Hot spots policing is used by many U.S. police departments and is consistently supported by evidence from evaluations of its effectiveness. Hot spot police work requires identifying small areas where crime is especially concentrated and directing police resources to these areas. This paper meta-analyses 22 reports in which random assignment was used to increase police activities in some hot spots. The analysis finds that the police activities reduced crime by .046 to .051 standard deviation units, with the most plausible estimate being .046. All the average effect sizes were statistically different from zero. However, about a third of the study-specific estimates differed reliably from zero and the moderator variables failed to explain variation in impact. Hot spots policing reduces crime, but its effect is small and inconsistent. We discuss implications for policing theory and practice, and provide suggestions to improve the use of meta-analysis for identifying effective programs and practices.