Start of Content Area

Function documentation Forecasting  Locate the document in its SAP Library structure


Use a forecast procedure to predict the future development of key figures. The forecast function uses historic data to calculate expected forecast values with statistical procedures.

This involves simple univariate forecast procedures.  


The forecast function has been newly implemented in BW-BPS. In standard cases the previous forecast function used within the bounds of SEM-BPS is no longer available. For compatibility reasons the system still provides the old function if you were using it within SEM-BPS. 

SAP recommends you use the newly implemented forecast function in BW-BPS for new developments. This explicitly enhances the functionality of the old (SEM-BPS) forecast function, even though it does not cover all aspects of it. (For example, you do not obtain forecasts according to Croston’s algorithm). For more information on the old (SEM-BPS) forecast function, see Forecasting (Old).


        Historic data that can serve as reference data for the forecast must be available.

        The planning level, in the context of which you create a forecast function, must contain at least one characteristic with a time reference (for example, fiscal year).

        The forecast period is copied from the selection in the planning package or planning level and therefore must be restricted there.


Forecast strategies

The following forecast strategies are available to you for performing calculations:

        Automatic model selection


        Moving average

        Weighted moving average

        Simple exponential smoothing (constant model)

        Linear exponential smoothing (trend model)

        Seasonal exponential smoothing (seasonal model)

        Trend-seasonal exponential smoothing (multiplicative seasonal component)

        Trend-seasonal exponential smoothing (additive seasonal component)

        Linear regression

The automatic model selection forecast strategy allows you to let the system select the forecast model that best fits the trend of the historic data (see Automatic Model Selection). 

If you already know that a particular forecast model is well matched to the time series trend, or if you explicitly want to use a forecast model for other reasons, you can select a particular forecast model (see Forecast Strategies).

Optional functions for the forecast strategies

The forecast strategies offer the following additional functions and possibilities:

        Outlier correction

        Setting negative forecast values to zero

        Logging statistical key figures

        Ignoring initial zeros

For exponential smoothing

        Optimization of smoothing factors for exponential smoothing

For forecast models with trend components:

        Trend dampening


To create a planning function of type Forecast, the following steps are necessary:


       1.      Creating a planning function of type Forecast

Time characteristic

A time characteristic is required for each forecast function in order to represent the time dimension of the forecast.

        The selected time characteristic is included automatically in the set of reference characteristics.

        The characteristic has to be restricted for the package. Values for this period are forecast.

Fields for reference data

The characteristics selected as reference data serve to determine the historic data on the basis of which forecast values are to be calculated.

Fields for conditions

You cannot include the selected time characteristic in the set of characteristics for conditions.

For more information on fields for conditions, see Planning Functions.

       2.      Creating parameter groups

Select a forecast strategy on the Forecast Parameter tab page. This determines how forecast values are to be calculated. Certain parameters have to be entered depending on the forecast strategy.


The system proposes Automatic Model Selection as the default forecast strategy and, in comparison to other forecast strategies, makes available the greatest number of parameters. Note that calculation is accordingly time-consuming.

On the Reference Data tab page, determine the period in the past on which the forecast is to be based.


Make sure to exclude non-posted special periods from the past period as otherwise these periods go into the forecast calculation with the value 0.

       3.      Executing the planning function

In comparison to the Execute Planning Function option, the Execute with Trace option is more time-consuming. As well as the tab page Log, the system also adds the tab page Trace.

        You can display the log for each data object for which the forecast was generated by clicking on This graphic is explained in the accompanying text in the Messages column.

        You can display the reference data and the values of the processed data objects before and after execution by clicking on This graphic is explained in the accompanying text in the Detail column.



End of Content Area