How You Configure Disaggregation of Statistical Forecasts

This topic explains how you can configure the disaggregation of statistical forecasts in the Demand Management, Demand and Supply Planning, and Replenishment Planning work areas.

Instead of having the forecast disaggregated by default to active product combinations at the lowest level, you can use a measure to disaggregate the forecast to product combinations of your choice. Thereby, you can override the default logic of using the forecast status measure that corresponds to the output measure of the selected forecasting profile for identifying active product combinations that receive the disaggregated forecast.

Configuring disaggregation of statistical forecasts is useful when your organization has a product that's frequently updated, but only the latest version is available for ordering. You can generate the forecast using the historical demand for old product versions that are currently available in the market and disaggregate the forecast to only the latest product version.

Note: You can use this feature in only a user-defined forecasting profile that's based on Bayesian machine learning.

Configure Disaggregation of Statistical Forecasts

Follow these high-level steps to configure the disaggregation of statistical forecasts:

1. Create a measure as follows:

   • The measure should be of the Number data type.

   • The measure should have all dimensions of the output measure of your user-defined forecasting profile except for the Time dimension.

     Any dependent measures of the created measure must also have the supported dimensions.

   • The measure expression should return 1 for product combinations that should receive the disaggregated forecast. For example, the expression could be as follows: Case(Lifecycle Phase When 'Active' Then 1 Else 0)

     You can also configure the expression using Item-level attributes.

     Preferably, the measure expression must not use these functions:

     • Archive

     • Bias

     • Fpos

     • Fsum

     • Lag

     • Lead

     • Log

     • Mad

     • Mape

     • MeasureOffset

     • Moving_Maximum

     • Moving_Minimum

     • Moving_Total

     • Pi

     • Power

     • Rand

   • You can populate the measure through the regular collections process or the file-based data import (FBDI) template for Supply Chain Planning measures.

2. Add the measure to the predefined Output Filters for Forecast Disaggregation measure group.

3. Add the measure and its dependent measures to the measure catalog of your plan.

   If you're using an Item-level attribute, ensure that it's present in the dimension catalog of your plan.

4. Add the ForecastOutputFilter forecasting parameter from the General Settings category to your forecasting profile.

5. Select the measure that you created for the forecasting parameter.

   Only these measures are available for the forecasting parameter:

   • The measures have the Number data type.

   • The measures have all the dimensions of the forecasting profile's output measure except for the Time dimension.

   • The measures are in the Output Filters for Forecast Disaggregation measure group.

When the plan is run, the forecast status measure corresponding to the output measure of the forecasting profile is selected for the forecasting parameter, forecast disaggregation occurs according to the default logic, and an error message is displayed on the Review Plan Messages page in these situations:

• The measure you selected for the forecasting parameter isn't in the plan's measure catalog.

• The dimensions of the dependent measures of the measure you selected for the forecasting parameter aren't supported.

• The Item-level attribute used in the measure you selected for the forecasting parameter isn't in the plan's dimension catalog.