Personalized Recommendation in a Retail Platform Under the Hybrid Selling Mode

Abstract

Retail platforms have widely implemented recommender systems to provide personalized recommendations to consumers, influencing sales significantly. However, under the hybrid selling mode where platforms offer both their products and third-party sellers’ products, the profitability of a recommender system and the optimal allocation of recommendations become critical considerations. This paper introduces a game-theoretic model to investigate these issues and unveil how a recommender system and its characteristics influence prices and profits. A key finding is that the recommender system increases prices and profits only if the commission rate is high and the system is profit-oriented or inaccurate. Surprisingly, higher recommendation accuracy does not always translate into higher profits; it is advantageous only in a consumer-oriented system. Moreover, the retail platform tends to allocate more recommendations to its own product than to the third-party seller’s product, a strategy known as self-preferencing. This strategy gives the platform a competitive edge and boosts its profit compared to the third-party seller. Furthermore, the degree of self-preferencing varies with the accuracy and orientation of the recommendation system. Specifically, in a consumer-oriented system, self-preferencing increases with accuracy, while in a profit-oriented system, it decreases with accuracy.