Multichannel Marketing and Hidden Markov Models

Understanding how customers' channel preferences evolve is crucial to firms in managing multiple channels effectively. This dissertation examines the underlying influences which cause customers to migrate over time into different unobservable experience states with higher propensity to purchase in a specific channel. I apply a multiple-segment Hidden Markov Model (HMM) to discover the dynamic behavior of customers facing multiple channels. The study in Chapter 3 takes an alternative way to incorporate heterogeneity using structural segments, which allows heterogeneity in both state transition and channel choice, and offers substantive and interesting insights regarding multichannel shopping patterns. In the empirical application, I identify two segments and two states in the multiple-segment HMM and examine different learning patterns and rate of experience development for each segment. My results show that over time, customers do not tend to move away from bricks-and-mortar stores as some experts expect as they gain more experience. Some customers perform multichannel-oriented behavior and show various evolving patterns. Customers also reveal different reactions to marketing communications for different combinations of channel tendency. Also, the proposed model suggests an effective way for a firm to dynamically segment and manage channel usage with its customer base. Based on empirically-derived insight regarding customer channel preference evolution with experience, marketers can allocate a firm's limited resources effectively and further refine marketing strategies. Furthermore, customer retention and churn has received increasing attention in the field of customer relationship management (CRM) in recent years. Chapter 4 provides a framework to estimate a relationship dynamics in a non-contractual setting whose customers' dropout time is not clearly stated and easily observed. I incorporate the effects of channel experiences and marketing communications on relationship dynamics, and use a nested structure to detect purchase preference and channel evolution simultaneously until "death" of a relationship, and identify a more (in)active-oriented channel. The proposed nested multinomial HMM addresses changes in preference of purchase incidence and channel choice across time with respect to various relationship states, and deals with the impact of marketing communications and channel experiences on customer retention as governed by transitions between relationship states.