Essays in Sequential Search

Abstract

Models of sequential search arise in a wide variety of contexts, including organizations hiring employees, firms developing technologies, marketers analyzing the behavior of consumers, and individuals searching for housing or investment opportunities. In many sequential search problems, decision makers must navigate complex uncertainties and trade-offs while exploring alternatives. In competitive markets, alternatives may have uncertain availability, as desirable options may disappear during the search process due to competition. Moreover, the acquisition of a selected alternative may not be guaranteed and may depend on the offer price chosen, introducing further trade-offs. In dynamic markets where conditions could change rapidly, the distributions of alternative values may be unknown due to limited prior information, requiring decisions to be based solely on observed data. These features of the search problem, common in dynamic and competitive markets, substantially complicate the problem, obscure effective search policies, and reduce overall value.In this dissertation, we study sequential search through the lens of sequential decision making under uncertainty and data-driven decision making. We consider two main search problems that arise in practical settings: Sequential Search with Acquisition Uncertainty and Data-Driven Sequential Search. Both problems are difficult due to complex sequencing, stopping, and pricing decisions. Given this difficulty, we focus on approximations of the search models. Through these approximations, we develop feasible search policies. We show that our policies not only are simple, interpretable, and easy to implement in practice, but also achieve provably strong performance guarantees. Specifically, for search problems with acquisition uncertainty, we rely on policies based on a well-chosen selection threshold. For data-driven search problems, we rely on policies based on a well-chosen search effort. Our main takeaway is that, despite the difficulty of these problems, our policies offer a simple and systematic way to search effectively. Our results (i) generate key managerial insights and shed light on complex search problems, (ii) facilitate the search process for decision makers, and (iii) provide guidelines for designing effective platforms and marketplaces involving sequential search applications.