The conventional method of obtaining a detailed picture of the energy performance of a building is by the installation of a dense network of air temperature sensors. The data from all the sensors is accumulated at a central point for analysis. However, it is argued that air temperature measurement is merely a secondary indicator of energy use and a more direct method of tracking energy flow is preferable. Focussing on the types of heating system where heat is produced at a central source and pumped around the building (for example, hydronic systems), it is suggested the heating system itself can function as a low-cost distributed sensor network to directly monitor energy use, although the information retrieved is mixed in a complex way that is dependent on the topology of the distribution system. It is claimed that high-resolution information can nevertheless be extracted by mapping the acquired data to the geometry of the system and by the use of adequate computer processing power to run the physical model. A method of testing the concept in a domestic context is described as a precursor to more extensive deployment. (C) 2008 Elsevier B.V. All rights reserved.