Skip to Content
SAP Integrated Business Planning for Supply Chain

Statistical Forcasting in SAP Integrated Business Planning for demand FAQ

During our “IBP - Statistical Forecasting” Special Topic Webinar on February 17, 2016, there were a lot good questions in the chat. Not all of them could be answered during the meeting. We therefore used the questions in the chat to create this FAQ document.

“SAP Integrated Business Planning Webinar –SPECIAL TOPIC: Statistical Forecasting

Recording Link:

Slides Link

Topics Covered:

  • Statistical Forecast Models in SAP IBP for demand
  • Statistical Forecast Models in SAP IBP for sales and operations
  • Review which Statistical Models are included in each SAP IBP application
  • Demo
  • Q&A

1. What is the new way of running the forecast and how it changed from 4.0 to 5.0 where before you ran the forecast at the base level of the target KF, and now you are able to run it at any level of aggregation from the target KF? Does that suggest that the target KF for the forecast is at the same level as the source KF (i.e. history)? 

Source and target key figure do not need to be on the same base planning level. Actually you need to define the level on which you want to forecast. During forecasting run the source key figure is aggregated and target key figure disaggregated according to the rules defined for these key figures.

2. Is it possible to execute Statistical Forecasting also in interactive mode/simulation mode via the Microsoft Excel Add-In? E.g. for What-If Analysis

Yes. You can execute statistical forecasting models also in simulation mode via the IBP Excel Add-In

3. Is it possible to execute Statistical Forecasting on multiple planning levels?
Yes. The user can for example freely select the planning level in his excel planning view and execute the forecasts based on this. The system checks that the attributes of the freely selected planning level (in the planning view) are contained in the base planning level of (at least one of) the output key figures (such as the statistical forecast KF, corrected history KF) of your forecast model.Only such forecast models can be selected that meet this criteria. In other words basically the system checks that the result can be stored correctly (can be disaggregated to the base planning level of the output key figure).

4. How does the user define a time series with a multiple level hierarchy?
This is done via configuration of the planning model.

5. Is it possible to analyze at which level the statistical forecast model should run in an ideal case? What about Forecast Level Optimizer (FLO)?
Currently not possible. This is a roadmap topic.

6. When scheduling demand jobs in the front end is it possible to specify successor/predecessor, replicating process chains functionality?   
Currently not possible. This is a roadmap topic.

7. Is it possible to execute the statistical forecast algorithms based on customized / customer specific key figures or do we have to use a standard  key figure (e.g. like APODPANT in APO)?  
You freely define the input and output key figures that you want to use. The key figures are not hard coded. You define which ones you want to use.

8. How often does IBP data need to be updated? In a weekly or daily basis?

That depends how often you want to run the statistical forecasting. It makes sense to update the sales history (e.g. from an ECC system) prior to that

9. How can I automize the statistical forecasting runs?
The statistical forecast runs can be scheduled regularly (e.g. weekly or monthly at a specific time) via IBP’s Microsoft Excel Add-In. In future, it is planned to further automize this process and also to provide more visibility across all scheduled jobs (roadmap topic).

10. Is an automatic forecast selection model available in the SAP IBP application for the Statistical Forecasting Models?
Yes. You can set up the forecast model in a way to pick the best algorithm based on the result of a certain error measure (e.g. MASE).

11. How can we know which forecast model has been chosen by the automatic forecast model selector for the products?

The forecast model selected by pick-best shows in the forecast log

12. What happens when MAPE calculation is executed and no values are found? In APO-DP as MAPE is a kind of percentage the jobs aborted. What would be result if no history is available in IBP?   

The job processes all selected planning combinations. If some cannot be forecasted the job does not fail.

13. Is it possible to separate the history horizon for statistical forecast calculation from the horizon of the forecast error calculation? E.g. using 36 months history for calculation of the statistical forecast versus only showing the forecast error based on the last 12 month? 

