Decentralisation reforms have shifted responsibility for public service delivery to local government, yet little is known about how their management practices or behaviour shape performance. The authors of this paper conducted a comprehensive management survey of mid-level education bureaucrats and their staff in every district in Tanzania, and employ flexible machine learning techniques to identify important management practices associated with learning outcomes. They find that management practices explain 10% of variation in a district's exam performance. The three management practices most predictive of performance are: i) the frequency of school visits; ii) school and teacher incentives administered by the district manager; and iii) performance review of staff. Although the model is not causal, these findings suggest the importance of robust systems to motivate district staff, schools, and teachers, that include frequent monitoring of schools. They also show the importance of surveying subordinates of managers, in order to produce richer information on management practices.