statistical forecasting

[Former Member]

I am running a statistical forecast on a group of materials using a composite forecast.  The selection I am using contains Material (numerous) - Sales Division (one div) - Sales Organization (one sales org).  It appears that running in the foreground or directly out of the data view is creating the stat forecast at the group level and disaggregating down to the materials. This is exactly what I want it to do.

I then run this same selection in the background with an aggregation level of Material - Sales Division - Sales Org.  It appears when running it this way that it is calculating the Stat forecast for each Material rather than as a group.

When I do the same background run but only at the Sales Division - Sales Org level I am not really sure what is going on.  When I look at the spool the MAPE that is calculated is the same regardless of what the selection is that I am using.

Does anybody know what is actually happening behind the scenes when running the stat in these three different ways?

[andrewkeng]

#1 - When you run stat. forecast interactively, the planning objects you selected on the shuffler has the materials, division and sales org. and this selection will be run as a group.  The forecast results will then be disaggregated to the lowest level.

#2 - When you run stat. forecast as background job, you have the option to choose aggregation level.  Even if you apply the same forecast model to the the same selection, the number of planning objects may not be the same as in #1. Forecast results will be the same, but not disaggregated to the lowest level.

#3 - Running stat. forecast in background with a different selection of planning objects than #2.  And again not sure your aggregation level was.  Chances are your forecast results and MAPE error would be different altogether.

So, you probably want to check in your planning job if Aggregation Level is at the lowest level of your CVC to make it comparable with the online transaction.