Building performance simulation has a key role in facilitating effective retrofitting of commercial buildings, an essential task to reduce Australia's greenhouse gas emissions. Simulation users commonly rely on assumptions for certain 'hard-to-measure' building and behaviour inputs, and the range of values use can significantly affect predicted energy consumption. New financing methods in Australia may attach financial risk to this simulation uncertainty. This paper reports on a study examining the sensitivity of total predicted building energy consumption to hard-to-measure simulation inputs for two template Australian office buildings; a 10-storey office tower located in a Central Business District (CBD), and a 3-storey office on the fringe of a CBD. The predicted energy consumption for the archetypes varied by more than 50% from baseline consumption for all Australian capital cities, using high and low assumptions. The input parameters that significantly influence energy consumption were found to be; cooling set-point, Information and Communications Technology (ICT) power density, ICT usage schedule, and lighting power density. A case study of a simple lighting upgrade to a 10-storey office building in Sydney showed that the payback period could vary from 2.4 to 10.3 years depending on the simulation assumptions used. (C) 2014 Elsevier B.V. All rights reserved.