Insights
Segmentation: Sorting Versus Understanding
Six standard approaches, why they fail silently, and two alternatives that predict behaviour
Most segmentation sorts customers. Very little of it explains them.
The distinction matters because a segmentation that sorts produces clean-looking groups that behave identically in the data and completely differently in the market. It fails silently, while producing a very convincing slide.
The six standard approaches
All six are legitimate. The question is never which is correct, but which actually predicts behaviour in your category.
| Approach | What it divides by, and where it works |
|---|---|
| Geographic | Region, climate, urban or rural. Works where physical access or climate genuinely changes need. |
| Demographic | Age, income, household, education. The default, and the one most often used past its usefulness. |
| Psychographic | Attitudes, values, lifestyle, personality. Harder to measure, better at explaining choice. |
| Behavioural | Purchase frequency, occasion, loyalty status, usage rate. What people do rather than who they are. |
| Product | Which variant, format or specification they buy. Often reveals segments nobody designed for. |
| Sales channel | Where they buy. In many categories this predicts more than any customer attribute. |
Why demographic segmentation fails silently
It is a proxy. It works where the proxy happens to hold, and gives no warning when it stops.
Consider a market where a large majority of residents are expatriates, drawn from dozens of countries, across an enormous range of income and length of stay.
Segment that market the way you would segment a domestic one – age, income, postcode, life stage – and the groups look immaculate and predict nothing. Two households can share an age bracket, an income band and a suburb and shop entirely differently, because one is three years into a posting and remitting money home, and the other has been there fifteen years and raised children in the country.
What actually predicted behaviour was closer to: how long have you been here, do you expect to stay, and where does your money go when it is not spent here?
The transferable point has nothing to do with that region. It is that every demographic segmentation rests on an assumption that demography and behaviour move together – and that assumption is worth testing rather than inheriting.
Alternative one: segment by occasion
Instead of asking who the customer is, follow one customer through a full day and ask what they need at each point.
A food business mapped a working consumer’s day hour by hour: the wake-up coffee, something eaten on the commute, arrival at work, the mid-morning break, a quick lunch, the afternoon slump, leaving work, an evening meal with friends, then a heavier dinner at home.
Each of those is a different occasion with a different need, a different price tolerance, a different format requirement and a different competitive set. The same person is in several segments during one day.
- It reveals gaps in the range. Occasions where you have nothing to sell become visible immediately.
- It sets the format, not just the message. Portable, single-serve, shareable, sit-down – the occasion decides.
- It identifies the real competitor. At eleven in the morning you are not competing with your category. You are competing with whatever else is nearby and quick.
- It works beyond food. Any category consumed in moments rather than as a lifestyle statement – media, transport, financial products, software – behaves the same way.
Alternative two: segment by role in a value chain
In business-to-business markets, the useful division is almost never demographic. It is what each party does.
A technical product mapped its market by function rather than by firmographics, and each group had a completely different reason to buy.
- Organisations that collect samples but send them elsewhere for analysis. Benefit: add analysis to your own services and keep the revenue.
- Organisations that already have an internal capability. Benefit: faster turnaround and higher capacity from the same team.
- Specialist analysis providers. Benefit: throughput, and relief from a shortage of qualified technicians.
- Universities and research institutes. Benefit: teaching value now, and a pipeline of trained future customers.
- Large engineering and consulting firms. Benefit: cost reduction across a portfolio of projects.
Two refinements worth stealing
The same exercise identified two groups that most segmentations would have missed entirely.
- The awareness-only segment. Large contractors who would never buy the product themselves, but whose expectations shape whether their suppliers need it. You market to them knowing they will not purchase.
- Channel partners as a segment. Resellers and training organisations have their own needs, their own economics and their own reasons to say yes. Treating them as a route rather than as a customer is how channels fail.
The test
Whichever approach you use, one question separates a segmentation that works from one that merely sorts.
Do members of this segment behave differently from members of the others – and can I name the behaviour?
If the answer is a description of who they are rather than what they do, you have a filing system rather than a strategy. It will produce clean charts and disappointing campaigns, and it will take two years to find out.