The public narrative surrounding efforts to improve low-performing K-12 schools in the United States has been notably gloomy. But what is known empirically about whether school improvement works, which policies are most effective, which contexts respond best to intervention, and how long it takes? We meta-analyze 141 estimates from 67 studies of post-No Child Left Behind Act turnaround policies. On average, policies had moderate positive effects on math and no effect on English Language Arts achievement on high-stakes exams. We find positive impacts on low-stakes exams and no evidence of harm on non-test outcomes. Extended learning time and teacher replacements predict greater effects. Contexts serving majority-Latina/o populations saw the largest improvements. We cannot rule out publication bias entirely but find no differences between peer-reviewed versus non-peer-reviewed estimates.