Customer segmentation is the practice of dividing a customer base into groups that buy differently — by demographics, geography, behaviour, psychographics or value — so that each group can be measured, targeted and sized on its own terms. For a marketing team, the output is a message. For a brand opening its next ten stores, the output should be something else entirely: a filter you can run across real trade areas.
The connection between the two is not obvious until you have signed a bad lease. A trade area is just a container; what fills it is people, and people are not interchangeable. Two sites with nearly identical population counts can forecast very differently because one holds a dense concentration of the group that actually spends with you and the other holds very little of it. Segmentation is how you learn the difference. Value-based segmentation is how you rank it.
Customer segmentation splits a customer base into groups that behave differently. The four classic bases are demographic, geographic, behavioural and psychographic; a fifth, value-based segmentation (usually RFM), ranks groups by what they are worth rather than who they are. For expansion, the useful output is not a persona deck. It is a geographic filter specific enough to score real trade areas, plus an honest read on how much of your revenue one segment is carrying.
What Are the Four Types of Customer Segmentation?
The four types are demographic (who customers are), geographic (where they are), behavioural (what they do) and psychographic (why they choose). Most brands use some combination of all four. A fifth base, value-based segmentation, sits on top of them by ranking groups according to what they are worth rather than describing them — which is the base expansion decisions actually run on.
| Base | Sorts customers on | Strongest at | Where it fails for site selection |
|---|---|---|---|
| Demographic | Age, income, household size, life stage | Sizing a market fast from data available everywhere | Two demographically identical neighbourhoods can perform nothing alike |
| Geographic | Where customers live, work and travel | Drawing trade areas and finding where demand already clusters | Describes where today’s customers are, not where tomorrow’s will be |
| Behavioural | Frequency, basket size, daypart, channel, loyalty | Finding your real repeat buyer instead of your average one | Lives in your own transaction data, so it cannot be observed in a market you have not entered |
| Psychographic | Values, attitudes, motivations, lifestyle | Explaining why one site beats its demographic twin | Hardest to source honestly and easiest to talk yourself into |
| Value-based (RFM) | Recency, frequency, monetary value | Ranking groups by contribution so you know which one to chase | Backward-looking: a segment worth a lot today may already be shrinking |
The four descriptive bases answer “who is this?” The value base answers “so what?” You need both: a profile with no value ranking tells you to chase everybody, and a value ranking with no profile tells you who matters without telling you where to find more of them. Psychographics are the hardest of the descriptive bases to source and the most likely to explain a demographic twin that underperforms, which is why they get their own treatment in our guide to psychographic segmentation.
How Do You Segment Customers?
Start from transactions, not personas. Build a customer-level table containing recency, frequency, average ticket and — wherever you can get it — home or work geography, then cut it into groups that differ materially on at least one of those measures. If a cut does not change spending behaviour, it is a label, not a segment.
- Use your own data first. Loyalty records, payment-card hashes and e-commerce accounts beat any purchased panel, because they describe people who have already chosen you.
- Cut before you cluster. Simple splits on frequency and ticket are defensible in a room. Unsupervised clusters are harder to argue for and easy to over-fit.
- Name groups by behaviour. “Weekday lunch regulars” survives scrutiny. “Urban achievers” does not.
- Attach geography to every segment. A segment you cannot locate on a map is useless for site selection, however elegant the profile.
- Re-cut annually. Segments drift. The group that carried you through your first ten stores may not be the one funding the next ten.
What is RFM analysis?
RFM analysis scores every customer on three things: recency (how long since they last bought), frequency (how often they buy) and monetary value (how much they spend). Each dimension is usually scored into quintiles, giving a three-digit score — a 555 is a best customer, a 111 has effectively left. It is the cheapest honest segmentation available, because it runs entirely on sales history you already own.
Its limit is that it describes behaviour without explaining it. RFM tells you the high scorers are worth chasing; it does not tell you what else is true about them, which is exactly what you need in order to find more of them in a city where you have no stores. That is why value-based segmentation is almost always paired with a descriptive base: value decides which group to target, demographics and psychographics decide where to look.
