XYZ

395. XYZ

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

where is mean periodic demand and is its standard deviation.

395.0.1. 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):

Class	CV range	Demand pattern
X	< 0.50	Smooth, predictable. Standard normal-distribution methods work.
Y	0.50 - 1.00	Variable but trackable. Use seasonality models, careful safety stock.
Z	> 1.00	Sporadic / lumpy. Traditional forecasting struggles; use intermittent-demand methods (Croston’s, etc.) or treat as project-driven.

395.0.2. 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).

395.0.3. Procedure

Collect periodic demand history (monthly is typical; daily for fast movers; quarterly for slow movers).
Compute and for each item.
Compute .
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.

Example

Given (same 6 SKUs as ABC, with 12-month demand histories):

Item	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec	Total
Laptop charger	18	20	19	21	20	19	20	21	19	20	21	22	240
Keyboard	20	25	35	30	28	22	40	45	30	35	28	22	360
Gaming chair	1	0	2	3	1	5	8	12	10	8	6	4	60
Cable	45	50	55	50	48	52	50	53	47	50	52	48	600
Sticker pack	50	100	150	80	120	100	110	90	100	100	100	100	1200
Phone case	10	20	5	30	0	50	25	10	20	15	5	10	200

Step 1 — compute and for each item

Item	(mean / month)	(SD)	CV = /
Laptop charger	20.0	1.1	0.06
Keyboard	30.0	7.5	0.25
Gaming chair	5.0	3.7	0.74
Cable	50.0	2.7	0.05
Sticker pack	100.0	25.6	0.26
Phone case	16.7	13.6	0.81

Step 2 — cut at thresholds (X < 0.5, Y < 1.0, Z ≥ 1.0)

Item	CV	Class
Laptop charger	0.06	X (very predictable)
Cable	0.05	X (very predictable)
Keyboard	0.25	X (predictable)
Sticker pack	0.26	X (predictable)
Gaming chair	0.74	Y (moderately variable, summer-heavy)
Phone case	0.81	Y (variable, on the edge of Z)

Hmm — none of the items hit Z under these thresholds. With only 12 monthly observations, CVs around 0.7–0.8 are common; the “true Z” range (>1.0) usually shows up in truly intermittent demand (months of zero, occasional spikes).

Step 3 — interpret

Laptop charger, Cable: stable workhorses. Forecast with simple moving averages or SES; carry minimal safety stock.
Keyboard, Sticker pack: mostly stable but with some seasonality / surge weeks. SES with seasonal adjustment, moderate safety stock.
Gaming chair, Phone case: lumpy. Watch for seasonality; carry more safety stock relative to mean. The gaming chair shows a clear summer pattern — deseasonalize to get a cleaner CV.

Compare to ABC: by value alone, laptop charger and gaming chair were both A items at $12,000/year. By variability, the laptop charger is X (smooth) but the gaming chair is Y (lumpy). They should get different inventory policies despite the same value rank — see the ABC-XYZ matrix.