Multivariate Analysis of Consumer Preference Structures Across Multiple Categories

Abstract

That consumers’ purchase behavior varies across categories is being documented actively by the marketing science community. The variation in such behavior can be attributed to the heterogeneity in consumer preferences across categories as well as the nature of categories (e.g., perishable goods). In this research, we implement a multivariate probit model specification that helps deduce how the nature of a category influences a consumer’s preference structure not only for that category but across multiple categories. We use scanner panel data across multiple categories to calibrate the model. Over and above critically evaluating the results from this model, we derive marketing metrics using customer survey data from the same panel of customers. We then deduce the (i) effectiveness of a retailer’s pricing and promotional policies, and (ii) suggest directions for improving customer relationship management. Given the complex nature of the modeling approach, we use Hierarchical Bayesian methods (MCMC) to obtain model parameters.