Preference Discovery in Product Search

TLDR: This paper developed a model of sequential search with preference discovery, and derived several key insights, including that novice consumers may achieve higher rewards net of search cost when there is more noise in initial beliefs about product attributes.

Abstract: Novice consumers, even when rational, sometimes discover and revise their preferences for certain product attributes as they evaluate products. Such preference discovery modifies consumers’ decisions with respect to (1) which products to evaluate in optimal sequential search and (2) which products to purchase. We develop a model of sequential search with preference discovery, and derive several key insights. First, we show that novice consumers may achieve higher rewards net of search cost when there is more noise in initial beliefs about product attributes. For expert consumers, less noise is better. Second, product recommendations can influence the future search path of the consumer, not merely introduce the recommended product into their consideration set. If a product exposes the consumer to a new, previously undiscovered attribute, the consumer may shift his search to a new part of the product space. Third, the value of a product recommendation is not necessarily directly related to the quality of the product. Counter to the literature on recommender systems, it is sometimes better to recommend an inferior product with the goal of helping consumers discover preferences. Non-benevolent agents can direct consumers to profitable products even though the consumer believes he is searching endogenously and rejects the agent’s recommendation. Finally, I present data from an incentive-aligned experiment to suggest that real consumers discover preferences.