ABC Analysis

abc analysis · abc classification · abc analisys · pareto analysis · abc inventory analysis · abc segmentation

Definition

ABC analysis is the classification of products or customers into three groups by the cumulative share of their contribution to a chosen metric: A is about 80% of the result, B about 15%, C about 5%.

The principle and the formula

ABC analysis rests on the Pareto principle: a smaller part of the items produces the larger part of the result. In retail this distribution shows up consistently — in typical catalogues 15–25% of the assortment accounts for around 80% of revenue.

The method turns that observation into a procedure: sort, compute the cumulative contribution, split at the boundaries.

1. Choose a metric and a period      (revenue over 12 months)
2. Sort in descending order          (largest contribution first)
3. Item share      = Item metric / Total across all items × 100%
4. Cumulative share = Sum of the shares of all items up to and including this one
5. Group:  A — cumulative share ≤ 80%
           B — from 80% to 95%
           C — above 95%

The 80/95 boundaries are a starting point, not dogma. If the distribution curve in your category is flat, applying 80% mechanically produces a group A the size of half the assortment and the point of prioritising is lost. The boundaries get moved to fit the actual shape of the curve.

A worked example

A category of eight items, metric — revenue for the quarter.

Item	Revenue, $	Share	Cumulative share	Group
Product 1	42,000	42.0%	42.0%	A
Product 2	23,000	23.0%	65.0%	A
Product 3	15,000	15.0%	80.0%	A
Product 4	8,000	8.0%	88.0%	B
Product 5	5,000	5.0%	93.0%	B
Product 6	3,500	3.5%	96.5%	C
Product 7	2,500	2.5%	99.0%	C
Product 8	1,000	1.0%	100.0%	C
Total	100,000	100%

Three items out of eight deliver 80% of the category revenue. That is the working output: a list you must never let go out of stock, and a list that deserves the best slots in the listing.

The choice of metric changes the answer

The most common mistake is to run ABC on revenue only and take every decision from it. Different metrics produce different group A lists:

Metric	The question it answers	Who lands in A
Revenue	Who produces the turnover	Expensive and fast-moving items
Gross margin, $	Who makes the money	High-markup items, not always expensive
Units sold	What loads the warehouse and logistics	Cheap, high-volume items
Orders containing the item	What brings the shopper in	Traffic drivers, often at low margin

An expensive fridge lands in A by revenue and in C by units. Batteries do the opposite. An assortment decision taken on one cut will be systematically wrong for the other. The working practice is to run ABC on revenue and on margin at minimum and look at the intersection: items in A on both metrics are the core of the category, while items in A by revenue and C by margin are a separate conversation about pricing or purchasing terms.

ABC for customers and the ABC x XYZ matrix

ABC for customers. The same procedure applies to the customer base: sort by revenue or margin for a period, split into groups. The result is usually more concentrated than for products. Group A is the set of customers whose departure shows up in the P&L, and they are the ones worth separate terms and priority in communications. For finer work, ABC is paired with RFM analysis: it adds recency and frequency, separating “a large customer who is active” from “a large customer who stopped buying six months ago”.

ABC x XYZ. ABC measures the size of the contribution, XYZ analysis measures the stability of demand through the coefficient of variation. Together they produce a matrix of nine groups:

	X (stable demand)	Y (fluctuating)	Z (irregular)
A	The core: never allow a stockout, hold buffer stock	Large contribution, needs an accurate forecast and safety stock	Large but unpredictable — risk of tying up cash
B	The stable middle, automatic replenishment	Standard inventory management	Replenish against actual demand
C	Small but predictable — minimal stock	Candidate for reduced depth	Candidate for delisting or make-to-order

It is this matrix, not bare ABC, that is fit for stock decisions: AZ and CZ require fundamentally different handling even though both are formally “unstable”.

How ABC is used in personalization

The results of ABC analysis move from a report into the storefront through three mechanisms.

Priority in recommendations and merchandising. Merchandising rules can raise the priority of group A items all else being equal: if the algorithm considers two products equally relevant to the user, the one that matters more to the category goes higher. The inverse rule is useful too — items with negative margin are excluded from the selections.

Protecting new products. A new product is in group C by definition — it has no accumulated sales yet. If the rules are hard-wired to ABC, new arrivals never get impressions and never leave C. So items younger than a defined age are taken out from under the ABC rules and assessed separately.

Protecting the tail. This is the main limitation, and the one that gets forgotten.

⚠

If listing results and recommendation blocks are filled only with group A items, the long tail stops receiving impressions, its sales fall, and at the next ABC recalculation part of group B drops into C. The process is self-reinforcing: the assortment gradually collapses to a few dozen items, and with it go both catalogue coverage and the store’s ability to serve non-standard demand.

The practical defence is quotas. Recommendation blocks and category pages fix a share of items outside group A, and the question of what share is optimal is settled by measurement rather than argument: an A/B test with different tail shares in the listing shows where the line runs between a focused storefront and an impoverished assortment.

Application checklist

The calculation period is no shorter than a full demand cycle for the category; for seasonal goods, a year at minimum.
The metric is chosen to fit the question, not defaulted to revenue.
Group boundaries are checked against the actual distribution curve rather than taken as 80/15/5 automatically.
ABC is calculated inside a category rather than across the whole catalogue at once: otherwise an entire small category ends up in C.
New products are excluded from the calculation or flagged separately.
Delisting decisions target items with no sales and negative margin, not the whole of group C.
The result is refreshed on a schedule rather than once: groups shift as demand shifts.
The storefront enforces a quota for items outside group A.

Frequently asked questions

How do you run an ABC analysis step by step?

Choose a metric and a period. Sort items in descending order of that metric. Calculate each item's share of the total. Calculate the running cumulative share. Assign group A to every item until the cumulative share reaches 80%, group B up to 95%, and group C to the rest. The 80/95 boundaries are a starting point and get adjusted to the real shape of the distribution curve.

Which metric should ABC analysis use?

It depends on the question. To assess contribution to turnover, use revenue. To assess profitability, use gross margin in currency. To plan warehouse operations and packing, use units sold. To assess the role in traffic, use the number of orders the item appears in. The group A lists differ across these metrics: an expensive product with rare sales lands in A by revenue and in C by units.

How is ABC analysis different from XYZ analysis?

ABC answers "how much does it bring", XYZ answers "how stable is it". ABC ranks by the size of the contribution, XYZ by the stability of demand through the coefficient of variation. They are used together: the intersection gives nine groups, where AX is large and predictable demand (never allow a stockout) and CZ is small and erratic (a candidate for delisting or for make-to-order).

Can ABC analysis be applied to customers?

Yes, and it is one of the most useful variants. Customers are sorted by revenue or margin for a period, then split into the same groups. Group A is a small part of the base that produces most of the revenue: it deserves separate terms and priority in communications. For behaviour-based segmentation, ABC is often paired with RFM analysis, which also accounts for recency and frequency of purchases.

Why not keep only group A in the assortment?

Because the groups are neither static nor isolated. Group C products often act as complements: the shopper comes for an item from A and fills the basket with items from C, so removing the latter lowers average order value. On top of that, once the C tail is cut, part of the former B products become the new C — the process repeats and the assortment collapses. Cuts should target items with no sales and negative margin, not the whole of group C.