Search Design and Broad Matching

We study decentralized mechanisms for allocating firms into search pools. The pools are created in response to noisy preference signals provided by consumers, who then browse the pools via costly random sequential search. Surplus-maximizing search pools are implementable in symmetric Nash equilibrium. Full extraction of the maximal surplus is implementable if and only if the distribution of consumer types satisfies a set of simple inequalities, which involve the relative fractions of consumers who like different products and the Bhattacharyya coeffi cient of similarity between their conditional signal distributions. The optimal mechanism can be simulated by a “keyword auction”with “broad matching”. (JEL D02, D47, D82). Consider a consumer who has a certain need and looks for a product that will satisfy it. When confronted with a particular product, the consumer can identify whether he wants it. However, when he first embarks on the search process, he is unable to provide an exact description of his need, and can only submit an imprecise “query”that fits several product types. How should a benevolent planner react to the consumer’s query? If he gives the consumer a single item, he risks ending up with a poor match. Instead, the planner may provide the consumer with a set of options to browse. If search is time-consuming, the planner should design the set in order to balance two considerations: maximizing the probability that the consumer will find what he is looking for, and minimizing the amount of time it will take him to find it. In other words, the planner’s problem is to design the consumer’s “search environment”. The consumer’s predicament is prevalent in other environments. When an employer approaches a Human Resource agency with an intention to hire a new worker, the most he can usually do is list a few characteristics that vaguely describe the kind of worker he is looking for. Likewise, when we look for a rental apartment through a real-estate agent, we typically describe only broad characteristics (location, size, amenities). A more modern example is online search, where users submit queries that often contain general and imprecise keywords. Imagine that you look for a specific piece of instrumental music on YouTube, but you forgot its name. You would recognize it instantaneously if you heard it, but the only information you can supply to YouTube is the piece’s genre. Finally, there are cases in which the consumer passively conveys information. For instance, in contemporary online platforms, the “cookies”on users’computers describe their navigation history, which may be correlated with their current needs. We attempt to capture situations of this kind with a simple model, in which every consumer is interested in only one type of product and can provide only one particular description or “query”concerning this product. Each product type is supplied by a continuum of firms; and when the consumer is given a subset of firms, he browses it using the most