On this page
Useful user segmentation is based on goals, motivations and how people decide. Age and city are only supporting detail, because people of the same age may use your product for completely different reasons. After reading this guide, you'll be able to use interviews to find segmentation dimensions, copy the interview guide and persona card template as is, and use behavior data to test whether the segments you found actually exist.
When to use it
- As the user base grows, people on the team describe "who our users are" in different ways
- Existing personas only cover age, city and spending, so neither product nor marketing can use them
- You're planning differentiated features, pricing or marketing and need to know how many types of users to plan for
- Feedback on the same feature is polarized across users, and you suspect different groups are behind it
Why segmenting by age and city isn't enough
Demographics are easy to collect and convenient for ad targeting, but they describe who people are, not why they make the choices they do. Two people who are both 30 and live in the same city may sign up for the same online course: one wants to earn a certification within six months to change jobs, the other just wants to learn something new on weekends. What they value, what they compare it to and how fast they decide are all different. Segmenting by age often produces conclusions like "people in every age group care about price," which doesn't help anyone make a decision.
Good segmentation meets three tests: people within a segment make decisions in similar ways; the differences between segments map to different product or marketing actions; and every segment can be found and selected in your data.
Segment by goals, motivations and how people decide
Interviews should clarify the four dimensions below. The examples in the table are illustrative and show what each dimension is asking about.
| Dimension | What to find out | Example |
|---|---|---|
| Goals | What they want to get done with the product | Rush restocking on weekdays, planned shopping on weekends |
| Motivation | Why now, and what triggered this choice | Had a baby, moved, changed jobs |
| How they decide | Who was involved in the decision, what they compared, and how long it took | Decided together as a couple, compared only two options, ordered the same day |
| Concerns | What would make them give up | Over budget, worried about after-sales support, too hard to learn |
The Jobs to be Done lens
Jobs to be Done is an idea popularized by Clayton Christensen: users don't buy products, they "hire" them to get a job done. His well-known example is commuters buying milkshakes in the morning. What they wanted was something they could hold in one hand, eat slowly on the drive, and that would keep them full until lunch. The competition was bananas and bagels, not other milkshakes.
Segmenting through this lens asks what job users are trying to get done. The same product is often hired by different people for different jobs. Those jobs are candidate segmentation dimensions, and they also remind you that your real competitors may not be in the same category.
Who to interview
- Select by behavior, not just demographics: include some heavy users, light users, recently churned users and users who only use one feature
- Cover different stages of the decision: users who just signed up, are comparing options, have paid, and have renewed
- Deliberately include users whose usage is clearly different from most people's; they often point to an overlooked group
- Plan about 12 interviews per candidate segment (see Guest, Bunce & Johnson, 2006), and add interviews if a new segment emerges during analysis
- [Opening] Thanks for joining. Today I'd like to understand how you use our product and how you chose it in the first place. There are no right or wrong answers.
- [Opening] Tell me a bit about yourself: what's the rhythm of your work and daily life like?
- [Core topic: goals] The last time you used it, what were you trying to get done?
- [Core topic: goals] If you didn't have it, how would you handle that?
- [Core topic: motivation] What got you looking for this kind of product in the first place? What was going on in your life at the time?
- [Core topic: how they decide] From first learning about it to deciding to use it, what was the process like? What other options did you compare?
- [Core topic: how they decide] Who else was involved in the decision? Whose opinion mattered most?
- [Core topic: concerns] Was there anything that almost made you give up on it? How did you get past that?
- [Core topic: usage] How many times a week do you use it these days? Usually at what time, and where?
- [Probes] You said it's "convenient." Can you give me one specific example?
- [Probes] What matters most to you here? If you could keep only one thing, what would it be?
- [Probes] Do you know anyone who uses it differently from you? What's the difference?
- [Closing] If you were telling a friend about it, how would you describe what it's for?
- [Closing] Is there anything you think I should have asked but didn't? Thank you.
A sample dialogue
Below is a sample interview for a grocery delivery app. The product and all content are fictional.
The same person makes decisions in two different ways on weekdays and weekends: on weekdays she looks at delivery time, on weekends at price. Don't rush to label her "not price sensitive." A more accurate dimension is the usage context: "rush restocking" versus "planned shopping." It also shows that the unit of segmentation is sometimes a situation, not a person.
