ABC-XYZ Classification — Multi-SKU, from CSV

ABC splits your portfolio by share of value — the 80/20 principle applied to your own data. XYZ splits it by demand variability. You choose the cut-offs and the tier names, then run the matrix.

Calculated locally in your browser. No server uploads.

Read the ABC-XYZ matrix & replenishment policy guide

Classification inputs

ABC uses the cumulative share of sales value. XYZ uses the coefficient of variation of monthly demand — the lower the variation, the more predictable the item.

ABC — share of sales value

% cum.
% cum.
C remainder
5%

XYZ — demand variability (CV)

% CV
% CV
Z above
50% CV

Running on a demo portfolio of 50 items — upload your operational CSV on the Upload page to use your own data.

How this ABC inventory analysis works

Each item's sales value over the last 6 months is calculated and items are ranked from highest to lowest. The running total decides the class: items within the first 80% of total value are class A, the next 15% are class B, and the remaining tail is class C. You can change these cut-offs and the class names.

XYZ adds demand stability. The coefficient of variation (CV = standard deviation of monthly demand ÷ mean monthly demand × 100) puts items with CV up to 25% in X, up to 50% in Y, and above 50% in Z by default. Items without monthly history should be read as unknown variability.

When to use the results

  • Give A items the tightest review: higher service levels, frequent counts and close supplier follow-up.
  • Use the XYZ class to decide how much buffer is needed: steady X items need less safety stock than erratic Z items.
  • Review C and CZ items for slow-moving or excess stock before reordering.
  • Feed the classes into the Safety Stock tool, which sets service levels by ABC class.

For a policy per segment, read the ABC-XYZ 9-box matrix guide.

Methodology & Guide

How Pareto (80/20) ABC classification and XYZ variability work, with a worked example.

Formula breakdown

Rank items by 6-month sales value, then:

Value share = item value ÷ total value × 100
Cumulative share = running total of value shares

ABC class (default cut-offs 80% / 15%):

A: cumulative ≤ 80%   B: cumulative ≤ 95%   C: the rest

XYZ class from monthly demand variability:

CV = σ(monthly demand) ÷ mean(monthly demand) × 100
X: CV ≤ 25%   Y: CV ≤ 50%   Z: above 50%
Value
Sales value over the last 6 months, from the Operations CSV.
A / B / C
Value tiers. A = the few items that make most of the value; C = the long tail.
CV
Coefficient of variation: how much monthly demand swings relative to its average.
X / Y / Z
Variability tiers. X = steady demand, Z = erratic or lumpy demand.

Worked example: a 1,000-SKU distributor

A distributor uploads 1,000 SKUs with 2,000,000 in 6-month sales value and keeps the default 80% / 15% cut-offs.

  1. Items are sorted from highest to lowest sales value and their shares are added up.
  2. The first 180 items reach 1,600,000 (80%) → Class A: 18% of SKUs carry 80% of value.
  3. The next 270 items add 300,000 (15%, cumulative 95%) → Class B.
  4. The remaining 550 items share the last 100,000 (5%) → Class C.
  5. One A item averages 400 units a month with a standard deviation of 80 → CV = 20% → X, so it is AX.

Focus planning time on the 180 A items; AX items suit tight automatic replenishment, while CZ items are candidates for make-to-order or delisting.

Quick FAQs

Do I have to use 80/15/5 cut-offs?

No. 80/20 is a rule of thumb, not a law. You can set your own A and B shares on this page; the Safety Stock and Reorder Point tools then use exactly the same classes for their service levels.

Why rank by sales value instead of units sold?

Value shows where your money and margin are. A cheap item can sell thousands of units and still matter less than a slow, expensive one. Unit volume is reflected separately through the XYZ demand-variability split.

Why did an item change class after a new upload?

Classes are relative: each item's place depends on everyone else's sales in the latest 6 months. A seasonal peak, a new product or a lost customer shifts the ranking, which is why it is worth re-running ABC regularly.

Full derivations, data safeguards and the privacy architecture are on the Methodology & Formulas page.