Energy analysis programs are often used for predicting the impact of energy-related building retrofit on the annual energy consumption and cost. However, due to many uncertainties involved in the development of the input file, the energy savings are often overestimated or under estimated. In order to increase the accuracy of predictions, this paper proposes the use of decision models under uncertainty, to determine the most profitable alternative, given the possible errors which may be introduced by the user in the input file. This approach is applied to a large existing office building in Montreal. The predicted annual cost savings, the payback period and the benefit-cost ratio are calculated using the building energy analysis software MICR0-D0E2.1E. The results are then analyzed using different criteria and, finally, some energy conservation measures are selected.