Find segmentation dimensions in your interviews
- Annotate each interview: for every participant, write one sentence each on their goal, trigger event, who was involved in the decision, what they compared against and their biggest concern.
- Look for the dimension with the widest split: whichever factor varies most across participants, and leads to different behavior, is a candidate dimension.
- Draft the groups: use one or two dimensions to sort participants into 3 to 5 groups. With too many groups, you can't treat each one differently.
- Check consistency within segments: people in the same segment should make decisions in similar ways. If a segment is still split, try a different dimension.
- Check whether it's actionable: does each segment map to different product or marketing actions, and can you find the matching people in your behavior data?
- Write persona cards: one per segment, with quotes and counterexamples.
- Segment name: name it after the job they're trying to get done, such as "office workers restocking in a rush," not something like "White-collar A"
- Job to be done:
- Trigger moment: what situation makes them start using or start looking
- How they decide: who's involved, what they compare, how long it takes to decide
- Values most: one or two things
- Biggest concerns: one or two
- Typical behavior: traits visible in behavior data, such as usage frequency, time of day and most-used features
- Representative quotes: two or three, with participant IDs
- Counterexample: one quote from a participant in the same segment that doesn't fit the persona
- Count in interviews: X / N people (shows only that this type exists, not its share overall)
- Share in the data: from querying typical behavior, about X%
- Actions: how product, marketing or service should treat this segment differently
How to validate with behavior data
Segments from interviews are only hypotheses. Interview samples are small and selected, so the relative size of each group doesn't represent your overall user base. Test them with behavior data before putting them to use.
- Translate each persona's typical behavior into criteria you can query, such as ordering on weekday evenings and choosing next-morning delivery
- Check each segment's share of all users to confirm it isn't a niche group that only showed up in interviews
- Compare retention, average order value, conversion and other metrics across segments. The segmentation only means something if the differences are clear
- After making differentiated changes by segment, look at metric changes per segment to confirm the changes work for the target group
- After user acquisition or the product changes significantly, recheck each segment's share
Common mistakes
Common mistakes
- Defining segments before you interview: you'll only find the groups you expected. Interview openly first and let the dimensions come out of the answers.
- Treating demographics as motivation: "Young people care more about price" may just reflect income differences. Probe for the specific trade-offs.
- Personas that are too neat: if you include only quotes that fit, readers will assume the boundaries between groups are clear-cut. Include counterexamples.
- Too many segments: you can't treat seven or eight groups differently. Three to five is usually enough.
- Using interview ratios to represent the market: ratios only show that such people exist. Use data to measure size.
- Filing personas away once they're done: every persona should map to specific actions and be checked against data regularly.
How to do it in heari
In heari, you can select participants by usage frequency, most-used features and churn status using behavior data from ThinkingAI Agentic Engine. AI drafts the study plan and interview guide, asks follow-up questions about goals, motivations and decision process based on each answer, and runs many interviews at once. The report groups findings by theme and shows how many people in each segment mentioned each one. Each finding comes with quotes and counterexamples, and the research lead can edit or reject it. Once the segments are in use, you validate them with behavior data.
FAQ
What's the difference between personas and user segmentation?
Segmentation is the rule for dividing users into groups; a persona describes each group. You define segmentation dimensions first, then write a persona for each segment. Personas without a segmentation rule can't be mapped to specific users in your data.
How many interviews do you need for user segmentation?
Segmentation research has to cover different groups, so it usually needs a larger sample than a single-topic study. One workable approach is to interview 20 to 30 people first to find candidate segments, then bring each segment up to about 12 interviews. This is a rule of thumb; the right number depends on how different your users are from one another.
Can Jobs to be Done and traditional personas be used together?
Yes. Jobs to be Done helps you find segmentation dimensions, namely the job users are trying to get done; personas then describe the people behind each job, including the trigger moment, how they decide and their concerns. Demographics such as age and city can be added to a persona as supporting detail.
How often should you update your user segments?
Revisit your segments whenever your product, pricing or user acquisition changes significantly, for example after entering a new market or running a large ad campaign. If nothing major has changed, check each segment's share and the differences in key metrics against behavior data every six months.
