Your team sells to everyone, and the results show it. Messages are generic, the product roadmap tries to please all buyers, and nobody can say which customers matter most. Customer segmentation is how you replace “everyone” with a few groups you can name, size and serve. This guide covers the bases of segmentation, how a segmentation study runs from objective to activation, how to judge a solution, and a worked example of picking a target segment by value. If you are newer to the basics, start with the four main types of segmentation in market research.
Key takeaways
- Customer segmentation analysis splits a market into groups that differ in needs, attitudes or behavior, so you can treat each group differently.
- A priori segments are defined before you look at the data, such as age bands or spend tiers. Post hoc segments are found in survey data with clustering.
- A segmentation study runs from a business objective through survey design, clustering, profiling and sizing to a typing tool that classifies new respondents.
- Judge every solution on six criteria: measurable, substantial, accessible, differentiable, actionable and stable.
- Segments pay off after the study, when trackers and customer surveys are tagged by segment and read with significance testing.
What is customer segmentation analysis?
Customer segmentation analysis is the work of dividing customers, or a whole market, into groups whose members are alike in ways that matter for a decision and different from the other groups. The point is action. Each segment should warrant its own target, positioning, product or service level.
There are two ways to form the groups.
- A priori segmentation. You choose the grouping rule before analyzing the data: three age bands, heavy versus light buyers, or customers in each region. It is quick and easy to explain, but it assumes you already know which variable separates people.
- Post hoc, data-driven segmentation. You measure many needs, attitudes or behaviors in a survey, then let a clustering method find groups of respondents who answer alike. It can find groups you did not expect, but it takes more design work and more judgment.
Most customer segmentation analysis in practice combines the two. A data-driven study finds the groups, then a priori variables such as age, region and spend describe them, so the business can find them in the real world.
What are the bases of customer segmentation?
In customer segmentation, the basis is the set of variables used to form the segments. It is the most important choice in the study, because the segments can only differ on what you feed in.
| Basis | Groups people by | Example variables | When it works |
|---|---|---|---|
| Demographic | Who they are | Age, income, household size | When needs really do follow life stage, and for media buying |
| Geographic | Where they are | Region, city size, climate | When availability, climate or regulation varies by place |
| Psychographic | What they value and believe | Attitude statements, lifestyle, personality | When brand choice depends on identity and values |
| Behavioral | What they do | Purchase frequency, channel, brand used, spend | When you want to grow, retain or win back buyers |
| Needs-based | What they want from the category | Importance of price, quality, convenience | When designing products, offers and positioning |
| Firmographic (B2B) | What the buying company is | Industry, company size, number of sites | When selling to businesses, often combined with buyer needs |
Psychographic segmentation explains motivation better than demographics do, but it needs carefully worded attitude batteries. Behavioral segmentation is often the easiest to act on, because behavior shows up in customer records and media targeting.
Needs-based segmentation is the classic basis for a full study. When needs are measured with plain rating scales, everyone tends to rate everything as important. A forced-choice method such as MaxDiff analysis gives sharper differences between people and cleaner inputs for clustering.
How does a customer segmentation study work, step by step?
A full customer segmentation study has more steps than most teams expect. Skipping the early ones is the most common reason a solution looks good on paper and is never used.
- Set the business objective. Write down the decision the segments must inform, such as which buyers to target with a new offer, or how to tailor service levels. The objective decides the basis.
- Choose basis variables. Pick the needs, attitudes or behaviors that relate to that decision. Keep the list focused, because every irrelevant item adds noise.
- Plan the sample size. Work back from the smallest segment you would act on. If you expect that segment to hold 15% of the market and want 150 respondents in it, you need 150 / 0.15 = 1,000 completes.
- Design the survey. Include basis questions, profiling questions (demographics, channels, brands used, spend) and outcome questions (purchase intent, satisfaction). Sample the whole category if you want market segmentation research, not only current customers.
- Prepare and standardize the data. Clean speeders and straight-liners. Then standardize the basis variables, because some people use the top of every scale and others the middle. Without this, segments form around how people answer rather than what they want.
- Run the clustering. Methods such as k-means, hierarchical clustering and latent class models each group respondents who answer alike. Our guide to cluster analysis compares them.
- Choose the number of segments. Produce several solutions, for example three to seven segments, and compare them (see below).
- Profile the segments. Crosstab each segment against everything you measured that was not a basis variable: demographics, channels, brands, spend and outcomes.
- Name the segments. Use short, plain names that describe what each group wants or does, such as “Time-pressed” rather than a made-up character.
- Size the segments. Report each segment’s share of the market and, where you can, its share of category spend.
- Build a typing tool. A typing tool is a short set of survey questions plus a scoring rule that assigns a new respondent to a segment. It lets you classify people in future surveys without repeating the full study.
- Activate. Brief the teams who will use the segments, add the typing questions to trackers and customer surveys, and agree which segments are the targets.
How do you choose the number of segments?
No statistic picks the number for you. Compare candidate solutions on three things. First, fit: how much tighter the groups get as you add a segment, which usually levels off at some point. Second, stability: run the analysis on two random halves of the sample and check that the same groups appear. Third, use: whether each segment has a clear story and a different action attached.
In our experience, the solution that wins on fit alone is often one segment too many. If two segments would get the same plan, merge them.
How do you evaluate a customer segmentation?
Clustering will return groups whether or not they are useful, so customer segmentation needs a separate quality check. Before you present a solution, run it through this checklist.
- Measurable. You can size each segment and describe it with data, not opinion.
- Substantial. Each target segment is large enough, in people or spend, to justify its own plan.
- Accessible. You can reach the segment through channels, media, sales teams or customer records.
