Quick answer
A customer segment is a group of customers who share a definable trait, such as behaviour, location or company size, and who are expected to respond differently from other groups. It is useful only if you would act differently on it.
What a customer segment is
A customer segment is a group of customers who share a trait you can define and who you expect to respond differently from other groups. The traits can be who they are (country, company size), what they do (bought twice in 90 days) or what they want (price over speed). The idea goes back to Wendell Smith's 1956 argument in the Journal of Marketing that a market is not one mass: it is several groups with different demands, and a business should adjust its offer to them (Smith 1956).
A segment is the group. Segmentation is the act of dividing customers into those groups. The two words are often mixed up; see audience segmentation for the process side.
The test of a segment is not that it exists in your data. It is that you would treat its members differently from everyone else. If you would send the same message and make the same offer to two segments, they are one segment.
Four common bases for segments
| Basis | Groups customers by | Example rule | Strength | Weakness |
|---|---|---|---|---|
| Demographic or firmographic | Who they are | Company size 50–200 | Easy to find in a CRM | Weak at predicting what they will do |
| Geographic | Where they are | Visitors from Germany | Clear for language, tax and shipping | Rarely says anything about intent |
| Behavioural | What they do | Bought twice in the last 90 days | Predicts the next action best | Needs event data and a time window |
| Needs-based or psychographic | What they want and believe | "Wants the cheapest option" | Explains why | Needs research; hard to apply per person |
Most working segments combine two bases, for example "returning customers (behaviour) in the EU (geography)". The more bases you stack, the smaller each group, so stop when a group is too small to compare.
A working checklist for a useful segment
Before you keep a segment, ask five questions. Marketing textbooks phrase them differently; this is a working version.
- Can you identify members? A rule that a system can evaluate ("did not buy in 90 days") beats a feeling ("price-sensitive types").
- Is the group big enough? If you cannot get a stable conversion rate from it, it will not guide a decision: with a few dozen people, one extra order swings the rate by several points.
- Can you reach it? If you have no channel that reaches only these people, the segment is analysis, not action.
- Does it respond differently? This is the real test, and the next section shows how to run it.
- Is there a different action? Different message, offer, support path or product change. If not, merge.
A worked example: do the segments behave differently?
Sample data: 4,000 customers on the mailing list. Everyone gets the same 10%-off email. Segments are defined by rules, not guesses.
| Segment (rule) | Customers | Orders from the email | Response |
|---|---|---|---|
| Repeat buyers (2+ orders in 90 days) | 600 | 72 | 12.0% |
| One-time buyers (1 order ever) | 1,900 | 95 | 5.0% |
| Lapsed (bought before, none in 90 days) | 1,100 | 33 | 3.0% |
| Never purchased (signed up only) | 400 | 8 | 2.0% |
| Total | 4,000 | 208 | 5.2% |
Check: 600 + 1,900 + 1,100 + 400 = 4,000 and 72 + 95 + 33 + 8 = 208, so the overall rate is 208 ÷ 4,000 = 5.2%.
Now compare a split that sounds sensible but changes nothing. Split the same 4,000 by country: 2,000 in the US and 2,000 elsewhere, each with 104 orders. Both respond at 104 ÷ 2,000 = 5.2%. Country gives you two labels and no difference in behaviour for this offer, so it is not a useful segment for this decision. The behavioural split spans 12.0% to 2.0%, a sixfold range.
The decision rule: keep splits where response differs enough to change the action. Here repeat buyers may not need a discount at all, while lapsed customers might get a win-back message instead of the same coupon.
One caution about the table above: small groups give shaky rates. A rough 95% margin for a rate is 1.96 × √(p × (1 − p) ÷ n), the normal approximation described in the NIST handbook. Repeat buyers: 12.0% ± 2.6 points. One-time: 5.0% ± 1.0. Lapsed: 3.0% ± 1.0. Never purchased: 2.0% ± 1.4. The last two ranges (2.0 to 4.0 and 0.6 to 3.4) overlap, so on this sample you cannot tell them apart and might treat them the same.
Customer segment, persona, cohort and cluster
| Term | What it is | Made by |
|---|---|---|
| Segment | A group defined by a rule over data | A rule you write or a model you run |
| Persona | A fictional profile that stands for a segment | Research and writing |
| Cohort | People who share a starting event, such as signup week | Time of first event |
| Cluster | A group found by an algorithm without labels | Statistical model |
| RFM group | A segment from recency, frequency and monetary value | Scoring recent purchases |
A persona is a story you tell about a segment; it is not the segment. A cohort is defined by when people started, a segment by what they are or do now.
Common mistakes
- Too many segments. Twelve groups of 150 people each cannot be compared. Start with three to five.
- Labels without actions. "Enthusiasts" with no plan attached is a slide, not a segment.
- Static labels on moving customers. Behaviour changes, so rules that recompute are safer than a list built once.
- Mixing bases in one name. "Young urban bargain hunters" stacks three bases; you cannot tell which one drives behaviour.
- Judging a segment by its size alone. A small group of repeat buyers can be worth more than a large group of browsers.
- Treating anonymous visitors as customers. You can only segment people you can recognise across visits.
For a longer walkthrough, read the blog on customer segmentation analysis.
Customer segments in TapCub
Take the email example above and build it in TapCub.
- Define the rule. Open All-platform analytics › Users › Segments & tags and create a rule segment, for example "did purchase in 90 days at least twice". Rules can use did or did not do, user properties, first seen, active, membership in another segment and more. Preview the count, then save.
- Let it refresh. Rule segments refresh daily (or on demand), and each segment keeps a 30-day size trend, so you can see Lapsed shrink after a win-back message.
- Use RFM when you have orders. An RFM segment takes a transaction event and an amount property and sorts people into 10 buckets (champions, loyal, potential loyalists, new, promising, needing attention, about to sleep, at risk, cannot lose, hibernating), with movement between buckets recorded.
- Analyse by segment. Use Analyze to apply a segment as a cohort in funnels and retention, and compare conversion between segments.
Limits to know: anonymous visitors have no profile, so they are never segment members; sign people in with tapcub.login first. TapCub does not send email or run campaigns, so you take a segment into your own sending tool; TapCub only measures the result.
Sources
- Wendell R. Smith, Product Differentiation and Market Segmentation as Alternative Marketing Strategies, Journal of Marketing, 1956.
- NIST/SEMATECH, e-Handbook of Statistical Methods: comparing two proportions, for comparing group response rates.
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Product Experience Specialist
Explores usability, user journeys, and feature adoption through practical examples of onboarding and everyday product use.