Sign in

Blog · Data quality and reconciliation

B2B customer segmentation from the ledger, not from personas

How to segment a B2B customer base from data the company holds: the two or three fields that make a segment useful for norms, size, sector and channel, why behavioural fields like order frequency and product mix belong in the measures rather than the segment definition, the minimum segment size that makes a norm mean something, the test that a segmentation is good, and why personas do not survive contact with a ledger.

The short answerA segment exists so that a customer can be compared to similar customers. Define it from fields that describe what the customer is, not what it does with you: size band, sector, and channel or type, from the customer master. Keep behaviour, order frequency, product mix, share, as measures computed within the segment, or the segment will be circular. Each segment needs enough members, thirty is a common floor, for its norm to mean anything, and the test of a good segmentation is that the spread of share of wallet within a segment is smaller than the spread across the base.

A company has personas from a workshop and segments from a marketing tool, and neither can be assigned to a row of the ledger by rule. Segmentation for analytics has one job: to put each customer beside the customers it should be compared to. This guide sets out the fields that do that, the ones that must not, the size floor, and the test.

What a segment is for

A segment is the set of customers whose norm applies to each other.

Share of wallet, products per customer, gap at norm, expected basket: every one of them compares a customer to its segment's norm. The segment is the denominator's definition.

Fields that define a segment

Field Why Source
Size band Bigger customers buy more; the norm scales Customer master: employees, sites, revenue band
Sector Different sectors buy different mixes Customer master: SIC or the company's own code
Channel or type Direct, distributor, chain, independent Customer master

Two or three fields. All from the master, none from the ledger.

Fields that must not

Field Why not
Revenue with the company The segment becomes "customers who buy a lot from us"; the norm is circular
Order frequency Same
Product mix Same
Share of wallet The measure the segment exists to compute

Behaviour is what the segment measures, not what defines it.

The size floor

Per segment cell, a minimum member count, thirty is common, below which the norm is greyed and the cell is merged upward. The report shows the count on every cell.

The test

A segmentation is good when a customer's measures are closer to its segment's norm than to the base's:

spread of share of wallet within segments < spread across the base

If the segments do not tighten the spread, they are not separating anything, and the fields are wrong.

A worked segmentation

Fields: size band (3), sector (6 after merging), type (2). 36 cells; 22 above the floor.

Cell Members Median products per customer Share of wallet, interquartile spread
Large, manufacturing, direct 84 5.8 18 points
Mid, manufacturing, direct 210 3.9 14 points
Small, services, distributor 31 1.6 22 points
Base, unsegmented 2,400 2.8 41 points

Within each cell the spread is a third to half of the base's. The segments separate. The norms per cell are usable, and the small services cell at thirty-one members is just above the floor and says so.

Where it goes wrong

Behavioural fields in the definition. Every customer at norm; no gap anywhere.

Too many cells. Norms from six customers.

Personas as segments. Not assignable by rule; no norm.

Segments never re-tested. The spread test once a year says whether they still separate.

Once a year, and on every norm

Mapped once, the customer master produces the segments, the counts per cell and the spread test, and every norm on the site is computed within them. Covirage builds this from the exports as they are. The sales intelligence solution describes the setup, and the norm guide covers what is computed inside each segment once it exists.

Questions people ask

Why not segment on behaviour?

Because the point of the segment is to say what a customer should be doing, from what similar customers do. If the segment is defined by what the customer does, high-share customers are in one segment and low-share in another, and every customer is at its segment's norm by construction. The norm says nothing.

How many segments?

As few as give a stable norm in each. Three size bands times eight sectors is twenty-four cells; if half have fewer than thirty customers, merge sectors until they do. The report shows the count per cell and greys the ones under the floor.

What about personas?

A persona describes a buyer's motivations for marketing. It is not on any row of the ledger, cannot be assigned to a customer by rule, and cannot produce a norm. Segments for analytics come from fields on the customer master; personas can sit on top for messaging.