In many instances, the social value of an innovation is much larger than the profits that a firm can obtain by selling the innovation on the market. When this is the case, a research contest can help align incentives and increase welfare. This paper examines the optimal design of research contests when the objective of the contest designer is the discovery and broad adoption of socially valuable innovations. We show that the contest designer benefits from conditioning the size of the prize on the market performance of the winner. The optimal contest features two quantity cutoffs and two prize levels. The low prize is awarded if the winner sells a quantity greater than the first cutoff while the high prize is awarded if the winner sells a quantity greater than the second cutoff.