"Bottom-up” engineering-economic modeling and analysis plays an important role in informing energy policies and shaping research priorities. For example, ex ante engineering estimates of the costs and benefits of energy saving measures guides public investment in energy efficiency to a significant extent. Once implemented, it is standard to conduct detailed evaluations to assess energy efficiency program impacts. However, these ex post evaluation activities typically stop short of trying to reconcile estimates of realized policy impacts with the ex ante predicted impacts. We are in the process of conducting a a state of the art evaluation of the nation's largest residential energy efficiency program, the Federal Weatherization Assistance Program. In addition to obtaining precise and credible measures of the impacts of weatherization assistance on household energy and expenditures, we are particularly interested in comparing the results of thousands of household-level energy audits against empirical estimates of realized energy savings. To the extent that we document a gap between the two, we aim to identify the source of the discrepancy. Possible explanations we will investigate include miscalibration of the engineering model, misspecification of the engineering model, poor installation of efficiency measures, and human behavior effects (including the so called “rebound” effect).