Several studies have shown that the actual energy performance of buildings can differ significantly from its designed value. An important part of this performance is constituted by the building fabric's thermal performance. A common method to evaluate the latter for an actual building is the co-heating test. The co-heating test comprises a quasi-stationary heating experiment followed by linear regression analysis of aggregated building performance data. This paper reviews related research work and cristallises the current state-of-the-art. The physical phenomena working behind the scenes of the generally assumed simplified heat balance are discussed. Statistical constraints generally prevent these from being uncovered fully during the analysis. Multiple linear regression is proposed as the most sensible method to analyse co-heating measurement data. A novel way to visualise such analysis, deduce building performance data graphically and compare different co-heating test results is presented. (C) 2014 Elsevier B.V. All rights reserved.