In addition to the automatic ABC classification, SKU Science also allows you to organize your forecasts according to the XYZ classification in order to identify those that are relatively stable or easy to predict, versus those that are more unstable or with high amplitudes.
Two options are available for the XYZ classification:
Coefficient of variation method
Forecast error method (Pro and Enterprise plan)
To access the rules to define the XYZ classification method, simply click on the gear icon, then select Rules.
You will then access this interface.
Classification method
Depending on your subscription, this list allows you to choose between 2 options:
Coefficient of variation
Forecast error
XYZ classification using the Coefficient of Variation (COV) method
The calculation of the coefficient of variation is a statistical method.
XYZ analysis using the coefficient of variation method divides items into three categories. X items represent the most historically stable items. Y items are generally characterized by moderate variation in demand. Z items are the most unstable, with large variations.
This method identifies the deviation of values from the average of all these values. This gives you an idea of how stable your article is over time.
The formula used for the calculation of the COV is the following:
CV is the coefficient of variation
σ is the standard deviation
μ is the mean
Class X limit %
Next, you need to specify the maximum coefficient for class X.
By default, it is set to 10%.
Thus, if the limit of class X is set to 10%, all items with a COV between 0 and 10% will be classified X.
Classes X and Y limit %
By default it is set to 30%.
If you set the Y class limit to 30%, then all items with a COV between 10% and 30% will be classified as Y, and those above 30% will get the Z class.
While this method is interesting to some extent for identifying stability, a product with high seasonality will be identified as a Z product, while potentially it can be easy to predict on SKU Science.
For this reason, we have introduced another method providing better results.
XYZ classification using the forecast error method
By selecting Forecast Error in the interface window, you get the following options:
XYZ lag
The reason why you are making your forecast usually drives the choice of the lag for the classification calculation. Explanations about the choice of the lag for the calculation are available here.
If you enter "2", the forecast considered for calculating the error vs the actual demand value will be the forecast created two months before each period for each item in your account.
The tables below help you understand how this works:
For item 0011-3, by choosing 'Lag 2', the table automatically updates with the values for the errors (and other KPIs).
On the error line, an average percentage is computed for this item, here 25%.
This information is computed for each item in order to establish the classification for a given lag. You can then check the results of the matrix generated according to your parameters.
Error type
Two options are available.
The first choice will calculate the XYZ classes from the user forecasts, taking into account all adjustments made by the user on the baseline forecasts. The baseline forecast is either computed by the platform or uploaded through an external forecast.
The second choice will calculate the XYZ classes only from the baseline forecasts, without considering any adjustments made by the user.
Class X limit %
Next, you need to specify the maximum percentage for class X.
By default, it is set to 10%.
Thus, if the limit of class X is set to 10%, all items with an average error for lag 2 between 0 and 10% will be classified X.
Classes X and Y limit %
By default it is set to 30%.
If you set the limit of class Y to 30%, then all items with an average error for lag 2 between 10% and 30% will be classified Y, and those above 30% will get class Z.
By clicking on Save you save your settings, and a new calculation of the XYZ classes is launched.
Once the calculation is complete, click on the Home tab to view the priority table.
Then simply click on the part of the matrix for a given combination of classes, to automatically get a view of these items. As an example, you can directly access the list of AY class items.
Do not hesitate to contact us to ask your questions directly in the tool interface, using the messenger at the bottom right of the screen.