Discovering heterogeneous consumer groups from sales transaction data

Abstract

We propose a demand estimation method to discover heterogeneous consumer groups. The estimation requires only historical sales data and product availability. Consumers belonging to different segments possess heterogeneous preferences and, in turn, heterogeneous substitution behaviors. For such consumers, the latent class consumer choice model can better represent their heterogeneous purchasing behaviors. In the latent class choice model, there are multiple consumer segments, and the segment types are not observable to the retailer. The expectation-maximization (EM) method is developed to jointly estimate the arrival rate and the parameters of the choice model. The developed method enables a simple estimation procedure by treating the observed data as incomplete observations of the consumer type along with consumer's first choice. The first choice is the choice before the substitution effects occur. We test the procedure on simulated data sets. The results show that the procedure effectively detects heterogeneous consumer segments that have significant market presence. © 2019