How to Model and Forecast Multiple Product Types i...

Former Member · ‎01-06-2016

In this CAR Tip in a Minute, I will illustrate how you can model and forecast the demand for multiple product types in the same client at the same time. For example, you might have categories in your assortment that include both never-out-of-stock products (NOS products) and short life cycle products (such as fashion). As of SAP Customer Activity Repository 2.0 FP2, you can configure this product type at the level of the individual product location, using different product-type-specific demand model definitions. The model definition for a specific product location is determined based on the material requirements planning type (MRP type) that is assigned to it in the master data system.

Choosing a Demand Model Definition

SAP delivers several model definitions as seed data. By default, modeling and forecasting will use the RTL model definition. RTL is a good choice for products available all year round (NOS products). For products that are available only for a short period (such as fashion products), the RTLSDP model definition might provide better results.

Example

You want to model and forecast future sales for a mix of products available all year round (such as milk and bread) and products available for only a short period (such as fashion products).
All products have the same time series source (in this example, point-of-sale data, POS_TS).
The MRP type of fashion products is ND, which allows you to distinguish them from other products.
Your goal is to use model definition RTLSDP for the fashion products and RTL for the NOS products.

Prerequisites

You have installed SAP Customer Activity Repository 2.0 FP2. For more information, see SAP Help Portal at http://help.sap.com/retail-car200fp2.
In particular, you have fully installed and implemented the following modules in SAP Customer Activity Repository:
- Unified Demand Forecast (UDF): modeling and forecasting engine
- Demand Data Foundation (DDF): required data layer
You have the authorization to run transaction SPRO and do changes in Customizing.

Procedure

Step 1: Create a New Model Category for Fashion Products

In transaction SPRO, execute the Define Model Categories activity under Cross-Application Components -> Demand Data Foundation -> Modeling and Forecasting.
Choose New Entries and add the following entry:
Save your changes.

Step 2: Assign the Fashion Products to the New Model Category

In transaction SPRO, execute the Assign Material Requirements Planning Types to Model Categories activity under Cross-Application Components -> Demand Data Foundation -> Modeling and Forecasting.
Choose New Entries and add the following entry:
Save your changes.

Note: The system assumes that all MRP types without a specific assignment to a model category belong to UMC (Unspecified Model Category). As a result, the fashion products are now assigned to ZSM, whereas all other products remain by default assigned to UMC.

Step 3: Assign the Demand Model Definition for the New Model Category

In transaction SPRO, execute the Assign Demand Model Definitions Customizing activity under Cross-Application Components -> Demand Data Foundation -> Modeling and Forecasting.
You will notice that model category UMC is already assigned to model definition RTL. You don’t need to do anything extra for the NOS products.
For the fashion products, choose New Entries and add the following entry:
Save your changes. That’s it!

Result

The next time you run modeling in production mode (that is, with a blank diagnostic ID) for time series source POS_TS, the system will use RTLSDP for fashion products and RTLfor NOS products.

Additional Notes

The MRP type is a two-character field that is set when you import the products from your master data system to your SAP Customer Activity Repository system.
Having a model category allows you to map multiple MRP types to a single model category, which simplifies the configuration in Step 3.
This example is for the POS_TS time series source. If you would like to change the model definition for other time series sources, you can make these assignments in Step 3.
This example is for the production mode (blank diagnostic ID). You can also specify a diagnostic ID, thus activating the diagnostic mode, and set up different model definitions this way.