In recent years, more and more consumers who cancelled their newspaper subscription have signed up for online-flyer portal sites. However, their site access logs are not necessarily linked with their purchase records stored in ID-POS data, which is a marketing problem from the sellers' point of view. As a solution to this problem, we propose a practical framework of a hidden Markov model which allows for the assessment of the flyer advertisement. We also present a hierarchical Bayesian model which sheds light on the information search and shopping behavior of consumers, given individual heterogeneity. Our results demonstrate that the subscribers of online flyers are more likely to visit stores as they have more flyer accesses, email advertisements from portal sites and the announcements of time-limited sales.
