Worldwide, gang proliferation is fought mostly with tough punishment strategies such as sweeps. In this paper, I study their causal effect on crime for arrested individuals and known peers following a difference-in-differences strategy. I also take advantage of the network structure I retrieved to assess peer effects and identify key players. I perform such an analysis with novel administrative data from the Metropolitan Area of Barcelona, where Latin gangs expanded rapidly and where a stark policy change occurred. Results show significant reductions in crimes of arrested individuals and their peers, particularly in crimes against the person. The areas of the sweeps benefit from improvements in crime, health and education. I further conduct an innovative counterfactual policy exercise comparing sweep outcomes with theoretically predicted crime reductions when removing key players. This exercise indicates that sweeps could have achieved a 50% larger reduction in criminal activity had key players been removed. In this way, a network analysis provides insights on how to improve policy design.