Syntha Lab RU

ABC-XYZ and Pareto: calculation, thresholds, link to safety stock and the range

The mathematics of range classification: cumulative revenue share and ABC bounds, the coefficient of variation and XYZ bounds, the 3×3 matrix, the effect of classes on safety stock and range decisions (what to hold widely, what to cut, what to make to order). The decision: the stock policy and range depth by class.

Math and analytics Brand and retail Verified: 2026-11-13

Open interactively →

Why this document

The category manager, buyer and planner read this when they decide which items to hold broadly, which to cut and where not to skimp on stock. ABC shows who brings the main revenue, XYZ shows how predictable demand is, and together they set different rules: a fast, stable item is held at a high service level, an unstable low-revenue one is trimmed. That is why one safety-stock formula for every item either overloads the warehouse or leaves the key product unprotected.

At a glance

What is this document and why is it needed?
The mathematics of range classification: cumulative revenue share and ABC bounds, the coefficient of variation and XYZ bounds, the 3×3 matrix, the effect of classes on safety stock and range decisions (what to hold widely, what to cut, what to make to order). The decision: the stock policy and range depth by class.
Who owns the document and how often is it rebuilt?
Range analyst and buyer; recalculated quarterly (replenished basics) and after each season; the result goes into replenishment rules, OTB and the range matrix.
What mistakes are the most common?
ABC on revenue with sales and new items → one-off events land in A → a mature period without sales, new items separately. XYZ without seasonality → every seasonal item lands in Z → divide by the seasonal index or classify within a season. The same policy for every cell → expensive stock on irregular demand → policy by cell, part of Z make-to-order.

For whom

Brand
A brand classifies models and colors by revenue and partner demand stability: classes decide what to hold in the warehouse for top-ups, what to make on pre-order and what to drop from the line. Classification is based on sell-out if available, not on shipments.
Retail
A retailer classifies SKUs by revenue or margin and by sales variation across stores and weeks: classes set stock depth, replenishment rules and shelf space. Classification is done by category separately, otherwise cheap and expensive categories mix.

What it consists of

ABC: cumulative share and bounds

  • SKUs are sorted by revenue in descending order and the cumulative share is computed; class A is SKUs up to 80% of cumulative revenue, B up to 95%, C the rest (bounds are illustrative and adjustable). In the example of 120 SKUs class A has 21 (18% of the range), the top 20% of SKUs give 82.8% of revenue.
  • Revenue is taken for a mature period without sales and without new items, otherwise items that were just discounted or just launched land in A; for margin decisions rank by gross profit, not revenue.

XYZ: demand variation

  • Coefficient of variation CV = SD/mean of weekly sales, with the sample SD (divisor n−1): X at CV ≤ 30% (stable demand), Y at 30-60% (fluctuating), Z at CV > 60% (irregular). The 30 and 60% bounds are the defaults in the ABC-XYZ calculation; there is no standard, and they are edited by category and series length.
  • CV is computed on reconstructed demand and on a series without promotion peaks or with events flagged; seasonal items are compared with the seasonal index, otherwise all seasonality reads as variation.

The 3×3 matrix and its meaning

  • The ABC × XYZ matrix in the example: AX 8 SKUs and 47.4% of revenue, AY 9 SKUs and 24.7%, AZ 4 SKUs and 8.3%; CZ 23 SKUs and 2.1%. Cell AX is the core: high service and automatic replenishment; CZ are candidates for removal or make-to-demand.
  • Brand: AX is held in core colors in the top-up warehouse, AZ and BZ are made in small lots or on pre-order; Retailer: AX is always on the shelf, CZ in a few stores or online only.

Link to safety stock

  • SKU safety stock SS_i = z_class · CV_i · μ_i · √L at lead time L = 4 weeks; service levels by classes A/B/C are illustratively 2.05/1.64/1.28 (98/95/90%). Total safety stock cost is RUB 2.71 million at cost.
  • Class Z is 13.3% of revenue but 23.8% of safety stock; class X is 54.8% of revenue and 35.7% of stock: irregular demand is expensive. The saving usually lies in taking part of Z out of permanent stock, not in lowering service on X.

