The method behind the UK Display Energy Certificate (DEC) improves the comparability of benchmarking by accounting for variations in weather and occupancy. To improve the comparability further, the incorporation of other features that are intrinsic to buildings (e.g. built form and building services) deserve exploration. This study investigates the impact of these features and explores ways to improve further comparability in benchmarking the energy performance of schools. Statistical analyses of approximately 7700 schools were performed, followed by analyses of causal factors in 465 schools in greater detail using artificial neural networks (ANNs), each designed to understand and identify the factors that have significant impact on the pattern of energy use of schools. Changes in the pattern of energy use of schools have occurred over the past four years. This fact highlights issues associated with static benchmarks. A significant difference in energy performance between primary and secondary schools meant that it was necessary to re-examine the way non-domestic buildings are classified. Factors were identified as having significant impact on the pattern of energy use. The characteristics raise new possibilities for developing sector- specific methods and improving comparability.