Building energy modeling is a common procedure for the analysis of energy efficiency retrofits. Smaller retrofits of isolated systems, such as equipment motors and lighting systems, can often be made without the need for complete energy modeling; however, when the retrofit affects multiple systems, such as those involving the building envelope or the heating or cooling system, or when the retrofits of motors and lighting systems are so significant that they affect the heating and cooling load of the building, a more complete energy analysis is necessary. Because the exact inputs to building energy models are never known, and some inputs to the model are stochastic in nature (e.g., occupancy, plug-loads, lighting loads, weather), deterministic prediction of energy use is not only invariably inaccurate, it is actually inappropriate. When simple deterministic energy savings without uncertainty are used in economic analyses (e.g., return on investment), it is difficult to analyze the risk/benefit of the retrofit investment with true accuracy. A stochastic simulation, which includes the effects of input uncertainty and stochastic inputs, is a more appropriate way to predict the building energy use. In this paper, we present a method for stochastic energy simulation that propagates probability characterizations of the input values through a computational engine to create probable energy use predictions. When this probable energy use is combined with forecasts of energy and construction costs, a probable estimate of return on energy efficiency measure investment is generated, and an economic risk/benefit analysis of the investment can be made. Such information is especially important to the growing energy service company market. The computational engine is based on the CEN/ISO monthly building energy calculation standards so its accuracy is well researched and validated, and the computational simplicity allows for efficient stochastic analysis.