Although simulation is increasingly being used in building system design, the full potential of simulation is usually not achieved. To improve system performance, designers generally guess dierent values of system parameters and redo the simulation. This is inecient and labor intensive. Also, if the number of parameters being varied exceeds two or three, the designer can be overwhelmed in trying to understand the nonlinear interactions of the parameters. However, techniques exist that allow automatic, multidimensional optimization of a simulation model, leading to better design with less eort. We describe how such optimization can be done using GenOpt, a generic optimization pro- gram. We show an example of how to use GenOpt to design an oce building such that source energy consumption for heating, cooling, and lighting is minimal with respect to selected design parameters. In this example, the optimization yields 14% energy savings. The additional time required to set up the optimization is about an hour. The measures found by using optimization not only decrease operating costs, but also lead to better daylighting usage, which results in higher comfort for the building occupants.