Trigg’s Tracking Signal

Trigg’s tracking signal checks whether the chosen forecast model is suitable for a certain location product, or whether it would be better to choose a new forecast model. To do so, it compares the results of one of the previous forecasts with the actual demand and checks whether a forecast error has exceeded a certain limit value too often. In this case, Trigg's tracking signal sets the Tracking Threshold: Date Exceeded date on the Interactive Forecasting screen. If you have scheduled the Perform Combined Forecast planning service in which Trigg’s tracking signal is integrated, the system firstly sets the exceed date and then checks certain stability criteria before triggering automatic model selection.

Features

Forecast for Model Evaluation

For each product, you can define which past forecast results the system should choose to compare with the demand. You do this via the SAP Easy Access menu under Start of the navigation path Advanced Planning and Optimization Next navigation step Service Parts Planning (SPP) Planning Forecasting Forecast Profile End of the navigation path on the General tab page in the Offset Forecast Creation field.

Check Current Forecast Model

Trigg’s tracking signal calculates the quotient of the smoothed sum of absolute deviation (SAD) and the smoothed mean absolute deviation (MAD).
- The system calculates the smoothed sum of absolute deviation according to the following formula:
  - You can specify alpha in the Smoothing Factor for Tracking parameter in the forecast profile on the Model Evaluation tab page.
  - SAD_(t-1) for the initial period is calculated as follows:
    SAD_initial =
    
    WeightFactor1 * (DHist_{curr.per. - tracking horizon - per. 3} — forecast_{curr.per. - tracking horizon - per. 3}) +
    
    WeightFactor2 * (DHist_{curr.per. - tracking horizon - per. 2} — forecast_{curr.per. - tracking horizon - per. 2}) +
    
    WeightFactor3 * (DHist_{curr.per. - tracking horizon - per. 1} — forecast_{curr.per. - tracking horizon - per. 2})
    
    You can define the weighting factors for the individual periods in Customizing under Initialize Base Values. For more information, see the Implementation Guide (IMG) for Advanced Planning and Optimization under Supply Chain Planning Service Parts Planning (SPP) Forecasting Initialize Base Values .
    
    You can define the tracking horizon in the forecast profile on the Model Evaluation tab page.
- The system calculates the mean absolute deviation (MAD) according to the following formula:
  - You can specify alpha in the Smoothing Factor for Tracking parameter in the forecast profile on the Model Evaluation tab page.
  - MAD_(t-1) for the initial period is calculated as follows:
    MAD_initial =
    
    WeightFactor1 * I(DHist_{curr.per. - tracking horizon - per. 3} — forecast_{curr.per. - tracking horizon - per. 3})I +
    
    WeightFactor2 * I(DHist_{curr.per. - tracking horizon - per. 2} — forecast_{curr.per. - tracking horizon - per. 2})I +
    
    WeightFactor3 * I(DHist_{curr.per. - tracking horizon - per. 1} — forecast_{curr.per. - tracking horizon - per. 1})I
If the quotient is greater than the tracking threshold value, the system sets a tracking signal. You can enter the tracking threshold value in the forecast profile on the Model Evaluation tab page in the Tracking Threshold parameter.
If the system sets x tracking signals in y periods, it is a sign that the quality of the current forecast model for this product is not ideal, and that another model should be chosen.
You can enter x in the Tracking Signals in Horizon parameter and y in the Tracking Horizon parameter in the forecast profile on the Model Evaluation tab page.

Reset Counter

If the system has set x tracking signals in y periods, it sets the Tracking Period Counter and the Tracking Signal Counter parameters in the forecast profile on the Model Evaluation tab page to zero. It also sets the date value in the Tracking Threshold: Date Exceeded parameter on the Model Evaluation tab page of the forecast profile to the date on which the system determined that x tracking signals occurred in y periods. The system also sets the date of the Last Tracking Run to today's date.