How can we get more ‘output,’ and of the right sort, from policing? The question has only taken on greater importance with recent, widely publicized instances of police misconduct; declines in public trust in police; and a rise in gun violence, all disproportionately concentrated in economically disadvantaged communities of color. Research typically focuses on two levers: (1) police resources, and (2) policing strategies or policies, historically focused on crime control but increasingly also on accountability, transparency, and fairness. Here we examine a third lever: management quality. We present three types of evidence. First, we show there is substantial variability in violent crime and police use of force both across cities and within a city across police districts, and that this variation is related to the timing of police leader tenures. Second, we show that an effort to change police management in selected districts in Chicago generates sizable changes in policing outcomes. Third, as part of that management intervention the department adopted a predictive policing tool that randomizes which high-crime areas it shows to officers. We use that randomization to generate district-specific measures of implementation fidelity and show that, even within the context of a management intervention designed to improve implementation of the department’s strategies, there is variability in implementation.