Bottom-up modelers typically predict a lower energy demand and a higher energy efficiency than top-down modelers do, leading to the notion of the energy efficiency gap. This difference is often ‘explained’ by combining bottom-up information with unrealistically high discount rates. In this paper we combine the bottom-up and top-down approaches in an energy demand model. The model has a top-down structure, but we employ bottom-up information to estimate most of its parameters, using the discount rate that firms say they use. This new approach provides a partial reconciliation of top-down and bottom-up methods, which proves to be very useful for policy analysis.