Forecasting using SAS Forecast Server, SAS/ETS, and more

Customize Disaggregation of Overrides

New Contributor
Posts: 3

Customize Disaggregation of Overrides

Hi There,

I have an interesting challenge that I think I'll have to write customized coding for. Thought I'd post it on this forum to get some statistics and expert ideas...

I have a data set of monthly volumes of a product by province, region, & then city.
I use Forecast Studio to create projects and use the rules of hierarchal forecasting in the tool for a bottom up reconcilation.

The timeseries trends show growth, however I still enter in an provincial override to almost 100% inflate that growth for the longterm forecast. FS disaggregates the overide by proportions of the reconciled forecast. My challenge here is I'd like to decrease those proportions and distribute out the disaggregations for the long term overrides more evenly.
For example if one product has 10% of the distributed override, then for next override, I'd like that allocate only 8% of the override, and then next month 5% etc etc. Since the granular timeseries trends are increasing, the way FS is disaggregating the override does not reflect this scenario.

Is there a way to code this situation or control/change the distribution of how the override is allocated in FS (other then the two options of: based off of proportions or evenly distributed) ?

Thank you kindly!
SAS Employee
Posts: 416

Re: Customize Disaggregation of Overrides

Posted in reply to SonyaApple
Hello -
In my opinion this cannot be archived with Forecast Studio today, if you want to use the existing overrides functionality (which is based on the 2 options you are outlining).
However, if you switch to SAS coding you should be able to implement the approach you are talking about, which reminds me of disaggregating daily forecasts to hourly forecast by using a day profile (low demand at start of day, high demand at noon, etc).
I might be completely missing your point - so maybe you could provide an example of what you would like to archive?
Ask a Question
Discussion stats
  • 1 reply
  • 1 like
  • 2 in conversation