Wednesday, 18 June 2014

AX 2012 R3 Demand Forecasting Basics

The two big features of the R3 release has clearly been the TMS and the new WAX. Unfortunately, Demand Forecasting a requirement for supply chain planning has also been added to  R3  . It is the first time we are seeing Demand Forecasting in any Microsoft Product, with a clear shift in the ideology to focus on the Supply Chain. In this post first i want to discuss the practice demand forecasting in the the supply chain, and then compare this with the functionality in AX 2012 R3. This will serve two purposes, not only will the rather hazy subject of forecasting be clarified, it will also shed light on how functional consultants can customise the module to make it more advanced.

Demand Forecasts the result of taking historical demand data from the company in forming the future plans for true predicted demand. To this effect the levels which forecasts can be made can determine the accuracy of the forecasts. The requirement is also dependent on company policy and at what level procurement is done.
  • Customer Level
  • Product Level
  • Product Group 
  • Territory
  • Region
  • Corporate Level

Before we look at the techniques of forecasting from a academic perspective, let me first discuss how to setup the demand forecasting module in AX 2012 R3.

Within the Master Planning Module traverse down to setup and Demand Forecasting Parameters. The form you should see can be seen in Figure 1.





As you see the Demand Forecast Unit is the first field that has to be defined. It is the unit in which the AX will calculate the forecast. The next two cubes are the Demand Forecast Intitial, and Demand Forecast Initial Accuracy cubes that have to be deployed to the Analysis Server. The deployment is a prerequisite for the Demand Forecasting. Demand Forecasting is AX is based upon the SSAS Forecasting algorithm that has been around for years within SQL Server but only now being introduced in R3.

Fourthly a path has to be defined where AX can output the excel file of forecasted values.

The next selection allows the user to define from what type of historical data should the forecast be based on. The options include almost all the modules in AX where items can be consumed.


The Forecasting algorithm section is the most important. There are two options to use the copy from historical data and SSAS Time Series forecasting models.

Copy from historical data only creates the forecast baseline from historical data, while the SSAS options does the same using the SSAS time series algorithm.

For more information on the SSAS Time Series Algorithm check out visit MSDN . In the event that this SSAS method is chosen then the user has the opportunity customising it by adding and removing features, but it is optional. Furthermore it requires thorough understanding of the algorithm .

The next setup is to define the detail of the forecast, as mentioned in the first list of this tutorial.


As you see the available dimensions in AX 2012 are unto the best practices of Demand Forecasting, thus finding a connection between academia and industry. The next and final stage of the setup is to define the Item Allocation Keys; to set up the items for which the forecast will work. Again you can define the algorithm parameters and transaction types for each allocation key group if necessary. Stay tuned for more on Item Allocation Groups.

Item selection is critical to demand forecasting. Practically it is not feasible to forecast and plan the supply of all the items. Thus the general practice is to use Pareto Analysis and ABC classification to determine which products are worth forecasting. It is a fact that forecasting is very time consuming and generally considered to be incorrect. In this regard AX is lacking. There is no functionality that is in built that can maintain ABC classification of products. Using the historical demand a simple functionality can be developed to update the ABC classification on a periodic basis. This will help reduce the time that is required by forecasters or planners. Another addition that can be made to AX is the use of Mean Average Deviation (MAD) to further classify products. After generating the baseline forecast, the forecast accuracy feature in Master Planning will calculate the Mean Absolute Deviation (MAD) for the historical demand and forecasts. With a little customisations it is possible to use this data to then reclassify products according to ABC to pinpoint the items which have the most deviation and then subsequently find a solution.

Hypothetical Solution

  1. Import the MAD figures into a table in AX 
  2. Run a activity to reclassify the ABC of specific products according to MAD (This can be a bi weekly or monthly activity)

Items with the most deviation will require tighter control on the ordering policies so they could potentially be promoted to the higher category.



Note Another perquisite to any of the features in the Master Planning Modules is to first set up the Intracompany Planning Groups.






1 comment:

  1. Varun....some of the screenshots on your blog are not showing up...plzzz check...like the latest post and here and discrete BOM --part 1 & part 2...

    Thanks for the post....Learning DAX... are you functional or tech....

    ReplyDelete