The effect of varying the averaging time of measured data used to calculate wind turbine power curves is examined. The effects of reducing the averaging time from 10 to 1 min, as recommended for small wind turbines, are investigated using power performance data recorded using a 15 kW wind turbine. Test site data have been processed according to the relevant international standard, IEC 61400-12-1, to provide power curves and annual energy yield predictions. A number of issues are explored: the systematic distortion of the power curve that occurs as averaging time is decreased, the errors introduced by the use of 1 min averaged power curves to calculate energy yield and the reduction of turbulence intensity as averaging time is reduced. Recommendations for improved small wind turbine testing and energy yield calculation are given. Copyright © 2012 John Wiley & Sons, Ltd.