Why Is Your Biggest Segment Rarely Your Most Valuable?
Because segment value is the product of three numbers and headcount is only one of them. Annual value per customer is average transaction value multiplied by visit frequency; total segment value is that figure multiplied by how many customers sit in the group. A large, cheap, infrequent segment can lose to a small, expensive, loyal one on both measures at once.
Work it through. A group that is a third of your base, spending $12 a visit across 38 visits a year, is worth $456 per customer. A group two-thirds its size, spending $44 across 18 visits, is worth $792 per customer — and because the gap per head is so wide, the smaller group can out-earn the larger one outright. Count customers and you get one ranking. Count dollars and you get another. Only the second one should be allowed near a site decision.
There is a third number almost nobody puts in the deck: how hard the segment is to find. A high-value group scattered thinly across every market is a worse expansion target than a slightly less valuable group that concentrates in places you can identify on a map. Value per unit of acquisition difficulty is the ranking that matters when the output is a site list rather than a campaign.
Rank Your Segments by What They Are Worth
Put your own segments into the table below: share of the customer base, average ticket, visits a year, and how hard each group is to reach on a one-to-five scale. The ranker computes annual value per customer and total value per segment, re-orders them live as you change the basis, flags the best value-to-difficulty ratio, and shows how much of your revenue rests on the single biggest group. The pre-filled rows are illustrative sample data — overwrite them with yours.
The five rows below are illustrative sample data, not Locate benchmarks. Overwrite every number with your own, add segments, and watch the ranking reorder.
Annual value per customer is average ticket × visits a year. Difficulty is your own read on how hard that group is to reach in a new market: 1 is everywhere, 5 is scarce.
Loyalty members, unique payment cards, whatever you can actually count. It scales the dollar figures; it does not change the ranking.
Dollars a year the whole group is worth. The ranking a CFO recognises.
The share of revenue in one segment at which you would call the base dangerously concentrated. There is no universal line — set your own.
Read it as: what one average customer in the base you described is worth over a year. Every segment below either pulls that figure up or drags it down.
Weekend family trips carries 28.7% of modeled revenue. The marker sits at your 40% line.
Weekday lunch regulars is the segment to build the next site list around: $1,550,400 of modeled annual revenue against an acquisition difficulty of 2, the best value per unit of difficulty in your table. Note the split: Weekday lunch regulars is your largest group by headcount at 34.0% of the base, while Weekend family trips carries the most revenue at 28.7%. Counting customers and counting dollars give different answers here. Concentration looks survivable: the top segment carries 28.7% of modeled revenue, under the 40% line you set. The base is spread widely enough that one group shifting will not move the whole P&L.
Estimate only. The pre-filled segments are illustrative sample data, and every figure here is computed from what you enter — ticket, frequency, share and your own difficulty rating. It models a steady-state year with no churn, no seasonality and no cross-segment movement, so treat it as a ranking tool rather than a forecast.
How a Segment Profile Becomes a Site-Selection Filter
A segment profile becomes useful the moment you can express it as conditions a trade area either meets or does not. This is the step most segmentation work never takes. You are not looking for places with a lot of people. You are looking for places with an unusual concentration of one specific group, which is a completely different search.
The translation runs attribute by attribute. Demographic traits become thresholds measured inside a drive time rather than a radius. Daypart behaviour becomes a daytime-versus- nighttime population question and a read on commuter flow. Psychographic traits, which you cannot buy cleanly, get proxied — by the co-tenants already winning in a centre, by the brands your best customers also shop, by what people in that market search for. Search demand is one of the few segment signals you can read before you enter: Semrush maintains a keyword database spanning over 28 billion keywords across 142 geographic databases with city-level volume, which makes it possible to check whether the thing your best segment searches for is actually searched in a city where you have no stores.
- →Define the segment by behaviour you can measure: ticket, frequency, daypart, channel.
- →Express it as trade-area conditions: drive-time thresholds, not radius rings.
- →Proxy what you cannot buy: co-tenancy, search demand, analogs among your own stores.
