A Bayesian multivariate Poisson regression model of cross-category store brand purchasing behavior

Abstract

The availability of cross-category transaction data in the retailing industry has enabled the investigation of interdependence in consumer purchase behavior across product categories. In this paper, we develop a multivariate count model to uncover and predict the pattern of cross-category store brand purchasing behavior. The proposed multivariate Poisson regression model, which we estimate using a Bayesian approach, provides flexibility in capturing cross-category correlations for sparse multivariate purchase data associated with infrequently purchased categories or purchasing in retail outlets such as warehouse clubs. We compare the goodness-of-fit of the proposed Poisson regression model with alternate benchmark models using customer purchase records across five product categories from a national warehouse club and find that the proposed model provides a superior fit. We also carry out a profitability analysis to illustrate the use of the model in planning cross-promotions.