Customer segmentation basics: three splits worth making

Aug 22
3 min read

Updated: 4 days ago

Introduction

Segmentation has a reputation for being an exercise in inventing personas with names and stock photographs. That version produces a document and no decisions.

The useful version splits your existing customer list along lines you can actually act on, using data you already hold, and it usually takes an afternoon.

1. Customer segmentation basics: a segment must change what you do

The test for any split is simple. If two groups would receive the same message, the same offer and the same priority, they are not segments — they are categories.

That rules out most demographic segmentation for small businesses. Knowing that a third of your customers are under thirty-five changes nothing unless you would treat them differently.

So work backwards. Decide what different treatments you are capable of offering, then find the splits that deserve them.

2. Split by value first

The most useful division in almost every business: what each customer is worth.

Sort your customer list by total spend over a year and divide into quarters. The top quarter usually accounts for a large majority of profit, and they are frequently receiving exactly the same treatment as everyone else.

This split immediately suggests action: better service, earlier access, a named contact, a phone call rather than an email. None of it requires new data.

3. Split by recency second

Value tells you who matters. Recency tells you who is at risk.

Group customers by how long since their last purchase, relative to their own normal interval. Active, overdue, dormant, gone.

Cross that with value and you have the single most actionable grid available to a small business. A high-value customer who is overdue is the most urgent contact you have, and almost nobody has a process that identifies them.

4. Split by what they bought, not who they are

The third useful division is behavioural: which product or service they purchased first.

This predicts a great deal — what else they might want, what problems they will have, how often they will return, and how price-sensitive they are.

It is also immediately usable, because a message referring to what someone actually bought is relevant by construction, without knowing anything about them personally.

5. Keep the number of segments small

Four to six segments is workable. Twelve is a spreadsheet nobody uses.

Each segment needs a defined treatment, and treatments cost time to design and maintain. If you cannot say what you will do differently for a segment, merge it into another one.

Start with two — your top quarter by value, and everyone else — and add segments only when you have exhausted the actions available for the ones you have.

6. Write the definition down as a rule

Segments must be reproducible from the data, not assigned by judgement.

"Spend over X in the last twelve months." "No purchase in more than twice their median interval." "First purchase was a service rather than a product."

Written as rules, segments recalculate automatically as customers move between them, which is the point. Manually assigned segments are accurate on the day they are made and wrong within a quarter.

7. Act on one segment at a time

The output of this exercise is a list of possible actions, and the temptation is to start all of them.

Pick the segment where the gap between current treatment and appropriate treatment is largest — usually high-value customers receiving standard service, or high-value customers who have gone quiet.

Do that one properly, measure it, then move on. Segmentation projects that try to serve six segments simultaneously produce six half-implemented treatments.

8. Check that segments actually behave differently

The verification step, and it is regularly skipped.

Once segments are defined, compare their behaviour: repeat rate, average order value, margin, how they respond to the same offer. If two segments behave identically, the split is not real and maintaining it costs you effort for nothing.

Recalculate quarterly, and watch movement between segments. Customers dropping from the top quarter is a more urgent signal than the overall retention rate, and it only becomes visible once the segments exist.

Conclusion

A segment is only worth having if it changes what you do. Split by value, then by recency relative to each customer's own interval, then by what they first bought.

Keep to four to six segments defined as reproducible rules rather than judgements, act on the single biggest treatment gap before adding more, verify that segments genuinely behave differently, and recalculate quarterly so movement between them is visible.