Context-aware power management (CAPM) uses context (e.g., user location) likely to be available in future ubiquitous computing environments, to effectively power manage a building’s energy consuming devices. The objective of CAPM is to minimise overall energy consumption while maintaining user-perceived device performance. The principal context required by CAPM is when the user is not using and when the user is about to use a device. Accurately inferring this user context is challenging and there is a balance between how much energy additional context can save and how much it will cost energy wise. This paper presents results from a detailed user study that investigated the potential of such CAPM. The results show that CAPM is a hard problem. It is possible to get within 6% of the optimal policy, but policy performance is very dependent on user behaviour. Furthermore, adding more sensors to improve context inference can actually increase overall energy consumption.