Occupant behaviour can cause major discrepancies between the designed and the real total energy use in buildings. A possible solution to reduce the differences between predictions and actual performances is designing robust buildings, i.e. buildings whose performances show little variations with alternating occupant behaviour patterns. The aim of this work was to investigate how alternating occupant behaviour patterns impact the performance of different envelope design solutions in terms of building robustness. Probabilistic models of occupants' window opening and use of shading were implemented in a dynamic building energy simulation tool (IDA ICE). The analysis was carried out by simulating 15 building envelope designs in different thermal zones of an Office Reference Building in 3 climates: Stockholm, Frankfurt and Athens.In general, robustness towards changes in occupants' behaviour increased with increasing thermal mass and with decreasing transparent area of the envelope. The importance of the robustness' evaluation is highlighted in this paper, in order to obtain optimized buildings' designs for more accurate and realistic energy predictions.