Being able to disaggregate total energy demand into components attributable to specific end uses provides useful information and represents a primary input into any attempt to simulate the impact of policies aimed at encouraging households to use less energy or shift load. Conceptually the estimation problem can be solved by directly metering individual appliances. Not surprisingly, this has not been widely practised and by far the most common estimation procedure has been the indirect statistical approach known as conditional demand analysis. More recently, with access to limited direct metering, both approaches have been used in combination. This paper reports on a substantial modelling exercise that represents a unique example of combining data of this type. The distinctive aspects are the extent and richness of the metering data and the fact that optimal design techniques were used to decide on the pattern of metering. As such, the empirical results are able to provide a very detailed and accurate picture of how total residential load is disaggregated by end uses. Significantly, the consumption of high penetration end uses such as lighting, which cannot be estimated by conventional conditional demand analysis, has been successfully estimated. Also, by matching our estimates of end-use load curves with some recent prices paid by distributors to purchase electricity from an electricity market pool, we have been able to determine the costs to distributors associated with servicing individual end uses.