In this paper we study the benefits of search costs in distributed multi-agent systems (MAS). These costs, often associated with obtaining, processing and evaluating information relating to other agents in the environment, can be either monetary or manifested in some tax on the agent's resources. Traditionally, such costs are considered as market inefficiency, and, as such, aimed to be reduced to the minimum. Here we show, in contrast, that in many MAS settings the introduction of search costs can actually improve market performance. This is demonstrated in three different settings. First we consider one-sided and two-sided (equilibrium-driven) search applications. In both settings we show that, while search costs may decrease the individual agents' outcomes, the overall market throughput may actually improve with the introduction of such costs. Next, we demonstrate a setting where, somewhat paradoxically, the introduction of search costs improves both the overall market throughput and the utility of each and every individual agent. We stress that we assume that the proceeds from the search costs are wasted, with no one directly benefiting from them. The importance of the results is for the design of MAS systems, where in many cases one should consider deliberately increasing (potentially artificially) the search friction to some desired level in order to improve the system's performance.