Power BI Forecasting vs. Excel Forecasting - why different numbers?

[mmace1]

Using the same source data, same confidence level, and default settings on both, Excel & Power BI give different numbers. 

The forecasts are pretty close (again though, not the same), and additionally Power BI seems to give tighter ranges for the same confidence level. 

Is there any overview as to how Excel & Power BI handle forecasting differently? 

[v-lili6-msft · Accepted solution]

hi, @mmace1 

You may refer to this blog:

https://powerbi.microsoft.com/en-us/blog/describing-the-forecasting-models-in-power-view/

https://powerbi.microsoft.com/en-us/blog/introducing-new-forecasting-capabilities-in-power-view-for-...

Does Power View use the same forecasting algorithm as in the Data Mining Add-ins for Excel?

No, this is a different tool, and a different (but similar) algorithm.

• Charting and forecasting in Power View does not require a connection to Analysis Services, and instead runs on data stored in Excel, with processing done “behind the scenes” on the Excel Data Model.
• Forecasting on Power View charts does not use the ARTXP or ARIMA algorithms provided by SQL Server Analysis Services. Although those algorithms were developed by Microsoft Research in response to common theories and research on time series forecasting, and solve many of the same problems, they are implemented on a different code base and might yield different results.

We recommend that you experiment with various methods of forecasting and compare results. Typically data scientists will try out many models and evaluate their accuracy before accepting a prediction as valid for business application.

Best Regards,
Lin

[v-lili6-msft · Accepted solution]

hi, @mmace1 

You may refer to this blog:

https://powerbi.microsoft.com/en-us/blog/describing-the-forecasting-models-in-power-view/

https://powerbi.microsoft.com/en-us/blog/introducing-new-forecasting-capabilities-in-power-view-for-...

Does Power View use the same forecasting algorithm as in the Data Mining Add-ins for Excel?

No, this is a different tool, and a different (but similar) algorithm.

• Charting and forecasting in Power View does not require a connection to Analysis Services, and instead runs on data stored in Excel, with processing done “behind the scenes” on the Excel Data Model.
• Forecasting on Power View charts does not use the ARTXP or ARIMA algorithms provided by SQL Server Analysis Services. Although those algorithms were developed by Microsoft Research in response to common theories and research on time series forecasting, and solve many of the same problems, they are implemented on a different code base and might yield different results.

We recommend that you experiment with various methods of forecasting and compare results. Typically data scientists will try out many models and evaluate their accuracy before accepting a prediction as valid for business application.

Best Regards,
Lin

[mmace1]

Thanks! 

Is there confirmation that the (2014) blog  data there is still accurate?  On the PBI side it mentions PowerView (now depricated), and on the Excel side it mentions the data add-in (it's built-in to Excel now, not an add-in).

[v-lili6-msft · Accepted solution]

hi, @mmace1 

My apologies,  It seems that there is no new information about this,  but the development route of them should be similar. So you could use it as a reference.

Best Regards,

Lin