Personalized Pricing through User Profiling in Social Networks

Abstract

User profiling allows product sellers to identify users’ willingness to pay and enable personalized pricing. However, users’ information exploited in profiling is usually private and hard to obtain accurately due to users’ privacy concerns. With the increasing popularity of social networks, where users reveal their private information through social interactions, more sellers today profile users through their social data. This paper is the first to study how a seller optimizes personalized pricing through user profiling on social networks, where users proactively react by controlling their social activities and information leakage. We formulate and analyze a dynamic Bayesian game played between users and the seller. First, users decide their social activities by trading off the social network benefit against the potential risk of revealing private information. Then, the seller exploits users’ profiles to determine the personalized prices for the profiled users and a uniform price for the non-profiled users. It is challenging to analyze the Perfect Bayesian Equilibrium (PBE) of this game due to i) the randomness in user profiling, and ii) the coupling among users’ activity levels and that between the seller’s pricing decisions and users’ social activities. Despite the difficulty, we propose to alternate backward induction and forward induction to successfully solve the PBE. We show the surprising result that users’ activity levels do not monotonically decrease as the profiling technology improves. Instead, when user profiling is of high accuracy, the seller strategically chooses a high uniform price to stimulate their increased social activities to profile more users.