- →Rank markets by segment density per trade area, not by raw population.
- →Check who already serves that segment there before you commit to a lease.
Run that filter across a market and the output is a ranked list of trade areas by segment density. It feeds directly into trade area analysis and sharpens the demographic screen described in our guide to demographic insights for site selection. One warning: wherever your segment is dense, somebody is usually already serving it, so read the field before you celebrate — that is the job of competitive intelligence.
Concentration Risk: When One Segment Carries the Revenue
Concentration risk is the share of total revenue that depends on a single segment. There is no universal danger line and we know of no published threshold worth quoting; as a working assumption we start worrying somewhere between a third and two-fifths, because past that point the portfolio is effectively a leveraged bet on one group continuing to behave as it does today. Set your own line and be explicit about it.
Expansion is where concentration either compounds or corrects. If every new site is matched to the same high-value segment, you are deepening the same bet in more markets, and a shift in that group’s habits — a competitor aimed squarely at them, a change in commuting patterns, a category trend — moves the entire P&L at once. The fix is not to abandon your best segment. It is to deliberately open some stores where a credible second segment is dense, underwrite those sites to a second forecast, and grade them as their own cohort rather than blending them into a portfolio average.
Why Is Customer Segmentation Important — and How Do You Know It Worked?
Customer segmentation is important because it converts an undifferentiated market into a ranked list of places where your actual buyer is concentrated. Without it you compare candidate sites on population and traffic counts, which quietly assumes everyone in a trade area is equally likely to spend with you. With it, you compare sites on the density of the group that demonstrably already does.
Segmentation earns its keep only if you grade it. Twelve to eighteen months after opening, split your stores into the ones that matched the target segment profile and the ones that did not, then compare actual revenue against forecast for each cohort. If the segment-matched sites do not outperform, one of two things is true: the segment is wrong, or your filter is not measuring what you think it measures. Both are fixable, and neither is discoverable without the grading step.
That loop — segment profile, trade-area filter, revenue forecast, signed lease, graded outcome — is the one Locate runs for multi-unit brands, with the analysis and the brokerage under one roof so nothing is lost in the handoff between knowing where to go and getting the deal done. If you want your own segments turned into a ranked market list, talk to us. Segment counts tell you who shops with you today. Segment value tells you who the next ten stores should be built around.
Common Questions
- What is customer segmentation?
- Customer segmentation is the practice of dividing a customer base into groups that buy differently — by demographics, geography, behaviour, psychographics or value — so each group can be measured, targeted and sized on its own terms. For a marketer it decides what to say to whom. For a growing retailer it decides something bigger: which group of people your next location should be chosen to sit among.
- What are the four types of customer segmentation?
- The four classic types are demographic (who customers are: age, income, household size), geographic (where they live, work and travel), behavioural (what they actually do: frequency, basket size, daypart, channel) and psychographic (why they choose: values, attitudes, motivations). A fifth base, value-based segmentation — usually run as RFM — sits on top of the others by ranking groups according to what they are worth rather than describing who they are.
- How do you segment customers?
- Start from your own transaction data, not from personas. Build a customer-level table with recency, frequency, average ticket and, where you can, home or work geography, then cut it into groups that differ materially on at least one of those measures. If a cut does not change spending behaviour, it is a label rather than a segment. Once the groups are real, attach a geographic signature to each one so it can be matched against trade areas.
- Why is customer segmentation important?
- Because it turns an undifferentiated market into a ranked list of places where your actual buyer is concentrated. Without segmentation you compare candidate sites on raw population and traffic counts, which treats every person in a trade area as equally likely to spend with you. With it, you compare sites on the density of the group that already does, which is the variable most likely to separate a store that hits its forecast from one that misses.
- What is RFM analysis?
- RFM analysis scores every customer on three dimensions: recency (how long since they last bought), frequency (how often they buy) and monetary value (how much they spend). Each dimension is typically scored into quintiles, producing a three-digit score where a 555 is a best customer and a 111 has effectively lapsed. It is the cheapest honest segmentation available because it needs no third-party data, only your own sales history.