An education system needs valid and objective information on student learning outcomes at a sufficiently disaggregated geographical level. Such data can inform policymakers on specific aspects in each geographical area that require attention and provide an indication of the returns to public investments in education. They are the starting point for any effort to improve the performance of an education system. Without information on learning outcomes, policy reforms or higher public investments in education are unlikely to be effective. Indonesia's education system does not currently have such information. In this paper, we take advantage of a government policy to change the mode of the national examination administration, from paper-based to computer-based testing, to estimate the quality of education in Indonesia at the district level. Our results indicate that education quality across the country's districts is highly heterogeneous. The gap in results between the highest-scoring and lowest-scoring districts shows that children in the latter have been attending schools for nine years with minimal learning outcomes. We find that, over the course of one year, the average quality of education has increased slightly and the variation in quality has declined slightly. However, these national averages mask significant heterogeneities across and within districts.