Best Fit Forecasting is a method that compares different models (which are maintained in a specific Best Fit Forecast profile) against the data in the system and ranks the various models based on the forecast errors of the outputs.
Using the standard IBP functionality, the Demand Planner can understand which algorithm has been chosen by the system while generating the forecast by navigating to application logs and filter the combinations and analyze the results. This can be done via Excel UI or Web UI.
Below are the configurations need to achieve the above requirement.
- Create an attribute “Best-Fit Model” and assign it to Product location MDT which is used in the segmentation process. The segmentation results will be stored in this attribute (Picture 1.2)
- Add the above-created attribute in the planning Area and assign it to the planning levels
- Create key figures to store individual forecast algorithms output values.(Only Stored key figures) Example: Create Key figure- Single exponential forecast to store the values generated by the single exponential forecast algorithm in best fit model
- Assign above created key figures (Step-3) in the forecast model. (Shown in the below picture 1.3)
- Create a calculated key figure as shown in picture 1.4. The key figure definition in this key figure is as follows.
- If Statistical forecast is equal to single-exponential forecast, then populate constant value 1 else validate the statistical forecast with double exponential forecast and if both values are same then populate constant value 2 and complete the key figure definition (if statement) with all loops as shown in below picture 1.4
- This key figure is used as segmentaion measure in ABC segmentation profile
6. Create ABC segmentation rules as shown in the below picture 1.5.a
- Maintain the attribute created in Step-1 as Target attribute to save segmentation results
- Maintain segmentation measure select the key figure which is created in step5
- Maintain calculation time scope as a current bucket. Select the future option and maintain the calculation horizon as 1.
7. Choose “(5) By Segmentation Measure (Single Values)” as segmentation method segmentation profile.
- This method compares one-by-one the predefined thresholds
8.Maintain the ABC segments as shown in picture 1.5.b
Note: Step-5, Step-6, and Step-8 should be in sync
- Create an Application Job Template that has both statistical forecasting as step1 and ABC segmentation as step2.
- Navigate to the Planning view to see the values in the “Best Fit Model” attribute after successful completion above job
In the planning view, a demand planner can view the “Best-Fit” forecasting algorithm chosen by the system while generating the forecast. This also enables the planner to analyze, react, respond quickly.