399. ABC × XYZ
Combine ABC (consumption value) and XYZ (demand variability) into a 3 × 3 grid. Each cell suggests a distinct inventory strategy.
399.0.1. The grid
| X (low CV) | Y (moderate CV) | Z (high CV) | |
| A (high value) | AX: lean, JIT, supplier integration, low safety stock | AY: forecast carefully, moderate safety stock, frequent review | AZ: high safety stock + close monitoring; consider supplier flexibility (consignment, drop-ship) |
| B (medium value) | BX: standard policy, automate | BY: moderate attention, periodic review | BZ: pad safety stock; review whether SKU rationalization helps |
| C (low value) | CX: two-bin / kanban, set-and-forget | CY: simple periodic, generous buffer (cheap to overstock) | CZ: candidate for elimination — low value AND unpredictable. If kept, build to order or accept high stockout rate |
399.0.2. Reading the cells
The value axis (ABC) sets how much management attention the item deserves. The variability axis (XYZ) sets how much safety stock (relative to mean demand) the item needs. Combine:
- High value + low variability → lean. Spend the management effort on tight forecasting; reap the savings from low safety stock.
- High value + high variability → expensive to misjudge. Cannot afford either over- or understock; need close monitoring and possibly contractual risk-sharing with the supplier.
- Low value + high variability → not worth managing carefully. Either overstock cheaply or eliminate the SKU.
399.0.3. Procedure
- Run [ABC analysis](abc.typ): assign each SKU an A / B / C class.
- Run [XYZ analysis](xyz.typ): assign each SKU an X / Y / Z class.
- Place each SKU into one of the 9 cells.
- Pick the inventory policy from the matrix above.
399.0.4. When the matrix isn’t enough
The 9-cell grid is a useful starting heuristic but doesn’t capture every dimension:
- Criticality (a 1 bolt that holds together a 1M machine — see [VED](ved.typ))
- Movement frequency / dead stock (see [FSN](fsn.typ))
- Lead time (long-lead items need more safety stock regardless of XYZ)
Most operations apply the matrix as a first-pass policy assignment, then refine on the dimensions above.
Example
Given (same 6 SKUs from ABC and XYZ):
Combine the two classifications:
| Item | ABC | XYZ | Cell + suggested policy |
| Laptop charger | A | X | AX — lean, JIT-friendly, low SS |
| Gaming chair | A | Y | AY — careful forecast (seasonal), moderate SS |
| Keyboard | A | X | AX — lean, automate replenishment |
| Cable | B | X | BX — standard , automate |
| Sticker pack | C | X | CX — two-bin, set-and-forget |
| Phone case | C | Y | CY — periodic, generous buffer |
Resulting matrix populated:
| X | Y | Z | |
| A | Laptop charger Keyboard | Gaming chair | — |
| B | Cable | — | — |
| C | Sticker pack | Phone case | — |
Step — interpret
- Top-left (AX) — laptop charger and keyboard: highest priority, spend the analytics effort here. Lean continuous-review (Q, r) policy with tight safety stock.
- Center (AY) — gaming chair: pay for a seasonal forecast; widen safety stock during peak season.
- Bottom-right empty: nothing extreme in this dataset. Real catalogs typically have CZ candidates (impulse / niche / one-off items) — those are the SKU-rationalization candidates.
Why the matrix beats ABC alone: if you used ABC alone, the laptop charger and gaming chair would get identical policies (both A). The matrix correctly distinguishes them — the smooth charger gets a lean policy; the lumpy chair gets safety stock and a seasonality model. Same value, different policy.