Representative, site-specific weather data is a key requirement for building performance simulation. In the UK, such data is available in two formats, Test Reference Years for analysing building services loads under 'typical' year conditions and Design Summer Years for estimating summer discomfort of naturally ventilated and free-running buildings. Currently, Design Summer Years are determined as a complete year based on the rank of the average dry bulb temperature from April to September. The simplicity of this approach does not take into account extreme temperature values in individual months or the incident solar radiation, both of which are however of great significance for the summer overheating performance of a building. This paper analyses the implications of this simplified approach for the resulting data. It is shown that there is no consistent relation between the Design Summer Years and the corresponding Test Reference Years and that, for some sites, building performance simulations using Design Summer Year files deliver unreliable results. Practical application: This paper demonstrates that the current approach for deriving Design Summer Years (DSYs) can lead to data series that are not representative for near-extreme summer conditions at a given location. It highlights that a new approach for deriving near-extreme summer years for building performance simulation is needed in order to overcome the inherent shortfalls of the current DSY data.