Not at this time. The horizon is set at the forecast model level

14. Is Character based forecasting possible in IBP for demand?

Currently not possible

15. Will the automatic assignment setting for forecast models overwrite the filter in excel add-on when you run forecast?
There is currently no “automatic” forecast model assignment in IBP. The assignment is a manual process that needs to be handled prior to running the forecasting. When using the “Forecast Model Assignment” option, the filters in Excel are overwritten by this.

16. Is there a concept of proportion factors in IBP?

Yes, there is a concept of proportion factors in IBP. However, it is not possible to calculate the prop. factors in IBP at this point in time.

17. Is there any capability for leading indicator forecasting?
Leading indicators (e.g. Corporate IT spending’s could indicate future hardware sales) can be used in IBP by leveraging Multiple Linear Regression. However the time lag is not analyzed by IBP automatically, you need to add the leading indicator with multiple time lags (offsetted with various number of periods) so that the regression detects which lags is appropriate. For this you need to create additional key figures where the leading indicator is offsetted with various lags and these key figures should be added to the regression model.

10. Would the offset in the forecast model also affect the seasonality (e.g. shift a holiday like Christmas)?

19. Can we define a frozen horizon which is not overwritten by the new forecast?

Defining a frozen horizon is separate from the statistical forecast and would be done with IBP planning area configuration.

20. Will there be the option to forecast quantiles?
Currently not possible. This is a roadmap topic.

21. Can we define a certain horizon in which the KPIs are calculated or will they be calculated for the complete ex-post forecast?
At the moment the whole ex-post forecast is considered. Further enhancements are planned for this topic.

22. Will IBP Demand do both bottom up aggregation and top down disaggregation? Or, are you limited to choosing one?

Bottom up aggregation and top down disaggregation are supported.

23. How does the system do time based disaggregation? From months to weeks if the weeks cross over?

Disaggregation is a setting by key figure and part of the planning area setup. In that case, it is already taken care of. The “week-to-month split” is planned to be available with the next IBP release.

24. If we choose a forecasting level, then we need to ensure disaggregation is defined to base planning level...or does the system automatically do it?   

The system is aggregating and disaggregation the key figures automatically according to the rules defined for these key figures.

25. Is the disaggregation done according to the rule of the TARGET / output KF?


26. Is it possible to have Daily, Weekly and Monthly buckets in a single forecast model?

No. you have to specify the time bucket (day, weeks, months) per model. It’s not possible to have multiple levels in one model.

27. I want my Demand Planner to see only specific forecast models. Can I control that?

This scenario is supported by configuring the permissions. There is a dedicated authorization object for forecast models.

28. How does IBP support “segmentation"?

There is already ABC segmentation functionality available in IBP. More advanced ABC support is on the roadmap. In addition also XYZ is on the roadmap.

29. I thought there was a plan to use only one model / planning area for different needs, instead of separate like we have today. 

Correct. With SAP Integrated Business Planning for demand, we introduced another topic specific model (SAP6). However, we are planning to release also a new sample model (SAPIBP1) with the next IBP release. This new model will contain a unified planning including S&OP, Supply, Demand, and        Inventory.

30. “Lifecycle management phase-in phase-out” - Is product lifecycle planning supported as part of demand planning?

Basic lifecycle planning can be modelled in implementation projects and was already done by some customers. However, extended lifecycle planning functionality including phase in and phase out specifications and a connection to the forecasting runs is a high prio topic on the roadmap.

31. Is “Like Modelling” covered in IBP for demand?

Currently not. Like Modelling is also part of our Product Lifecycle Planning roadmap topic and is planned to be available in a later release.

32. How does the Statistical Forecasting Engine deal with new SKU's with mainly “pseudo demand”, e.g. false demand when a new SKU gets launched and only a week or so of actual sales data is already available?. Does it use both the pseudo and the new actuals data?

Via configuration in IBP, it is possible to come to a basic solution for such cases. However, this topic will also be addressed as part of the roadmap for “product lifecycle planning”.

