Ex-Post Forecast in IBP

[miaguo]

Dear All,

Good day.

In sap IBP, ex-post forecast is calculated and leveraged if MAPE is used for algorithm choice among them in one forecast module.

The ex-post value is important, may I ask how the ex-post got calculated, especially for those beginning/initial period which has no more historical periods exist before.

Thanks a lot.

Best Regards,

Mia

[LauraTozzo]

Hi Mia,

ex-post forecast can be calculated differently depending on the algorithm.
Here you can find some general information about the ex-post forecast:
https://help.sap.com/docs/SAP_INTEGRATED_BUSINESS_PLANNING/feae3cea3cc549aaa9d9de7d363a83e6/8c481b3a...

and here you can find some details on the settings for the error measure:

https://help.sap.com/docs/SAP_INTEGRATED_BUSINESS_PLANNING/feae3cea3cc549aaa9d9de7d363a83e6/d9b213f8...

In general, the ex-post forecast is calculated for past periods for which actual demand history is also available. The system calculates the forecast accuracy measurements by comparing the differences between the actual values and the ex-post values.

Best regards

Laura

[miaguo]

Hi Laura,

Thanks a lot for your feedback. That's really helpful.

And regarding the ex-post calculation, if it use the whole historical period set in forecast model, let's say 36 month. Then the first period in system will be Aug 2019. So the Ex-post Forecast will be calculated leverage 36 historical period data as full set from Aug2019, instead of calculated from beginning period using first month data until to last period in time order?

In system, we got a testing data like below, Auto ARIMA was selected by system as best fit. but how to explain why we got the ex-post for July2019, if it will be included into the MAPE calculation.

Thanks a ton.

Best Regards,

Mia Guo

sample.png

[LauraTozzo]

Different forecast algorithms might have different initialization periods, meaning they can start creating ex-post forecast in different time buckets. This can impact the error calculated for the ex-post forecast. In IBP, there are different options to come over this issue.

1. Using Error Measure Calculation Horizon on the General tab of the Forecast Model, one can set how many of the latest historical time periods should be considered in error measure calculation, making sure that error is measured based on identical amount of time periods for all algorithms
2. Using Test Phase Periods on the Forecasting Steps tab of the Forecast Model, you can eliminate the difference in accuracy between different forecast models. Some algorithms have more accurate ex-post forecast then others, but it does not necessarily mean that their forecast will also be more accurate. For example, the exponential smoothing type of algorithms, especially in case of high smoothing coefficient, tend to calculate very accurate ex-post forecast. To work with the forecast accuracy instead of that of the ex-post forecast, you can specify the number of test phase periods. IBP will use preserve this number of the latest historical time periods for testing purpose, and let each algorithm calculate forecast for this period from the remainder of the history horizon. The calculated forecasts are compared with the actuals, and the most accurate algorithm is selected. Please note that this method should be avoided if there are significant changes in the sales history, falling into test phase periods. Such changes can be detected with Forecast Automation.

You can read more about this topic here: https://help.sap.com/docs/SAP_INTEGRATED_BUSINESS_PLANNING/feae3cea3cc549aaa9d9de7d363a83e6/65b1da1c...