Brand and retailer: classes in decisions

  • Brand: a brand classifies model-color by revenue and sell-out variation; AX are core colors for top-ups, AZ and BZ are small lots or pre-order, CZ are candidates for removal from the line.
  • Retailer: a retailer classifies SKUs within a category and by store group; AX are always on the shelf and in automatic replenishment, CZ are limited to a number of stores, online or made to customer order.
  • Brand: for wholesale classes determine the depth of the seasonal buy at a partner and the number of offers: partners are offered fewer class Z items while AX is guaranteed by top-up delivery.
  • Retailer: for stores classes determine shelf space and display minimum: AX get the best place and a higher minimum, CZ go to the lower zone or online.
  • Brand: brand safety stock is held in the central warehouse by class; Z items save stock by moving to pre-order production while X items get a high service level.
  • Retailer: retailer safety stock is split between the warehouse and stores by class: for X stock in the store, for Z stock in the warehouse with transfer on order, which lowers total safety stock.
  • Brand: a brand ABC is computed on gross profit after partner returns and not on shipments: otherwise items that partners later return land in A.
  • Retailer: a retailer ABC is computed on gross profit after markdowns and shrink and accounts for item links in the receipt: class C companion items are not removed while they support class A sales.
  • Brand: a brand looks at class migration by season: a model that moved from A to C gets a removal or color-change decision; new items are classified by pre-orders and not by history.
  • Retailer: a retailer looks at class migration quarterly for the replenished basics: a move into Z starts a review of replenishment rules, a move from C to A raises the display minimum.

Where the data comes from

Data sourceFields
Sales by SKU and weekRevenue and units for a mature period (at least 26-52 weeks), without sales; reconstructed demand; category and role.
Stock parametersLead time and its spread, service levels by role, cost, MOQ and multiple.
Range matrixRoles (core, fashion, image), new items, delistings, number of stores carrying the item.

How to build it step by step

  1. Step 1. Choose the period and measure for ABC: revenue without sales (for stock) or gross profit (for the range); exclude items younger than a season.
  2. Brand: Step 2. Compute by model and color on partner sell-out; separate channels (D2C and wholesale) so the class is not set by one channel.
  3. Retailer: Step 3. Compute by SKU within a category and separately for store groups (flagships, malls, street) so rare items are not lost.
  4. Step 4. Sort in descending order, compute the cumulative share and assign A, B, C by the bounds (80/95); store the bounds in policy.
  5. Step 5. Compute the CV of weekly demand on the reconstructed series and assign X, Y, Z (30/60% by default, the bounds are editable); for seasonal items divide by the seasonal index.
  6. Step 6. Build the 3×3 matrix: SKU count, revenue share, stock value by cell.
  7. Step 7. Set the policy by cell: service, replenishment rules, stock depth, number of points, make-to-order or delisting.
  8. Step 8. Compute safety stock by SKU and the sum by class; compare the stock share with the revenue share.
  9. Step 9. Check transitions between cells over a year (migration) and class stability; items on a class boundary are assessed separately.
  10. Step 10. Fix the process: recalculation frequency, owner, revision thresholds; results go into the range matrix and replenishment rules.

Formulas

MeasureFormulasNote
Cumulative shareC_k = Σ_(i ≤ k) R_(i) / Σ R, R_(i) revenue in descending orderA: C_(k−1) < 80%; B: < 95%; C: the rest (the SKU crossing a bound goes to the higher class).
Coefficient of variationCV_i = s_i / μ_i, s_i = √[ Σ(y − μ_i)² / (n − 1) ]; X: ≤ 0.30; Y: 0.30-0.60; Z: > 0.60μ and s on reconstructed weekly demand, divisor n−1; the 30/60% bounds are the default and editable; with rare sales CV is unstable.
SKU safety stockSS_i = z_c(i)·σ_i·√L = z_c(i)·CV_i·μ_i·√LA simplification at a constant lead time L; z by class.
Stock costV = Σ_i SS_i · c_i; class share = V_class / Vc_i unit cost; compare with the revenue share.
Pareto concentration indexP_20 = C_(⌈0.2·N⌉)Revenue share of the top 20% of SKUs; the index grows with revenue inequality.
Class migrationM_(ab) = number of SKUs that moved from cell a to cell b / number of SKUs in cell aHigh migration means the bounds are sensitive and the policy must be more flexible.

