This paper describes a new methodology to assess the robustness of building performance in the long term with a probabilistic approach. The aim is to include uncertainties related to climate change predictions as well as the intrinsic uncertainties in weather files describing them. A case study focussing on refurbishment strategies of a realistic building in Turin is presented to demonstrate the methodological steps. The main outcome is that it is advisable to have outcomes in terms of ranges of energy consumption instead of single output values to evaluate energy efficient design solutions in both present and future years.