Predicting And Economically Exploiting UtilityThresholds With Utility-Based RecommendationSystems

Consumers spend a lot of time and effort on online product search in order to find the best match to their preferences. Recommendation systems promise to decrease these search costs. Much recent work has focused on refining methods for finding the best alternatives for a consumer. The question of how many of these alternatives the consumer actually wants to see has remained largely unanswered. This paper proposes utility thresholds as a novel approach to identifying the optimal recommendation set size. Beyond improving recommendation systems, utility thresholds improve business decision support by enabling willingness-to-pay estimation and individually optimal product configurations. Our empirical evaluation shows that utility threshold prediction is better for factors related to the recommendation process than personal factors. Search costs are reduced and willingness-to-pay estimates improved.