Entering content frame

Process documentation Statistical Error Analysis Locate the document in its SAP Library structure


Statistical error analysis is a technqiue used in forecast accuracy reporting. A series of previous forecasts for a particular period is stored and each deviation of this series is compared to the actuals for the same period. The deviation can then be projected into the future.

This visualization of expected deviation not only places the forecast in context; it also enables "what if" planning. For example, it enables your company to evaluate future scenarios for maximum, planned and minimum production volumes, stock, sales revenue, and cash flow. The real impact of forecast accuracy reporting is realized when coupled with a simulation at the significant nodes of the supply chain, in particular, for S&OP.


The analyst must have a comprehensive understanding of the both the historical and projected business environment including market behavior, competitor activity, the product, product life cycle, and activities relating to pricing, promotion and distribution.

In addition the analyst should understand the business constraints relating to supply, production and distribution. This comprehensive understanding will enhance the quality of the corrected historical data and ensure that projections are kept in context.

  1. You have stored the forecasts from previous periods. See also Forecast Storage.
  2. In Demand Planning, use macros for this purpose. For example, you might have one macro that stores for a given month what was forecasted in the prior month, a second macro that stores for a given month what was forecasted two months prior to that month, and a third macro that stores for a given month the sum of three months' forecasts ending with the forecast for that month. The following forecasts would be seen in the month of March:







    March's forecast for March

    Prev Month

    December's forecast for January

    January's forecast for February

    February's forecast for March

    -2 Month

    November's forecast for January

    December's forecast for February

    January's forecast for March

    3 Month Rolling

    November’s forecast for

    December’s forecast for Dec+Jan+Feb

    January’s forecast for Jan+Feb+Mar

    Include this information in your data view, if you wish to use it while forecasting. Otherwise, create a separate data view.

  3. You have updated your historical data by loading actuals for the period(s) just completed.
  4. You have corrected history such that outliers are suppressed, previous promotion impacts removed, and business issues like previous stock transfers, refurbishments and so on, are not counted as sales.

Process Flow

  1. Review the corrected history to validate that it makes sense in the business context.
  1. Review the business dynamics and decide which data range is relevant.

For instance, if the cumulative procurement lead time for the significant components is only one month, it may not make sense to include errors generated by forecasts prior to one month, as strictly they are irrelevant from a procurement standpoint. In the table below, the errors from months (-3) and (-2) should therefore not be included in any statistical average calculation as they would make the forecast error worse. This would lead to unnecessary contingency in the forecast projections.

In this example, a product has three original forecasts for the month of November. The actual consumption for the month of November was 150 units.


Units forecast for November

Actual sales

% Error compared to actual sales

August (-3)




September (-2)




October (-1)




November (1)





  1. Run standard macros to calculate the forecast errors.

The % error for each of forecast is calculated as the percentage difference between the forecast and the actual. A positive error indicates an over-forecast while a negative error indicates an under-forecast.


An understanding by your company not only of the accuracy of the forecast, but also of how to provide for the variance.


The magnitude of the forecast error will depend greatly on industry, product complexity and market dynamics. Forecast errors by themselves reflect not just the quality of the forecast, but also the volatility of the business dynamics.

Leaving content frame