The New Microsoft Dynamics AX Demand Forecasting

Microsoft Dynamics AX2012 R3 introduced statistical based demand forecasting to Microsoft Dynamics AX Demand Forecasting. The new Microsoft Dynamics AX (V7), as you may know, is first released for Cloud only and uses Azure machine learning rather than the SSAS data mining models supported in Microsoft Dynamics AX2012 R3. Below is an introduction to some of the major changes users will see with this new interface.

• Item allocation keys can be accessed in master planning workspace rather than having to navigate to Inventory Management.

• The UI is now right in client, rather than exporting to the SSAS cube in Excel, however, forecasts can still be exported to Excel for creating KPIs in conjunction with Power BI.

• Manual adjustments are now highlighted by the values being Bold rather than the red triangle icon from the Excel file in R3.

Changes to forecast models or APIs:

• ETS – Exponential Smoothing has been added.
• STL – Seasonal Trend Decomposition has been added.
• ARIMA – Autoregressive Integrated Moving Average, is still available.
• Mixed option is still available.
• ARTxp is no longer available.
• Automatic model selection can be used if you are unsure of which API to use.

• Demand forecast is not used in Master Planning until it is authorized for use, similar to the R3 periodic import being required before the forecast was actually used in Microsoft Dynamics AX.
• The solution is much more scalable, with no dedicated hardware required and the ability to process up to 200 item allocation keys simultaneously.
• Can design custom algorithms easily and ISVs can publish their own solutions on Azure marketplace if the APIs in Microsoft Dynamics AX are not suitable for your industry or company.

While there are still some features missing that companies see in other large forecasting packages, like out of the box KPIs, or comparison of many models simultaneously, this is a huge step in the right direction, further enhancing the Microsoft Dynamics AX reputation as a best of breed ERP solution.

Note: You will need an Azure Machine Learning account to generate a forecast using the AML strategy. The “Copy over historical demand” option for forecast generation is still readily available without requiring such an account.

For more information please download the “How to choose algorithms for Microsoft Azure Machine Learning” Cheat Sheet: