Crucial empirical data (currently absent in building energy models) on central heating demand temperatures and durations are presented. This data is derived from the first national survey of energy use in English homes and includes monitored temperatures in living rooms, central heating settings reported by participants, along with building, technical and behavioural data. The results are compared to model assumptions with respect to thermostat settings and heating durations. Contrary to assumptions, the use of controls did not reduce average maximum living room temperatures or duration of operation. Regulations, policies and programs may need to revise their assumptions that adding controls will reduce energy use. Alternative forms of heating control should be developed and tested to ascertain whether their use saves energy in real-world settings. Given the finding that detached houses are heated for longer, these dwellings should be particularly targeted in energy efficiency retrofit programs. Furthermore, social marketing programs could use the wide variation in thermostat settings as the foundation of a ‘social norm’ program aimed at reducing temperatures in ‘overheated’ homes. Finally, building energy models that inform energy policies require firmer foundations in real world data to improve policy effectiveness. Greater coordination of data collection and management would make more data available for this purpose.