We analyzed the peak load reductions due to a residential direct load control program for air-conditioners in southern Ontario in 2008. In this program, participant thermostats were increased by 2 degrees C for four hours on five event days. We used hourly, whole-house data for 195 participant households and 268 non-participant households, and four different methods of analysis ranging from simple spreadsheet-based comparisons of average loads on event days, to complex time-series regression. Average peak load reductions were 0.2-0.9 kWh/h per household, or 10-35%. However, there were large differences between event days and across event hours, and in results for the same event day/hour, with different analysis methods. There was also a wide range of load reductions between individual households, and only a minority of households contributed to any given event. Policy makers should be aware of how the choice of an analysis method may affect decisions regarding which demand-side management programs to support, and how they might be incentivized. We recommend greater use of time-series methods, although it might take time to become comfortable with their complexity. Further investigation of what type of households contribute most to aggregate load reductions would also help policy makers better target programs. Crown Copyright (C) 2011 Published by Elsevier Ltd. All rights reserved.