Strategic portfolio classification
The importance (margin at risk) × predictability (error) matrix and the strategy each quadrant calls for (A–Z, A–Y, B–Y, B–Z).
Why not all SKUs deserve the same attention
A typical portfolio can have hundreds or thousands of SKUs. Manually validating every SKU's forecast with the same level of rigor is impossible: the Demand Planner's time is finite. The question is where to concentrate that time so it has the greatest possible business impact.
The strategic classification answers that question by crossing two independent dimensions: how important the product is (how much it costs the business if the forecast fails on it) and how well it can be forecast (how much historical error the model has on that SKU).
The two axes of the matrix
Importance axis: A vs. B
The A/B classification measures margin at risk: how much profit is exposed if the forecast fails on that SKU. This is not just sales volume. A low-volume product with a high margin can be A; a high-volume product with a very low margin can be B.
The question is: if this SKU's forecast fails and causes a stockout or a surplus, how much profit does the business lose?
- A: high margin at risk; the consequence of a forecast error is significant.
- B: low margin at risk; the consequence of an error is limited.
Predictability axis: Y vs. Z
The Y/Z classification measures how well the system can forecast that SKU, measured empirically by the model's historical error on that product.
- Y: high predictability; the model has historically had low error on this SKU.
- Z: low predictability; error is high and this SKU's future is genuinely uncertain for the model.
In the AInventory interface, quadrants may appear as AZ, AY, BY, BZ (without a hyphen). These are equivalent to A–Z, A–Y, B–Y, B–Z respectively.
The four quadrants and their strategy
A–Z: high importance, low predictability — maximum priority
This quadrant concentrates the highest business risk: products where the margin exposure is large and the model struggles to forecast them. This is where the Demand Planner should invest the most time.
The strategy is intensive human validation: market intelligence, calls to Sales, analysis of the root causes of low predictability. Uncertainty does not disappear by simply applying another model; sometimes demand is genuinely hard to predict because it depends on external factors (negotiations, new competitors, key customer decisions). Human judgment with contextual information is the only mechanism that can compensate for that uncertainty.
If the percentage of the portfolio in A–Z exceeds approximately 25 %, AInventory recommends concentrating validation resources in that quadrant and reviewing whether there are long-term actions to reduce uncertainty (better customer segmentation, additional data, framework agreements).
A–Y: high importance, high predictability — protect
The model performs well on these SKUs and the margin at stake is high. The strategy is to protect what is working: do not intervene without solid evidence. An unnecessary manual adjustment in A–Y can destroy significant value (see FVA and the 50/50 Rule).
B–Y: low importance, high predictability — automate
The model performs well and the impact of an error is limited. These SKUs are natural candidates for full automation: let the model run without human intervention and free up team time for the higher-risk quadrants.
B–Z: low importance, low predictability — limited tactical actions
Margin exposure is low and the model struggles to predict demand. The strategy is not to invest much effort: "stop forecasting the noise." Actions are tactical and bounded: more conservative inventory policies to absorb uncertainty, or portfolio decisions (is it worth keeping this SKU?).
How the classification is calculated
The A/B and Y/Z classification is calculated in a prior process outside the forecast module. AInventory consumes the results as percentages of the portfolio in each quadrant. The KPI the system monitors is precisely the distribution of the portfolio across quadrants over time.
TODO: link to the A/B–Y/Z classification process when documented.
Relationship with the glossary
The terms margin at risk, predictability, A–Z quadrant, and Demand Planner have entries in the bilingual glossary.