Predictive FCST - including only account/signal history or also other dimension values/variables

[axel_]

Hi Folks,

can anyone tell me if the predictive FCST feature (executed in a table) does use only the historical values of the key figure/account or will it consider also all other historical dimension values.

Basically trying to figure out if a predictive scenario where I explicitly define the dimension values as variables beside the signal value (account) is more accurate or if the predictive FCST executed in a table will anyway do the same.

Thanks and best regards,

Axel

[achab · Accepted solution]

Hello Axel, great question, thanks for asking. If you refer to the predictive capabilities that are currently provided in SAP Analytics Cloud planning, forecast considers only date to predict the output variable (key figure). It's recommended that you give it a try to Smart Predict predictive scenario if you want to have a greater control to your data model including additional variables (dimensions). Please note that for any dimension that you include in a data model to be used by Smart Predict time series forecasting you have to provide values not only for historical but also for the values you want to forecast for the feature.

Example: I have historical values for my KPI until end of 2018. I want to forecast all the values for my KPI for 2019. Then I'll need to integrate in the data model values for the dimensions for 2019.

Including additional variables based on calendar data, economical conditions, external events is a great way to potentially improve the forecast accuracy. The difficult part is to provide values for future dates. Easy for calendar data or specific events (Black Friday, Independence Day, Sapphire...) less easy for other variables (temperature, oil price, stock value...).

I hope this helps,

Kind regards,

Antoine Chabert

[axel_ · Accepted solution]

Hello Antoine,

thank you very much for your prompt answer!

Best regards,

Axel

[achab]

Hello Axel,

My pleasure and good to know this was useful!

Are you looking to run a specific use case combining Smart Predict and SAC/Planning capabilities?

Kind regards,

Antoine