Group-based incentive pay is attractive in contexts where production is complex and interdependent, yet freeriding is a paramount concern. We assess the introduction of group-based performance pay in a modern industrial production setting using difference-in-difference estimation. Performance increased by 19 percent, with three quarters coming from increased performance of existing workers and the remaining from selection; workers became more efficient and were absent less often. We find little evidence of freeriding; quantile regressions show increased performance throughout the distribution of workers. Features of the design and implementation process created trust, a common goal, and a shared identity, which limited freeriding.