Forecast Not Generated For Item/Store Where Sporadic Sales is Expected/Normal As Automatic Outlier Calculation Does Not Handle These Exceptions (Doc ID 3049776.1)

Last updated on SEPTEMBER 25, 2024

Applies to:

Symptoms

Its not possible to generate a forecast for the item/store where the automatic outlier calculation has marked all sales periods as if they were outliers. The multiplier for automatic outlier calculation (PP_LLC_OUTLIER_THRESHOLD) in AI Foundation (AIF) only has a global value. However, there are situations in which it is necessary to treat this parameter at the exception level, either by changing the multiplier or even deactivating it in a specific case.

With an item/store with sporadic sales, and some weeks with limited sales and many weeks with zero sales, the automatic outlier calculation marked all sales periods as if they were outliers, when they are not - they are normal sales. As a result, the pre-processed sales were all zero and did not generate a forecast for the item/store.

For this item/store, it is necessary to deactivate the outlier calculation, or register an extremely high multiplier so that those specific sales are not marked as outliers. In addition to allowing the exception control explained above, it is also important that this parameterization is available to the analyst within Retail Demand Forecasting Cloud Service (RDF), not just in AI Foundation (AIF).

Cause