First Ranked First to Serve: Strategic Agents in a Service Contest

Abstract

We model a service provider whose agents endogenously determine their participation and performance level. This paper is motivated by two-sided marketplaces and work-from-home call centers, which crowdsource incoming demand to a pool of agents. The agents are ranked in a prespecified number of priority classes based on their relative performance. Higher performers are more utilized and agents' idle time is not compensated. We study which priority class structure creates the best incentives for agents' participation and performance. We show that the optimal design of such a “service contest” is often coarse, namely, it contains only a few priority classes. Discarding available information on agents' relative rankings, or, deploying coarser priority classes, can paradoxically provide higher incentives for agents' participation and performance.