What is a Good Forecast?:

If the real demand during a given time period matches with the forecast, we may say, the forecast was good. In other terms, if the forecast error is minimum, we say the forecast was good and the forecast technique that provides such a forecast is appropriate. However there are several ways of measuring error, like average or mean deviation, sum of deviations, absolute deviation, standard deviation of errors, and percentage of demand variation determined by a functional equation. It can be essential to consider any one or more of such measurements when selecting the best forecasting technique.

An acceptable test of a forecasting technique is to try it out on past data. For this it is significant that reliable past data is available. If the forecast technique has any basis on past data and if the data are available, then it is a good practice to develop forecast on the basis of the first half of available historical data and test it on the second half. A major problem often faced is that we have past records of sales or shipments rather than demand. The past data provides figures of not what was actually demanded, but what was shipped. That is, the data about lost demand because of product unavailability, delayed shipments, or substituted product is more often not available. There is thus lack of correspondence of shipments with demand.