Empirical analysis of the impact of product diversity on long-term performance of recommender systems

This study explains how the product diversity affects long-term performance of recommendation systems. We examine how the number of product categories offered to customers is related to customer churn incidence. We collect a large scale panel data consisting of product category, revenues and customer churn information from a large offline retailer. We find that as the number of product categories recommended increases, the likelihood that customers churn strikingly decreases after controlling for the number of individual products being recommended. Our results suggest that companies can achieve better outcomes in their recommendation systems by explicitly incorporating the diversity of products being offered to their customers. Further, simulation results show that our proposed diversity-based recommendation strategy can save the company approximately $26 million per year (7.5% of the company's annual revenue) by preventing customer churn.