Automatic Model Evaluation The system uses the following methods to analyze forecast performance:
The system uses Trigg’s tracking signal to establish whether a forecast model is still valid, or if it would be better to choose a new one. To do so, the system looks for non systematic forecast errors.
A non systematic forecast error occurs if the calculated forecast values are incorrect without a recognizable pattern. For example, if the demand forecast for a product for the month of May is much too high, and much too low for the month of June, this is a non systematic forecast error.
The system uses tripping to analyze the performance of an existing model and to initialize the base values and the corresponding deviation. To do so, the system looks for systematic forecast errors.
A systematic error occurs if a specific forecast error occurs time and time again. For example, if the demand of a product has quadrupled since June 2004 yet the system still calculates the forecast on the basis of the demand history since 2000, the demand forecast will be regularly and permanently incorrect. This is a systematic forecast error.
The system applies Trigg’s tracking signal only to demand forecast, whereas it can apply tripping to demand, order items, and the average demand size per order item (demand/item).
You can schedule Trigg’s tracking signal in the forecast service profile as a regular planning service of the Planning Service Manager (PSM). In this case, Trigg's tracking signal sets the
Tracking Threshold: Date Exceeded
date on the
Interactive Forecasting
screen.
The Trigg’s tracking signal function is also integrated in the
Perform Combined Forecast
service. You can also schedule this as a regular planning service in the forecast service profile. If you schedule this combined service in the PSM, the system checks certain stability rules and, if necessary, triggers
automatic model selection
after setting the exceed date.
You can schedule tripping in the forecast service profile as a regular PSM planning service or as part of the
Perform Combined Forecast
planning service.
For more information about the forecast service profile, see the IMG for
Advanced Planning and Optimization
under
.
For more information about the PSM, see Use of the Planning Service Manager in SPP and PSM Services for Forecasting .