Monitoring a Forecasting System

sistently lag; or it may be that the existence of unsuspected seasonal variations is causing biased errors. It is desirable to sense such situations as quickly as possible so that a more appropriate forecasting model may be introduced. One method of monitoring a forecasting system which has been examined in detail by Brown1 is to compute a "tracking signal" which is defined as the sum of the forecasting errors divided by the Mean Absolute Deviation. The latter quantity, which is simply a convenient measure of the noise in the system, is obtained by a simple smoothing process upon the absolute forecasting errors.