We amend the canonical matching model by assuming diminishing returns to labor. We put the model to the twin test of generating a high volatility of labor market variables in response to productivity shocks (the “Shimer puzzle”) and a moderate response to changes in unemployment benefits and find that it passes that test. It does not feature wage rigidity, nor is it based on a small surplus calibration. Diminishing returns introduce a distinction between marginal and average surplus. With a standard (large average surplus) calibration, we can have a small marginal surplus, and thus a strong response of hiring to productivity shocks, while obtaining a measured response of unemployment to changes in benefits.