For any spatial estimation IDW often performs better than other spatial interpolation techniques. The reason lies behind the way it models spatial variability. Kriging and even thin plate spline methods perform smoothing while modelling spatial variability whereas IDW models it by fitting straight lines through the observations. Therefore the estimated values are more similar close to the observed location and get decreased by the distance weights as the estimated locations get further away. This phenomenon is known as 'Bull's eye effect'. The observations likely come with observation errors that can be reduced by performing smoothing in spatial variability modelling. So, though IDW performs better in case of satisfactory number of observations it's not necessarily good because it incorporates the observation errors in estimation.