Evaluating the energy consumption of different design solutions during the design process is essential to reach energy targets. Conventionally, building energy performance is evaluated with energy simulations using a single input weather file (design or typical year) referring only to present weather conditions. Recent works, for example by Gaterell at al. for sites in the United Kingdom, suggest, however, that it is necessary to also include weather files describing climate conditions in future years. That is, the effect of predicted climate change on building performance. Buildings have a life span of 50 to 100 years and so must perform satisfactorily under both current and future climate, adjusting to take advantage of opportunities and to moderate potential damage caused by a changing climate (Wilde et al. 2008). Given the fact that climate is a singularly stochastic phenomenon and that we cannot predict the weather in future years with complete certainty, weather inputs necessarily have intrinsic uncertainties. Using a single weather file in building simulations, regardless of its source or generative algorithm, could lead to inaccurate energy consumption forecasts, and therefore wrong design decisions. In any case, the development of typical weather files was not always done keeping in mind forecasting, but rather what-if analyses. The innovation in our study consists in the introduction of more than one input weather file in building simulation to represent both present and future years. Our methodology aims to assess the robustness of different design solutions over many possible future climate projections, i.e. the sensitivity of a design or device to uncertain climatic outcomes. We focus on the robustness of refurbishment measures compared to the robustness of the non-refurbished building. We assess each intervention in terms of energy consumption. We decided to focus on existing buildings since they represent the biggest opportunity to reach energy targets today in Europe. In fact, the majority of the European building stock was constructed before any energy regulations and tends to perform poorly for energy consumption and comfort. We propose three indices to make comparisons between refurbishments. The Robustness Index (RI) compares the robustness of the each solution in terms of its range of energy usage. The Energy Saving Index (ESI) assesses the refurbished models in terms of energy usage difference in comparison with the non-refurbished model or base-case. The Gather Index (GI) summarizes the results of the two previous indices to compare the performance of the different refurbishments in terms of both robustness and energy efficiency. As a case study, we evaluated the robustness of an existing dwelling with twenty-two realistic refurbishments in Turin, Italy. The retrofit solutions focus on the thermal properties of the envelope by varying the U-value, the solar heat gains, the thermal mass and the air tightness of the envelope. The main outcome of our study is that interpreting numerical simulations of future energy consumption based on single-point estimates of input/output data (i.e. one typical weather file) is risky. Results are better treated as being probabilistic rather than deterministic, as is the norm today. Discussing outcomes in terms of ranges instead of single values improves estimates of outcomes, making it possible to identify robust design solutions for policy and investment decisions.