Long-term consumer selection strategies for a heterogeneous consumer base when demand depends on purchase and advertising histories

Abstract

For many retailers with an existing consumer base, direct marketing, particularly, one-to-one marketing actions, is crucial. However, to decide which consumer should receive which kind of advertising at which point in time is challenging, particularly if the individual consumer's demand is history dependent, that is, influenced by his/her previous sales as well as previously received advertising. To effectively manage a heterogeneous consumer base over time, in this paper, we propose a relaxed dynamic programming-based consumer selection strategy optimizing long-term objectives under budget constraints. To be able to apply our model under unknown consumer behavior, we integrate exemplary predictions for history-dependent demand based on observable data. Using different synthetic consumer behavior models, we analyze under which conditions history-dependent consumer behavior can be estimated accurately from historic data. Finally, we evaluate the performance of our approach based on data-driven demand predictions. Our evaluations show that our model is general applicable and outperforms established consumer selection heuristics as well as solutions for greedy short-term models by up to 20%. The selection policies obtained are discussed and managerial insights are inferred.