Incorporating Nonrandom Direct Marketing Activity into Latent Attrition Models

Latent attrition models serve as the foundation for analyzing transactional customer relationships to derive forecasts of customer behavior and value. Yet, extant models often overlook the effects of direct marketing and how it is individually targeted. We propose a parsimonious model that allows for direct marketing to impact the frequency with which transactions occur, the monetary amount of a transaction, and the latent customer lifetime. Additionally, our modeling framework takes into account the way in which organizations may target direct marketing at the individual level. Using donation data from a non-profit organization, we find that direct marketing increases the donation rate and the amount donated for active donors. However, direct marketing is a double-edged sword for some donors because it accelerates their attrition. We use our model to examine the financial impact of direct marketing activities, deriving the residual lifetime value (RLV) associated with alternative direct marketing policies and the incremental RLV as a measure of the effectiveness of an additional piece of direct marketing. We examine how these metrics relate to summaries of past donation activity (recency, frequency, and average donation amount) and discuss potential shortcomings of basing direct marketing decisions on such heuristics.