How accurately can official energy performance calculations assess the real energy use in high-performance houses? This question was investigated by analysing 537 dwellings. Data on building characteristics and calculated performance from the Flemish Energy Performance of Buildings (EPB) database were complemented with data from energy utilities and surveys of inhabitants, their socio-demographic characteristics and user behaviours. While the real and theoretical energy uses were strongly correlated, the official calculation method overestimated the heating energy use of most houses while neglecting important electricity end uses. The prediction error varied strongly between individual cases. Two options within the calculation procedure had a significant impact on these prediction errors: the use of default values for the air tightness of the building envelope and the reported return temperature of the space heating system. The simplified calculation of net domestic hot water consumption and the real heating of the master bedrooms also affected prediction accuracy. However, extrapolations are hazardous due to the risk of selection and non-response biases implied by the approach and the need for further research into the causalities. Nonetheless, the findings stress the importance of accurate input data and realistic default values for calculation models used for high-performance buildings