Copper losses in low voltage distribution circuits are a significant proportion of total energy losses and contribute to higher customer costs and carbon emissions. These losses can be evaluated using network models with customer demand data. This paper considers the under-estimation of copper losses when the spiky characteristics of real customer demands are smoothed by arithmetic mean averaging. This is investigated through simulation and by analysis of measured data. The mean losses in cables and equipment supplying a single dwelling estimated from half-hourly data were found to have significant errors of 40%, compared to calculations using high resolution data. Similar errors were found in estimates of peak thermal loading over a half-hour period, with significant variation between results for each customer. The errors reduce as the demand is aggregated, with mean losses for a group of 22 dwellings under-estimated by 7% using half-hourly data. This paper investigates the relationship between the demand data time resolution and errors in the estimated losses. Recommendations are then provided for the time resolution to be used in future measurements and simulation studies. A linear extrapolation technique is also presented whereby errors due to the use of averaged demand data can be reduced.