Marshall Burke and fellow researchers study productivity in smallholder farms to understand variation across the adbundant but understudied firms. They use a novel framework, satellite data, and machine learning to understand such variation, and they find that output measurement error contributes significantly to this discrepancy in productivity.