Formulas in notation

ABC bounds
A: C_{k−1} < 0,80; B: C_{k−1} < 0,95; C: otherwise
The 80/95 bounds are illustrative.
Safety stock
SS_i = z_{c(i)} · CV_i · μ_i · √L
z depends on the class.
Class stock share
s_K = Σ_{i∈K} SS_i·c_i / Σ_i SS_i·c_i
Compared with the class revenue share.
Pareto concentration
P_20 = C_{⌈0,2N⌉}
Revenue share of the top 20% of SKUs.

Worked example

ABC × XYZ matrix: SKU count, revenue share and safety stock cost

CellSKUsRevenue share, %Safety stock, RUB thousand
AX847.4889
AY924.7909
AZ48.3438
BX145.361
BY146.4169
BZ72.9133
CX272.117
CY140.817
CZ232.173

120 SKUs, revenue of RUB 240 million a year, CVs are generated in code with a fixed seed. Class A: 21 SKUs and 80.5% of revenue; class Z: 13.3% of revenue and 23.8% of safety stock. The 80/95% and 30/60% bounds (the defaults of the ABC-XYZ calculation, editable) and the 98/95/90% service levels are illustrative. The data is illustrative.

Figures

Pareto curve: revenue share by SKU share
Line chart: Cumulative revenue, %: 0 → 100; Uniform distribution, %: 0 → 10002040608010001030507090100Cumulative revenue, %, 0: 0Cumulative revenue, %, 5: 56.4Cumulative revenue, %, 10: 70.2Cumulative revenue, %, 20: 82.8Cumulative revenue, %, 30: 89.2Cumulative revenue, %, 40: 93.2Cumulative revenue, %, 50: 95.8Cumulative revenue, %, 60: 97.6Cumulative revenue, %, 70: 98.7Cumulative revenue, %, 80: 99.5Cumulative revenue, %, 90: 99.9Cumulative revenue, %, 100: 100Uniform distribution, %, 0: 0Uniform distribution, %, 5: 5Uniform distribution, %, 10: 10Uniform distribution, %, 20: 20Uniform distribution, %, 30: 30Uniform distribution, %, 40: 40Uniform distribution, %, 50: 50Uniform distribution, %, 60: 60Uniform distribution, %, 70: 70Uniform distribution, %, 80: 80Uniform distribution, %, 90: 90Uniform distribution, %, 100: 100%
  • Cumulative revenue, %
  • Uniform distribution, %
Data as a table
MeasureCumulative revenue, %Uniform distribution, %
000
556.45
1070.210
2082.820
3089.230
4093.240
5095.850
6097.660
7098.770
8099.580
9099.990
100100100

Pareto curve: revenue share by SKU share. The top 20% of SKUs give 82.8% of revenue; the curve rises fast and then flattens into the tail. Axis: share of SKUs by descending revenue, %.

ABC × XYZ matrix: SKU count
Heat map: 3×3; min 4, max 27XXYYZZAA8A / X: 89A / Y: 94A / Z: 4BB14B / X: 1414B / Y: 147B / Z: 7CC27C / X: 2714C / Y: 1423C / Z: 23
  • 4 … 27 SKUs
Data as a table
XYZ
A894
B14147
C271423

ABC × XYZ matrix: SKU count. Dark cells show the SKU count: most items are in CX (27) and CZ (23); in the most revenue-heavy cell AX 8 SKUs give 47.4% of revenue.

Safety stock by XYZ and ABC classes
Bar chart: X 0.89, Y 0.91, Z 0.4400.20.40.60.811.2XXClass A: 0.89Class B: 0.06Class C: 0.02YYClass A: 0.91Class B: 0.17Class C: 0.02ZZClass A: 0.44Class B: 0.13Class C: 0.07RUB million
  • Class A
  • Class B
  • Class C
Data as a table
MeasureClass AClass BClass C
X0.890.060.02
Y0.910.170.02
Z0.440.130.07

Safety stock by XYZ and ABC classes. Class Z holds 23.8% of stock for 13.3% of revenue; the bulk of stock sits in Y.

