To optimally allocate their marketing mix across customers, firms need to consider the evolution of their customers over time. Changes in the marketing environment, as well as intrinsic changes in preferences or needs, may discretely shift customers into different buying-behavior states. The ability to identify the dynamics in customer behavior and the drivers of these dynamics, present an opportunity for firms to influence the movement of customers to more favorable states of buying behavior. Accordingly, we address the following managerial questions in this dissertation: (1) how can firms dynamically segment their customer base? (2) what are the short- and long-term effects of marketing activities? and (3) how should firms allocate and target their marketing resources to maximize long-term profitability?

To address these questions we propose a non-homogeneous hidden Markov model that accounts for dynamics in customer behavior, the long-term impact of marketing actions, and customer heterogeneity. We capture dynamics by allowing customers to transition over time among a set of latent states of buying behavior. We develop a unique and flexible approach to capture the enduring effect of marketing actions by incorporating a non-stationary transition matrix that is dynamically affected by these actions. To optimally allocate marketing activities, we solve a partially observable Markov decision process that takes into account the evolution in customers' behavior and the possibly enduring effect of marketing actions.

We apply the model in the context of direct-to-physicians marketing by a major pharmaceutical firm. Our model suggests that physicians transition over time among three behavioral states, showing a high degree of dynamics. Furthermore, the marketing activities have varying degrees of short- and long-term effects that depend on the physician's prescription-behavior state. Specifically, we find that (i) both detailing and sampling have mostly long-term effects; (ii) detailing and sampling have a total duration impact of approximately 10 and 5 months, respectively; (iii) detailing is most effective as an acquisition marketing tool, whereas sampling is most effective as a retention tool. Using a counterfactual analysis, our optimization results show that the firm could increase prescriptions and profit by as much as 53.8% and 82.6%, respectively. Moreover, our analysis suggests that the firm should decrease its current detailing and sampling efforts by 30% and 14%, respectively. The integrative framework we propose provides important marketing implications for managing customers and maximizing long-run profitability.