong

    Tourism Management

    (2004)
  • E.W. Anderson et al.

    Customer satisfaction, market share and profitability: findings from Sweden

    Journal of Marketing

    (1994)
  • E.W. Anderson et al.

    Strengthening the satisfaction-profit chain

    Journal of Service Research

    (2000)
  • E.W. Anderson et al.

    The antecedents and consequences of customer satisfaction for firms

    Marketing Science

    (1993)
  • J. Antony et al.

    Evaluating service quality in a UK hotel chain: a case study

    International Journal of Contemporary Hospitality Management

    (2004)
  • G. Armstrong et al.

    Marketing: An introduction

    (2000)
  • L.R. Beach et al.

    The service quality improvement strategy: identifying priorities for change

    International Journal of Service Industry Management

    (1995)
  • B. Berman

    How to delight your customers

    California Management Review

    (2005)
  • C.M. Bishop

    Neural networks and their applications

    Review of Scientific Instruments

    (1994)
  • D.R. Brandt

    How service marketers can identify value-enhancing service elements

    The Journal of Service Marketing

    (1988)
  • J.I. Crompton et al.

    An investigation of the relative efficacy of four alternative approaches to importance–performance analysis

    Journal of the Academy of Marketing Science

    (1985)
  • J. Cronin et al.

    Measuring service quality: a reexamination and extension

    Journal of Marketing

    (1992)
  • A.L. Dolinsky et al.

    Adding a competitive dimension on importance–performance analysis: an application to traditional health care systems

    Health Marketing Quarterly

    (1991)
  • J.A. Eklof et al.

    On measuring interactions between customer satisfaction and financial results

    Total Quality Management

    (1999)
  • L. Fausett

    Fundamentals of neural networks

    (1994)
  • Cited by (118)

    • Service quality evaluation and service improvement using online reviews: A framework combining deep learning with a hierarchical service quality model

      2022, Electronic Commerce Research and Applications
      Citation Excerpt :

      These methods of service quality evaluation have the shortcoming of relying on experts' subjective scoring data and often lack sufficient samples (Wei et al. 2015). In addition, the Kano model (Hsu et al., 2018; Bi et al., 2019; Qi et al., 2016) and importance-performance analysis (IPA) (Deng et al., 2008) are often used for service quality evaluation. In addition to comprehensive evaluation methods, multivariate regression is applied to rank the importance of service quality dimensions (Palese and Usai, 2018).

    View all citing articles on Scopus
    View full text