Useful Notes on Tracking Signal Forecasting Error

Automatic monitoring mechanism to keep a track of accuracy of forecasting models is known as tracking signal. It is the ratio of cumulative forecasting error to the mean absolute deviation (MAD). Cumulative forecasting error or running sum of forecasting error (RSFE) is indicated by:

ⁿ ∑ (A_t -F_t), where, A_t = Actual demand for period ‘t’ and F_t = Forecasted demand for period ‘t’

Tracking Signal is recalculated every time with the availability of actual demand data and forecasting is updated accordingly. TS should be fairly small and then only one can say that the forecasting is accurate or nearly accurate. If a TS moves outside some pre-established range, it is called tripping of signal.

Every time the signal trips, forecasting method is adjusted. Usually, the limit of the TS is taken to vary between 3 and 8. Let us now illustrate TS with an example of actual demand data, forecasted demand data and the resultant value of TS. Let us assume that up to period 5, we have made an initial estimate of the MAD to test accuracy of our forecasting model. For each subsequent period, we have computed TS value to understand, however, whether it is within the accepted range of 3 to 8.

In this case, the TS is less than the limit of 3 to 8 and is up to period 9. In periods 10 and 11, it undershoots the limit. As of now, the forecasting model is considered to be functioning well. However, if the trend persists even in subsequent periods, the forecasting model may have to be reassessed. Using a tracking control chart, we can plot the data to explain the situation.