← Blog

Customer segmentation for researchers

6 November 2026 · 5 min read

Illustration of customer segments

Ask five people in the same company who their customers are and you will get five different answers. The marketer describes demographics, the product manager describes power users, the founder describes a persona from three years ago. Customer segmentation exists to replace that muddle with a shared, evidence-based map: who you serve, how the groups differ, and which differences actually matter.

It is also one of the places where market research and UX research genuinely meet. Market researchers usually build the segmentation; UX researchers inherit it, and they are often the first to notice when a segment dissolves on contact with a real participant. This guide is for both sides: what good customer segmentation looks like, and why so much of it ends up in a drawer.

Three ways to slice: demographic, behavioural, needs-based

Demographic segmentation groups people by who they are: age, income, location, company size, industry. It is the easiest kind to build, because the data already sits in your CRM, and the weakest kind to act on. Two people with identical demographics can want completely different things from the same product, and the same person can be a bargain hunter in one category and a premium buyer in the next.

Behavioural segmentation groups people by what they do: how often they buy, which features they use, what channel they arrived through, whether they churn. It is far more predictive, because past behaviour is the best available signal of future behaviour. Its limit is that it captures the what without the why, so it can tell you a group exists but not what to build for it.

Needs-based segmentation groups people by the underlying problem they are trying to solve. It is the hardest to build, because needs do not sit in a database waiting to be exported, and it is the most useful for product and proposition decisions. If the framing sounds familiar, it is the same instinct behind jobs to be done: people who share a job have more in common, commercially, than people who share a birth year.

The segmentations that hold up usually blend all three. Needs define the segments, behaviour validates them, and demographics help you find and reach them.

Segments are not personas

These two get conflated constantly, and the confusion causes real damage.

A segment is a quantified group: mutually exclusive, measurable, and sized. You should be able to say roughly what share of the market it represents, and work out which segment any given customer belongs to.

A persona is a qualitative portrait: a memorable, humanised character that helps a team design and write for one kind of person.

They work best as a pair. A persona without a segment behind it is fiction with a name. A segment without a persona is a spreadsheet nobody remembers in a design review. Build the segments with rigour, then wrap the most important ones in personas the team can hold in their heads.

When customer segmentation research is worth it

A proper segmentation study is a serious investment, so it should be commissioned like one. It earns its cost when the differences between groups would genuinely change a decision:

  • You are choosing which audience or market to prioritise
  • Pricing or packaging needs to differ by group
  • Messaging that lands with one audience is clearly failing another
  • The roadmap is being pulled in incompatible directions by different user types

Keep it light, or skip it entirely, when the product serves one core need, or when the organisation has no appetite to treat different customers differently. A segmentation nobody intends to act on is expensive shelf decoration.

The classic failure: segments nobody can action

Most failed segmentations die the same death. The study produces six beautifully named clusters, "the Aspirational Explorer", "the Pragmatic Optimiser", and then someone in marketing asks the fatal question: how do I know which one I am talking to?

A segmentation earns its keep only if the segments are:

  • Identifiable · you can tell which segment a real customer sits in, from data you already hold or a handful of classifying questions
  • Substantial · large or valuable enough to justify different treatment
  • Reachable · you can get to them through some channel
  • Distinct · they would respond differently to a different product, price, or message
  • Actionable · someone owns a decision that changes depending on the segment

Attitude-only clusters fail the first test most often. If a segment is defined purely by how people feel, and those feelings do not correlate with anything observable, the segmentation can never leave the report.

How researchers actually build one

The pattern that survives is qual then quant. Interviews and field visits surface the needs, the jobs, and the language, giving the segmentation qualitative ground to stand on. A survey across a proper sample then measures those needs and behaviours at scale, and cluster analysis groups respondents statistically. Finally, a short set of golden questions is derived, so future customers can be classified without rerunning the whole study. If the interplay between those two modes of evidence feels fuzzy, our guide to qualitative and quantitative research walks through it.

Segmentation has a reputation problem because so many segmentations were built as decks rather than decisions. Treat it as a tool with one job: letting different customers be treated differently, on evidence. Start from the decisions the organisation is actually willing to make, build on needs and behaviour rather than demographics alone, and give every candidate segment the same two-part test: can we find these people, and will we do anything differently once we have? When the answer is yes, segmentation is some of the highest-leverage research there is.


People of Research is a community and newsletter for market researchers, UX researchers, and the insight people who help teams see their customers clearly. Join us and get the good stuff in your inbox.

Join 7,500+ researchers, free

One email a week, a global community, and events near you. No noise, just the people and ideas that matter.