33. How does IBP handle “Dead SKUs”? E.g. Materials that have no more sales but their history is required for roll-up. How are they handled?

You can leave the data in the system but exclude them from the forecasting. However, this topic will also be addressed as part of the roadmap for “product lifecycle planning”.

34. Is it possible to see alerts prepared in WebUI, in Excel Add-in Alert Dashboard?

No. These two forms of alerts are different.

35. What about data realignment?     
Data Realignment as you know it from APO DP is not part of IBP right now. It is also a topic on our roadmap.

36. What is the scope for an integration with open source statistics tools like "R"?

We are looking into this, but this is on the long-term roadmap at the moment. Currently, we are enhancing the statistical methods that are already available in the Predictive Analysis Library (PAL) in HANA, e.g. ARIMA is on the roadmap.

37. Are you considering ARIMA and other models that APO DP did not have and other tools have?

Including the ARIMA model is on the roadmap for IBP for demand.

38. Does IBP allow for data transformations as Box & Jenkins?

The planned ARIMA model that is on the IBP roadmap would do the data transformations to arrive to a stationary time series. See PAL documentation for  further details:       

39. Are the statistical algorithms the same for 1st, 2nd and 3rd order exponential smoothing and the APO Constant, Trend and Seasonal trend models? 

There are slight differences in the implementation as IBP leverages the algorithms from PAL.See PAL documentation for further details:       

40. For exponential smoothing models, what is the process for parameter determination (Alpha, Beta, and Gamma)?

There is a dedicated algorithm called “Automated Exponential Smoothing”, which allows for an automated parameter determination in IBP for demand.

41. How does the promotion elimination algorithm identify promotion lifts?

The algorithm uses a consumption logic to eliminate the uplifts

42. In case we don't want to eliminate the impact of promotion on our sales history, will there be any special configuration requirement for that?

No. In case you don’t want to use the promotion elimination, you can also use your data the way it is. You need to specifically configure the forecast model so that promotion elimination is executed.

43. Are you planning on developing group seasonlity functionality?

Currently not possible. This is a roadmap topic.

44. Will cell fixing be possible?

Currently not possible. This is a roadmap topic.

45. What is the difference in the offering between conventional APO-DP Stat Forecasting vs IBP Stat Forecasting?

Currently, pretty similar scope of both solutions. More advanced statistical methods planned for IBP demand.

46. What data volumes can forecasting accommodate?
For a statistical forecasting that is using time series data, there is no specific limit if proper sizing of the hardware is considered.

47. Is it possible to integrate to different ECC systems? Can you integrate through SAP PI/PO or HCI. Or would you need a normalized data set?

Integration is via HCI only. Basically it is possible to integrate multiple ERP system to IBP. If master data is not harmonized across the ERP systems, master data ID’s need to be mapped outside of IBP. If it’s required that order data is sent to IBP then in general the ERP order ID is not unique across the ERP systems and thus the corresponding data types of the IBP data model needs to be enhanced by e.g. the system ID of the source ERP system. The Response and Supply module of IBP which is new with IBP 6.1 currently only support one ERP system as source.

48. How is the forecast released to SNP? Can we maintain location split and period split for it?

The forecast is released to APO SNP via HCI. Location split and period split are not out-of-the-box supported, but can be modelled inside IBP if necessary using key figure disaggregation.

49. Will it be able to mass-export data and reload data back into IBP?

Functionality available with the HCI Integration

50. How is IBP integrated with BW? Are there plans to have an integrated BW cockpit like APO? Like APO-BW cockpit, e.g. for data loading or backups? 

IBP is not using any BW functionality. However, IBP can be integrated with a BW. Data loading between IBP and BW is covered by IBP and HCI functionality. Backups are handled by SAP Cloud Operations as IBP is a cloud solution.

51. Is there any planned integration with the Unified Demand Forecast (UDF) in CAR?

We are looking into this topic of UDF, but this is far out from a roadmap perspective.


Former Member

No comments