An Improved BG/NBD Approach for Modeling Purchasing Behavior Using COM-Poisson Distribution

Abstract

Abstract — The concept of Customer Lifetime Value (CLV) attempts to account for the anticipated future profitability of each customer during his lifetime with the firm. In non-contractual context, in which the firm does not observe customer defection, the measurement of the CLV metric presents the challenge of choosing the appropriate model that provides satisfactory prediction of the future purchasing behavior of customers. The most prevalent models in non-contractual setting are the Pareto/NBD and the BG/NBD which are based on statistical distributions and assume that the number of transactions follows a Poisson distribution. However, many applications have an empirical distribution that does not fit a Poisson model. In this paper we propose an improved BG/NBD approach for modeling purchasing behavior using COM-Poisson Distribution, which is a generalization of the Poisson distribution to a two-parameter distribution, offering more flexibility and fitting better real world discrete data. An empirical study based on customer credit card transactions shows that the proposed model has better forecasting performance than competing models.