Optimal Pricing Policy with Recommender Systems

We look at one of the informational novelties introduced by the existence of internet market, namely "the recommender systems". A recommender system is a system employed by some internet sellers, which collects data from all previous customers about their experiences and makes inferences from this data to recommend a product to an active customer. The recommender system can also be interpreted as a peerto-peer system where the seller provides a platform for the buyers to share their experiences. We interpret the role of a recommender system as reducing uncertainty for the customers, which creates some additonal surplus to be distributed between the customers and sellers employing such systems. We differentiate the customers with respect to the extremity of their preferences, which also implies di®erent valuations for decreased uncertainty. We show that an internet seller employing such as system can extract a non-negligable share of this surplus from the customers through higher prices in the presence of a competitive fringe without recommender systems. However optimal pricing by the seller with the system leads to a less than full market share, since the seller nds it optimal to leave out the buyers with moderate taste to the fringe. Thus the optimal pricing mechanism does not employ the recommender system at the e±cient level, in other words there is under utilization. We also nd that the overall under-utilization might entail over-utilization of the system for some products and under-utilization for others.