Forecastability (1).png

17 May 2022

ABC XYZ Segmentation in SAP IBP

The ABC XYZ segmentation has become a cottage industry on its own in the last 20+ years. I remember presenting this in an IBF conference many years ago simply as "Planning by Exception" that included many elements - product segmentation, customer Pareto, Error analysis, Delta filters, etc. With a new name and a moniker, it is not ABC-XYZ. By definition, it is a nine-blocker in a 3 by 3 grid regardless of whether the grid makes sense or not.

Many tools provide the option to have ABC-XYZ updated seamlessly - Kinaxis Rapid Response and SAP IBP are two such tools that have this functionality and feature this prominently.

Here is a brief critique of the functionality provided by SAP in its integrated Planning tool.

IBP allows you to easily set up the volume thresholds to come up with a Pareto calculation using the historical data - typically at 70% for High, 25% for Medium, and the balance as Low volume. When we set this up, we sometimes use a long-tail filter so 4% goes to C and the last 1% tail goes to D. It will be interesting to see how many SKUs get flagged as D. We also set up another class called O or E that may include inactive SKUs.

Call Us For A Free Consult


SAP IBP allows you to set up the right metric to use for volume segmentation - units, weight, or revenues. Perhaps if you want to steer your process with value, you will choose revenues. If you are a liquor shop, you may want to use ounces or barrels of alcohol. If you are selling potatoes, then Pounds or bags it is!

Up to this point, it is as easy as apple pie. But the hard part is the trick that involves setting up the planning levels to use in the segmentation. The attribute can be set up at a single level or compound level so the segmentation value will be available at all levels of detail for planning. However, the calculation level could be different. For example, I need the segmentation at the Sales Org -Customer-Product-Location level but I want to sum up the numbers to the level of Sales Org and Product. In this case, the ABC would be a Pareto for each Sales Org and will segment the products independently for each Sales division.

ABC segmentation in SAP IBP at the Product level

The XYZ classification is meant for segmenting products by volatility or ex-post forecast error. The system purports to classify according to demand patterns. The system is using Co-efficient of Variation to classify demand into Constant demand, Patterned demand, and Sporadic Demand.

Although most consultants use the CV or the Co-efficient of Variation to do segmentation for Forecasting, we all know the CV will overshoot given it will classify even predictable demand patterns as undesirable given the CV will be high for a trending pattern or a seasonal pattern.

Sporadic or intermittent demand generally can have higher CV but the intermittency trumps the effects of high volatility. The forecasting characteristics are different so I am not sure how we can interpret very high CV to indicate intermittency or vice versa.

Unfortunately, SAP IBP gives you only two options, the CV and CV-Squared. When you put the two pieces together, IBP gives you the nine-blocker. The SAP help manual provides the following as the interpretation of the ABC-XYZ classification.

Some observations:

  1. CV will be low if you have a constant demand pattern. So AX, BX, and CX all pass the criteria. However, the others do not.
  2. The Y class is supposed to be patterned demand - but it may or may not have a pattern. All we know is that CV is higher than the constant demand category.
  3. The Z class may or may not be sporadic. Surprisingly, this could have products with good patterns and steep growth curves.

If your demand pattern is highly seasonal or sporadic, your CV will be high. The higher the amplitude of the swings, the higher the CV, and in some cases, this may be even larger in magnitude than when the demand pattern is sporadic.

Similarly, CV may be high for products that are recently launched and have just a few months of history - there is no provision for New Products or products with a short history.

There are a few nice options including the K-means method to classify demand rather than using CV. Alternatively, you can also use the time series options to classify demand using the dataset that is adjusted for trend and seasonality.

For more information to make sense of IBP and navigate through all of the infinite possibilities that SAP IBP provides for demand planning, you can contact us.

Leave a Reply