A Bayesian Network Model of the Consumer Complaint Process

Abstract

The purpose of this article is to present a Bayesian network model of the consumer complaint process. The outputs of the Bayesian model—conditional probabilities—provide much insight into the determinants and subsequent behavioral outcomes (e.g., full repatronage, limited repatronage, and exit; negative word-of-mouth behavior [WOM], no WOM, and positive WOM) of the complaint process. By distinguishing between (a) noncomplainers, (b) satisfied complainants, and (c) dissatisfied complainants, the model provides a rich, descriptive overview of the broader complaining behavior process. The model revealed several interesting findings; for example, the probability that a noncomplainer or a dissatisfied complainant will completely exit is quite low. The probabilities that a satisfied complainant will intend to fully repatronize the retailer and engage in positive word of mouth, on the other hand, are quite high. The advantages and limitations of Bayesian network models are discussed vis-à-vis structural equations models and partial least squares.