In the UK, the Standard Assessment Procedure (SAP) is used to rate the energy performance of existing dwellings whenever they are let or sold. This study investigates which of the inputs to SAP account for the most variance in energy rating across existing gas central heated houses in England. Data from the English Housing Survey (EHS) 2009 are used to generate a representative set of dwellings and variance-based global sensitivity analysis is then applied to assess each input's contribution to the variance in the calculated ratings. It is demonstrated that heating system efficiency, external wall U-value and dwelling geometry account for 75% of the variance of the energy rating of gas central heated houses in England. This suggests that improving heating system efficiencies and wall U-values of the worst performing dwellings will go a long way towards improving their energy rating and potentially reducing their energy consumption. It is also demonstrated that dwelling geometry has a much bigger influence on the calculated carbon emissions (accounting for 80% of the variance) than it does on the SAP energy rating (accounting for 30%), meaning that significant improvements in energy rating might not be accompanied by significant reductions in carbon emissions. © 2014 © 2014 The Author(s). Published by Taylor & Francis.