The energy reference area ERA is one of the most important reference values for practitioners and scientists to quantify the energetic performance of a building. Its fast and easy prediction from generally available building characteristics is a regular need. Here we present a comprehensive statistical model to predict the ERA based on a multiple regression approach and we motivate why the often chosen bivariate regression, i.e. simple regression should be disfavoured. Basis for the model development was a comprehensive data set of the Swiss buildings stock with about 30'000 buildings and 23 descriptive features. The final model only needs four building features to model the energy reference area. These features are the footprint, the living area, the number of floors and the buildings' usages. The main key performance indicators during the model development were the coefficient of determination R 2 and the uncertainty given by the root-mean-square-error RMSE, which was determined on a test set (about 10% of the dataset). The here presented multiple model has an R 2 of 99.7% and gives low uncertainties for the predicted ERA ranging between 25% and 10% of the predicted value. Thus this work allows estimating the energy reference area and its level of uncertainty fast and easy. We are convinced that our efforts will support practitioners as well as scientists in their daily work by providing more accurate and less biased estimations of the ERA.