By
now you it should be somewhat apparent that exponential smoothing techniques
tend to lag behind the turning points of the actual time series data. The usefulness of the forecast is best
determined by evaluating its associated error.
Of course, smaller error values are better than larger errors when
comparing different values of the smoothing parameters. The same holds when comparing different
forecasting models. ORS provides
several useful measures of forecast error.
Choosing which error term to focus on is part of the 'art' of
forecasting. Generally, in business and
economics the focus shifts to MAD (or MAPE) and MSE.