Search Before Trade-offs Are Known

Abstract

Search, broadly defined, is a critical managerial activity. Our contribution is a model of search for multiattribute alternatives, and our focus is on parallel search, where the decision is about the number of alternatives to explore. Most of the search literature considers univariate alternatives, and it can be applied to a multiattribute setting provided that the trade-offs to be used at the final selection stage were known at the search stage. However, uncertainty about trade-offs is likely to occur, especially in settings that involve parallel search (e.g., vendor selection, new product development, innovation tournaments). We show that incorporating uncertainty about trade-offs into a model changes its search strategy recommendations. Failing to account for such uncertainty, which is likely in practice, leads to suboptimal search and potentially large losses. For parallel search and a multivariate elliptical (e.g., normal) distribution of the alternatives, the solution is equivalent to univariate search with appropriately adjusted standard deviation. We prove that, in this setting, the optimal number of alternatives to explore increases if uncertainty about trade-offs increases, and we discuss the value of information about uncertain trade-offs.