HMM For Conversion Rate Dynamics in Online Retail

In this study, we use a proprietary data set from an online marketplace to study the conversion rate dynamics in online retail. We examine how seller-level covariates, such as online sellers’ pricing and product strategies, marketing efforts, service responsiveness, reputation scores, product quality ratings, and other attributes, affect conversion rates. Specifically, we address the following research questions: (i) How do sellers’ covariates affect their conversion rates? (ii) Is the relationship between the conversion rate and sellers’ covariates state dependent? (iii) If the relationship is state dependent, what are the factors that determine the states and the state transitions? A hidden Markov model is adopted in the effort to answer these questions. The estimation results indicate that there are two states that affect conversion rate dynamics. The relationship between the conversion rate and the sellers’ covariates is state dependent, that is, given the different states of sellers, the effects of the sellers’ covariates on their conversion rates are different. We also estimate the thresholds between states. The results provide important guidance for sellers regarding what attributes they need to improve, to increase their conversion rates.