Probability Models for Customer-Base Analysis

Abstract

As more firms begin to collect (and seek value from) richer customer-level datasets, a focus on the emerging concept of customer-base analysis is becoming increasingly common and critical. Such analyses include forward-looking projections ranging from aggregate-level sales trajectories to individual-level conditional expectations (which, in turn, can be used to derive estimates of customer lifetime value). We provide an overview of a class of parsimonious models (called probability models) that are well-suited to meet these rising challenges. We first present a taxonomy that captures some of the key distinctions across different kinds of business settings and customer relationships, and identify some of the unique modeling and measurement issues that arise across them. We then provide deeper coverage of these modeling issues, first for noncontractual settings (i.e., situations in which customer “death” is unobservable), then contractual ones (i.e., situations in which customer “death” can be observed). We review recent literature in these areas, highlighting substantive insights that arise from the research as well as the methods used to capture them. We focus on practical applications that use appropriately chosen data summaries (such as recency and frequency) and rely on commonly available software packages (such as Microsoft Excel).