398. XYZ

Classify items by demand variability, measured via the coefficient of variation (CV) of demand over time:

where $𝜇$ is mean periodic demand and $𝜎$ is its standard deviation.

398.0.1. The variability split

• X items: low CV — predictable demand. Forecast easily, lean inventory.
• Y items: moderate CV — somewhat variable, perhaps with seasonality or trend.
• Z items: high CV — volatile, sporadic, or irregular.

Typical thresholds (industry-dependent):

398.0.2. Why classify by variability

ABC tells you which items to focus on; XYZ tells you how predictable each item is. Both matter for inventory policy:

• X items: tight forecasting works → lean inventory, low safety stock margin.
• Y items: forecast error grows → higher safety stock; may benefit from time-series models with seasonal terms.
• Z items: forecasts unreliable → either (a) carry generous safety stock and accept overstock, (b) build to order, or (c) consider eliminating the SKU.

XYZ doesn’t replace ABC — they’re orthogonal. Use them together in the ABC-XYZ matrix (next file).

398.0.3. Procedure

1. Collect periodic demand history (monthly is typical; daily for fast movers; quarterly for slow movers).
2. Compute $𝜇$ and $𝜎$ for each item.
3. Compute $\text{CV}=𝜎/𝜇$.
4. Cut at the chosen thresholds.

Notes:

• Use enough history (at least 12 periods, ideally 24+) so CV is stable.
• Detrend if you have a strong trend — otherwise CV is inflated by drift, not noise.
• For seasonal items: deseasonalize first or use within-season variability.