Validity, Measurement, and Control of Forecast Errors:

Every user of forecast is concerned with the reliability of its process. But everybody understands that no procedure can do any better than utilize past and current information to estimate the future. A simple but effective method of checking the validity of a forecasting model is to apply it on past data and discover how well it would have predicted the present. A good fit to historical data does not, of course, guarantee accuracy in futuristic forecast, but a bad fit should definitely be taken as a warning signal.

All forecasts should also state the estimate of the expected error. The most commonly used measures of accuracy (or expected error) are the mean squared error (MSE), the mean absolute deviation (MAD), and the tracking signal.