Investigating the non-linear effects of e-service quality dimensions on customer satisfaction $ Adam Finn n University of Alberta School of Business, Edmonton, Alberta, Canada T6G 2R6 article info Keywords: e-Service quality Customer satisfaction Non-linearity Retail websites abstract The literature on service quality initially focused on identifying the service attributes that drive overall measures of customer satisfaction. More recently, the assumption that attribute-level performance is linearly related to customer satisfaction has been challenged. Inspired by Kano’s work on product quality, service researchers have used questionable methods to classify service attributes as attractive, one-dimensional, or a must-be, based on the observed shape of their satisfaction response functions. Valid assessment of the shape of satisfaction response functions for services requires crossed service by respondent ratings data to control for differences in respondent’s scale use in service assessment. Application of a recommended approach identifies download speed as a must-be performance dimension that interacts negatively with site functionality as the only non-linearity for online retailers. Currently used methods produce quite different results. & 2010 Elsevier Ltd. All rights reserved. 1. Introduction Service managers need to understand how perceptions of their performance on service quality dimensions influence levels of customer satisfaction. The literature shows positive effects of customer satisfaction on such desirable outcomes as repeat purchase (Szymanski and Henard, 2001), retention (Bolton, 1998), loyalty (Anderson and Sullivan, 1993), retailer sales performance (Gomez et al., 2004), and profitability (Anderson et al., 1994; Bernhardt et al., 2000). Reports about which service attributes drive levels of customer satisfaction are available for such application areas as banking (Levesque and McDougall, 1996), software (Kekre et al., 1995), and online shopping (Szymanski and Hise, 2000). However, the initial assumption that attribute-level perfor- mance is linearly related to overall customer satisfaction has now been challenged (Mittal et al., 1998; Anderson and Mittal, 2000). Negative asymmetry or decreasing returns, often attributed to prospect theory (Kahneman and Tversky, 1979), has been found using regression analysis and cross-sectional survey data in a health care and an automobile setting (Mittal et al., 1998), for hypermarkets (Ting and Chen, 2002), for a supplier in the automotive industry (Matzler et al., 2004), and for an educational program e-portal (Cheung and Lee, 2004). Service attributes have even been classified as attractive, one-dimensional, or must-be by Kano et al. (1984), based on the shape of their response functions. 1 For example, Ting and Chen (2002) determined a ‘children’s playroom’ was attractive, ‘quick checkouts’ was one- dimensional, and ‘sufficient parking area’ was a must-be for Taiwanese hypermarkets. Non-linear response functions mean managers not only have to know if an attribute is a driver of customer satisfaction, they need to know the shape of the response function for each driver to identify priorities and allocate resources. If a response function is linear, an improve- ment in quality has the same effect on customer satisfaction, no matter where the service provider is located on the attribute. If the response function is non-linear, the effect of the im- provement depends on the slope of the function at the point where the provider is located on the attribute. But as detailed below, to draw conclusions the published research has relied on samples that truncate the distribution of quality and satis- faction that is investigated, treated categorical ratings as if they were all ratio scale data, and employed regression models that could be expected to produce biased parameter estimates. The purpose of this research is to compare an appropriate method for identifying any non-linear effects of service quality dimensions on customer satisfaction with the methods that are currently being used in the literature. Valid identification of the Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/jretconser Journal of Retailing and Consumer Services 0969-6989/$ - see front matter & 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.jretconser.2010.09.002 $ Support for this research was provided by a Social Sciences and Humanities Council of Canada Initiative on the New Economy Research Grant. n Tel.: + 1 780 492 5369; fax: + 1 403 492 3325. E-mail address: adam.finn@ualberta.ca 1 Oliver (1997, p. 152) identified the same satisfaction response functions, which he labeled as for monovalent satisfier, bivalent satisfier, and monovalent dissatisfier needs. Journal of Retailing and Consumer Services 18 (2011) 27–37