Editor's note: Michael S. Garver is professor of marketing at Central Michigan University, Mt. Pleasant, Mich. Richard Divine is professor of marketing at Central Michigan University. Dominic Nieto is a student at Michigan State University.
Leading organizations use customer experience (CX) research to obtain marketplace feedback to drive strategic and tactical decisions. At the heart of this strategy, researchers and executives want to know how to improve the customer experience. While CX research is a powerful management tool, research suggests that the majority of CX programs need significant improvement (Allenson 2016). Most CX research improvements put forth by authors either focus on how to better use CX data to drive improvements or on improved research methods to collect and analyze data. This article will examine new CX research analysis tools for conducting key driver analysis.
Key driver analysis is a common procedure for statistically inferring the importance of CX attributes (independent variables) that drive the customer experience(dependent variable). Yet there are critical assumptions and limitations of key driver analysis that are frequently ignored. Two of these limitations are ignoring the existence of different key driver segments in the customer base and multicollinearity associated with independent variables or CX attributes. Academic researchers have developed new statistical modeling techniques to overcome both of these limitations. For example, latent class regression (LCR) has been employed to identify key driver segments in the customer base while correlated components regression (CCR) has been developed to overcome problems associated with multicollinearity. The purpose of this article is to put forth a process for using both LCR and CCR to conduct key driver analysis with CX data.
In a key driver analysis, CX attributes are given importance scores that inform management ...