Concomitant Variable Latent Class Models for Conjoint Analysis

We propose a unifying framework for benefit and demographic segmentation based on the analysis of rank-order choice data collected in conjoint studies. The model identifies a number of unobserved segments, estimates the conjoint model within each segment, and at the same time estimates the association of segment membership with concomitant (consumer descriptor) variables. Analyses of synthetic datasets, and an illustrative application of the model to a conjoint analysis study on consumers' preferences for banking services are provided. The analyses of synthetic data support the validity of the concomitant variable latent class model. The validity of the model is also confirmed in an empirical study, where the preference patterns are shown to be logically consistent with the respondents' banking behavior