HOW CAN CONSUMER PREFERENCES BE LEVERAGED FOR TARGETED UPSELLING IN CABLE TV SERVICES?

摘要

Internet TV has attracted a significant amount of attention from the conventional cable TV serviceproviders, by providing customized TV programs at preferred time slots. The cable TV service providersare seeking to retain their customers by giving them a better experience: by understanding theircustomers’ preferences and upselling them the right products to cater to their interests. It is not easy tounderstand customer preferences though, since customers are not able to watch channels to whichthey have not subscribed. This makes it difficult to predict what they will like to watch, as a result. In thispaper, I discuss my ongoing research on TV viewership behavior. I model customer preferences usinga technique called latent Dirichlet analysis (LDA), by considering channel viewing behavior as a similarprocess of article generation. Customer preferences over unsubscribed channel are calculated from theLDA model. My model achieves better prediction performance as a result. I also present a quantitativecase study to show that the observed channel viewing behavior makes sense.