Revenue and demand variation by SKU
Scatter plot: 60 points; regression line: y = 3.128 + −2.303·x01020304000.20.40.60.810.25; 35.50.28; 24.20.52; 13.20.54; 7.350.13; 5.090.74; 4.850.58; 4.260.67; 2.990.45; 2.90.42; 2.110.37; 1.920.5; 1.840.79; 1.760.22; 1.590.29; 1.470.15; 1.050.18; 1.050.56; 1.010.28; 0.960.15; 0.870.15; 0.820.71; 0.820.36; 0.770.17; 0.680.12; 0.60.48; 0.570.57; 0.530.65; 0.50.61; 0.480.21; 0.470.77; 0.410.14; 0.40.9; 0.360.87; 0.340.81; 0.310.28; 0.290.5; 0.280.44; 0.260.1; 0.230.12; 0.230.15; 0.220.86; 0.20.14; 0.190.11; 0.190.15; 0.160.22; 0.140.65; 0.140.5; 0.120.62; 0.110.24; 0.10.38; 0.090.72; 0.070.5; 0.060.84; 0.060.65; 0.050.32; 0.040.78; 0.030.53; 0.030.12; 0.020.7; 0.01Demand coefficient of variation↑ Annual revenue, RUB million
  • points
  • regression line: y = 3.128 − 2.303·x
Data as a table
Demand coefficient of variationAnnual revenue, RUB million
0.2535.46
0.2824.16
0.5213.16
0.547.35
0.135.09
0.744.85
0.584.26
0.672.99
0.452.9
0.422.11
0.371.92
0.51.84
0.791.76
0.221.59
0.291.47
0.151.05
0.181.05
0.561.01
0.280.96
0.150.87
0.150.82
0.710.82
0.360.77
0.170.68
0.120.6
0.480.57
0.570.53
0.650.5
0.610.48
0.210.47
0.770.41
0.140.4
0.90.36
0.870.34
0.810.31
0.280.29
0.50.28
0.440.26
0.10.23
0.120.23
0.150.22
0.860.2
0.140.19
0.110.19
0.150.16
0.220.14
0.650.14
0.50.12
0.620.11
0.240.1
0.380.09
0.720.07
0.50.06
0.840.06
0.650.05
0.320.04
0.780.03
0.530.03
0.120.02
0.70.01

Revenue and demand variation by SKU. Revenue is weakly tied to variation (correlation -0.1): large items exist in the X, Y, Z zones; the trend line is nearly flat.

Specifics by category

Apparel
Basics fall into AX, seasonal and fashion into Y and Z; classification is done within a season and by role, otherwise seasonality reads as irregular demand.
Footwear
Models are classified, not sizes; the size run within class A follows the curve; Z classes by size (edge sizes) are covered by transfers, not stock.
Accessories
Many cheap items with low turnover: ABC by revenue gives a long C tail; for the tail decide by shelf space and a cap on item count, not by safety stock.
Kids
Sizes go obsolete so the class is computed by age groups and seasons; school items are X in August and Z the rest of the time: classify by windows.
Men
More class X items: steady demand, a large basics share; stock is held deeper and the number of Z items is small.
Women
More Y and Z classes because of fashion and colors: the decision on Z is make-to-order, small lots and a fast reorder; migration between cells is higher, recalculation more frequent.

Common mistakes

  • ABC on revenue with sales and new items → one-off events land in A → a mature period without sales, new items separately.
  • XYZ without seasonality → every seasonal item lands in Z → divide by the seasonal index or classify within a season.
  • The same policy for every cell → expensive stock on irregular demand → policy by cell, part of Z make-to-order.
  • Classification for the whole chain without slicing by category and point → a cheap tail hides important items → compute within category and store group.
  • All C items are deleted → the basket breaks (companion items) → check item links in the receipt before delisting.

Russian specifics

  • For labeled light-industry goods delisting and C-class leftovers are tied to label-code accounting: stock with codes must be written off or sold under the rules, not simply removed from the catalog. Rule: Decree 1956 of 31.12.2019: Government Decree of 31.12.2019 No. 1956 (marking rules for light-industry goods) (rule check not completed).
  • C-class leftovers are measured at the lower of cost and net selling price (FSBU 5/2019): slow items require a provision and ABC-XYZ is a basis for calculating it. Rule: Order 180n of 15.11.2019: Ministry of Finance Order of 15.11.2019 No. 180n (FSBU 5/2019 Inventories) (rule check not completed).

Open interactively →