Evaluations of energy efficiency programs reveal that realized savings consistently fall short of projections. This paper decomposes this 'performance wedge' using data from the Illinois Home Weatherization Assistance Program (IHWAP) and a machine learning-based event study research design. The authors find that bias in engineering models can account for up to 41% of the wedge, primarily from overestimated savings in wall insulation. Heterogeneity in workmanship can also account for a large fraction (43%) of the wedge, while the rebound effect can explain only 6%. They find substantial heterogeneity in energy-related benefits from IHWAP projects, suggesting opportunities for better targeting of investments.