Today energy saving rates due to retrofits in residential housing are either measured by experiment on site or predicted by system simulation. In the latter case mathematical models of the components are developed, verified by laboratory tests and combined to represent the thermal characteristics of the investigated system. Then energy demand is calculated before and after adopting the investigated retrofit. There still remains the question, are energy savings precalculated by system simulation: obtained in fact, in other words do they agree with results of field experiments. Long-term- measurements in 51 residential buildings before and after renewal of the boiler prove, that due to varying occupancy behaviour and weather conditions no reliable results are obtained by such experiments on site. Therefore the correctness of simulation results cannot be confirmed by field measurements. However, the results measured can be explained by system simulation, calculating the energy consumption before and after renewal of the boiler under uniform boundary conditions. Thereby the strong influence of occupancy behaviour on energy consumption and energy saving rate is confirmed.