- Differentiable. Segments respond differently to offers and messages, and the profile differences pass significance tests.
- Actionable. You can design a distinct product, message or service for each target segment.
- Stable. The segments hold up on a split sample and stay recognizable over a planning cycle.
Accessibility is the criterion most often failed. A segment defined only by attitudes, with no link to anything in your customer file or media plan, is hard to use.
How do you size segments and pick a target? A worked example
Here is an illustrative example, not a real study. A home coffee brand surveys 1,200 category buyers before launching a coffee subscription. A needs-based clustering produces four segments. The survey also asks annual coffee spend, and top-2-box purchase intent for the subscription concept.
| Segment | Respondents | Share of sample | Aged 18 to 34 | Buy coffee mostly online | Average annual spend | Purchase intent (top-2-box) |
|---|---|---|---|---|---|---|
| A: Price-driven | 384 | 32.0% | 41% | 38% | $120 | 123 / 384 = 32.0% |
| B: Brand loyalists | 264 | 22.0% | 22% | 25% | $260 | 53 / 264 = 20.1% |
| C: Time-pressed | 336 | 28.0% | 35% | 62% | $180 | 151 / 336 = 44.9% |
| D: Explorers | 216 | 18.0% | 58% | 54% | $150 | 82 / 216 = 38.0% |
| Total | 1,200 | 100% |
Age and online buying are profiling variables. They were not used to form the segments, so they show whether each group can be found and reached.
Step 1: compute value share
Segment size is not segment value. Multiply each segment’s respondents by its average spend to get its share of category spend in the sample.
| Segment | Respondents × spend | Value | Value share |
|---|---|---|---|
| A | 384 × $120 | $46,080 | 46,080 / 207,600 = 22.2% |
| B | 264 × $260 | $68,640 | 68,640 / 207,600 = 33.1% |
| C | 336 × $180 | $60,480 | 60,480 / 207,600 = 29.1% |
| D | 216 × $150 | $32,400 | 32,400 / 207,600 = 15.6% |
| Total | $207,600 | 100% |
Price-driven buyers are the largest segment (32.0% of people) but hold only 22.2% of spend. Brand loyalists are only 22.0% of people but hold the most spend (33.1%).
Step 2: weight value by purchase intent
Now ask how much of that spend sits with people who say they would buy the subscription. Multiply the intenders in each segment by the segment’s average spend.
- A: 123 × $120 = $14,760
- B: 53 × $260 = $13,780
- C: 151 × $180 = $27,180
- D: 82 × $150 = $12,300
- Total: $68,020
Time-pressed buyers hold $27,180 / $68,020 = 40.0% of the spend among likely subscribers, about double any other segment. Brand loyalists are valuable but happy with how they buy now, so the subscription is a weak fit for them.
Step 3: check the gaps that drive the decision
Is the Time-pressed lead on intent over Explorers (44.9% vs 38.0%) real? Use a two-proportion z test.
- Pooled proportion = (151 + 82) / (336 + 216) = 233 / 552 = 0.422
- Standard error = √(0.422 × 0.578 × (1/336 + 1/216)) = √(0.2439 × 0.00761) = √0.001856 = 0.0431
- Difference = 0.449 − 0.380 = 0.069
- z = 0.069 / 0.0431 = 1.60
That is below 1.96, so the intent gap is not significant at the 95% level. The case for Time-pressed as the primary target does not rest on intent alone. It rests on size and spend: the segment is larger, spends more and is just as reachable online. Explorers become a sensible secondary target.
How do you use customer segments after the study?
Customer segmentation earns its cost only if people use the segments for years, not weeks. Three habits make that happen.
Tag every tracker and survey by segment
Add the typing tool questions to your brand tracker, customer satisfaction surveys and concept tests. Each respondent is then assigned to a segment, so every future study can be read through the same lens. Without this step, the segmentation lives in one slide deck and fades.
Read results by segment with significance testing
Once surveys are tagged, run results as banner tables with the segments as columns. Test each difference before acting on it, because segment bases are smaller than the total and small gaps are often noise. Crosstabs and banner tables with weighting and significance testing make it practical to compare segments in every wave and to see when one starts to move.
Find out what drives each segment
Segments often value different things, so the levers that lift satisfaction for one group may not move another. Key driver analysis run separately by segment shows which attributes matter most to each. Teams that want an ongoing read can see what drives satisfaction and loyalty by segment and journey, rather than waiting for the next study.
When you plan offers or product lines for a target segment, TURF analysis helps you choose the combination of items that reaches the most people in that segment.
What are the common mistakes in customer segmentation?
- Too many segments. Eight segments may fit the data slightly better than five, but no marketing team can brief, target and track eight. Choose the smallest number that still gives each group a different plan.
- Segments nobody can find. Attitude-only segments with no profile on age, region, channel or customer records cannot be reached. Check accessibility before you present, not after.
- Profiling on the basis variables only. Segments always differ on the variables that formed them, so that proves nothing. The useful profile is on variables the clustering did not see: demographics, spend, channels and outcomes.
- Skipping the typing tool. Without a short classifier, you cannot tag future surveys, and the segmentation stops being used once the study ends.
- Never refreshing. Markets shift, and so do segments. Track segment sizes in your tracker each wave, and revisit the study when sizes or profiles drift.
How do you get started with customer segmentation?
Start with the decision, not the method. Write one sentence on what the segments should change, such as who you target with a new offer, and list the needs, attitudes or behaviors that would separate buyers on that decision. Then check what you already have: past surveys, a tracker with spend and channel questions, or customer records with purchase history may support an a priori cut today, while a full segmentation study is planned for the next budget cycle.