Using ABC and XYZ Segmentation

• ABC segmentation is the prioritization of planning objects based on their relative importance.

  For example, you may want to categorize the combinations of product and customer based on the revenue key figure.

• XYZ segmentation is the classification of planning objects based on their demand volatility.

  For example, you may want to categorize the combinations of product and location based on the confirmed quantity key figure.

Segmentation helps you define more specific alerts and reports and generate more accurate results for demand planning and inventory planning. Among others, segmentation allows you to do the following:

• Tailor your forecast strategy and inventory optimization to the characteristics of segments

• Assign more-appropriate forecast models to planning objects

• Analyze forecast accuracy by product segments

• Identify planning objects with a relatively high or relatively low level of forecastability

• Identify inventory items that require closer attention

• Define alerts for specific segments only

• View analytic charts with regards to specific segments

ABC segmentation is the prioritization of planning objects based on their relative importance.

The calculation is performed as follows:

• For each planning object, the system calculates the total value of the segmentation measure, which is a selected key figure, for the selected segmentation period. For example, it calculates that PROD01 produced a total revenue of USD 51,000 during the last 10 weeks.

• The system sorts the planning objects in decreasing order of the calculated total values. For example, the planning object that produced the most revenue in the last 10 weeks is moved to the first place.

• The system uses the selected segmentation method and the predefined thresholds to assign the sorted items to segment A, B, or C.

Segmentation methods:

• By Pareto Principle (Sorted and Cumulated %):

  This method is based on the application of the Pareto principle on a set of planning objects within a specific time horizon. The Pareto principle or 80/20 rule states that for many events, roughly 80% of the effects come from 20% of the causes. The goal is to rank products or other planning objects to identify those that contribute the most and the least to the overall value of the segmentation measure (for example, revenue) in percentage. Therefore, all items are assigned to A, B, and C segments in the descending order of their relative importance. Note that the sum of the percentages that are used to define the thresholds must always be 100%.

  Note

  Zero and negative values are automatically assigned to the last segment.

• By Pareto Principle (Sorted and Cumulated Values):

  This method is based on the application of the Pareto principle on a set of planning objects within a specific time horizon. The Pareto principle or 80/20 rule states that for many events, roughly 80% of the effects come from 20% of the causes. The goal is to rank products or other planning objects into absolute value. Therefore, all items are assigned to A, B, and C segments in the descending order of their relative importance.

• By Number of Items (Sorted %):

  If you choose this method, the system calculates the segments based on the relative number of items that produce various shares of the segmentation measure.

• By Number of Items (Sorted Values):

  If you choose this method, the system calculates the segments based on the number of items that produce various shares of the segmentation measure.

• By Segmentation Measure (Single Values):

  If you choose this method, the segments are calculated based on the total value produced by each item in terms of the segmentation measure. The sums are compared one-by-one to the predefined thresholds.

XYZ segmentation is the classification of planning objects based on their demand volatility, which can be established using two different calculation strategies: Calculate Variation and Aggregate over Periods.

The main difference between the two strategies is that Calculate Variation calculates the variance values during the segmentation runs, while Aggregate over Periods works with values that were previously calculated by other tools such as the Manage Forecast Error Calculations app.

• Calculate Variation:

  If you choose this calculation strategy, the system evaluates the fluctuations in the historical demand or consumption during each segmentation run and classifies planning objects based on their regularity or propensity for planning.

  The following calculation methods are available for this purpose:

  • Coefficient of Variation (CV):

    If you choose this method, the system assigns each item to various segments based on their coefficient of variation, which is the standard deviation divided by the mathematical mean value for a specific time frame:

    

    In the formula above, σ is the standard deviation and X̅ is the mathematical mean.

  • Coefficient of Variation Squared (CV Squared):

    If you choose this method, the system divides the square of the deviation with the square of the mean, getting CV squared as a result. CV squared is considered more convenient than CV for evaluating demand fluctuation.

  Consider the Results of Time Series Analysis:

  When forecast automation is run in the Application Jobs app, the system analyzes the time series values stored in the selected input key figure and saves the identified properties in the background. If you select Consider Results of Time Series Analysis for XYZ segmentation, the system will automatically use the analysis results to remove any trend or seasonality from the time series data if necessary and calculate CV or CV Squared based on this transformed time series. Both trend and seasonality are known to distort the segmentation results because they indicate volatility despite often being predictable.

  Let’s say, for example, that the CV calculated for a planning object is 0.6 with an observed trend left untouched, which means that the item would be assigned to segment Y. With the trend removed, the CV calculated for it is 0.2, which means that the item would be assigned to segment X. Since the second CV value is lower than the first one, the system automatically chooses X as the segmentation result.

  If the time series data is intermittent, it may happen that the CV value is higher after the cleansing than before. In this case the planning object gets assigned to the segment that was calculated based on the original segmentation measure values.

  Note: The analysis results can only be considered if time series analysis has already been executed for the input key figure of XYZ segmentation.

• Aggregate over Periods:

  If you choose this calculation strategy, the system performs the following steps:

  1. It retrieves the segmentation measure values that were calculated by another tool such as the Manage Forecast Error Calculations app.

  2. It processes the segmentation measure values using the aggregation method that you select.

  3. It compares the results of the calculations to the thresholds you specified and assigns the planning objects to XYZ segments accordingly.

  This calculation strategy is useful, for example, if you want to classify the planning objects based on a forecast error measure such as MAPE.

  The following aggregation methods are available:

  • Minimum: The smallest value produced by an item is compared to the thresholds.

  • Maximum: The largest value produced by an item is compared to the thresholds.

  • Average: The average of all values produced by an item is calculated and compared to the thresholds.

  • Sum: The values produced by an item are summed up and the total value is compared to the thresholds.

  Note

  This calculation method is identical to segmentation method no. 5 in ABC segmentation.