The Ground Rule
Start With Real People
A persona without research is a guess wearing a name tag.
Most personas are written in a conference room. Someone says, "Our users are probably like this." Everyone nods. A name gets picked, a stock photo gets pasted, and the guessing is over. Except it isn't. The guessing has just been laminated.
Research fixes this. Not a five-figure study. Not a lab. Research means going to where your users are and paying attention — because the whole point of a persona is to understand real people well enough to serve them. That's the business. Everything else is decoration.
Here's the toolkit. Five methods. Most cost nothing but attention.
| Method | What it finds | Sample | Time |
|---|---|---|---|
| Interviews | Motivation, context, real language | 5–8 per segment | 1–2 weeks |
| Surveys | How common a pattern is | 100+ responses | 3–7 days |
| Analytics mining | What people actually do | All users | An afternoon |
| Ticket archaeology | Recurring pain, verbatim words | 50–200 tickets | 2–4 hours |
| Field observation | Environment, workarounds | 3–6 sessions | 1–3 days |
Method 1
User Interviews
One honest conversation beats a hundred assumptions.
If you do one thing on this page, do this. Sit across from a user — in person or on a call — and ask about their work. Thirty minutes. No pitch, no demo. Just curiosity.
The magic question is "Tell me about the last time you…" Not "what do you usually do" — people idealize their usual. Ask about a specific recent instance and they relive it. The workarounds come out. The frustration comes out. The moment they almost quit comes out.
Steve Portigal, whose Interviewing Users is the standard text on the craft, teaches that the first answer is rarely the real one. Someone says "it was frustrating." Don't nod. Ask "frustrating how?" Then "what did you do next?" The insight lives two follow-ups deep.
Recruiting is simpler than it looks. Email your users a personal note and a small gift card. If you don't have users yet, go where the problem lives — communities, forums, LinkedIn. Avoid friends and coworkers. They'll be kind. Kindness is noise.
Don't write a script; write a guide. Five to seven open questions, then let the conversation breathe. We've collected the exact prompts that work — grouped by goals, frustrations, and context — in our library of persona interview questions. Steal them. That's what they're for.
Record everything (with permission) and write down exact words. Users' verbatim language becomes your persona's quotes, your UI labels, and your marketing copy. Paraphrase kills it.
Method 2
Surveys
Surveys count what people say. They don't watch what people do.
That's the limitation and the use, in one sentence. Jakob Nielsen's first rule of usability is blunt: don't listen to users — watch what they do, because self-reports are unreliable and predictions are worse. A survey that asks "would you use this feature?" collects polite fiction at scale.
So use surveys for the one thing they're great at: counting. Run them after interviews, never instead of them. Interviews find the pattern. Surveys tell you how common it is. "Six of eight people I talked to dread the Monday report" is a hunch. "71% of 240 respondents rank reporting as their top time sink" is a persona field with receipts.
Three rules keep survey data honest:
Under ten questions. Long surveys select for patient people, and patient people are not your persona.
Ask about the past, not the future. "How many times last week did you export a report?" beats "how often do you export reports?" and demolishes "would you pay for faster exports?"
One question, one thing. "How satisfied are you with speed and design?" collects noise. Split it.
Method 3
Analytics Mining
The data never flatters you. That's the gift.
In interviews people tell you what they think they do. Analytics show you what they did. The gap between the two is where the best persona insights hide — so mine your product data before and after you talk to people, and investigate every contradiction instead of picking a winner.
Usage clusters. Daily power users, weekly check-iners, signed-up-and-vanished. These clusters are often your persona segments, sitting in plain sight.
Drop-off points. Where do people stall? Funnels and heatmaps surface friction users are too polite — or too resigned — to mention.
Feature adoption. You believe everyone uses the advanced filter. The data says 8% do. Now you know something true about who your users really are.
Session recordings. Watch five. Just five. The hesitation, the wrong click, the back button, the pause before giving up. Five recordings teach more than any dashboard.
Remember what analytics can't do: explain. The data says what. Only humans say why. That's why this method never stands alone.
Method 4
Support-Ticket Archaeology
Your support inbox is research somebody already paid for.
Every ticket is a user, unprompted, telling you exactly where your product failed them — in their own words, with their own emotion, at the moment it happened. No recruiting. No scheduling. No performance for the researcher. Dig through the last three months like an archaeologist: you're not reacting to individual complaints, you're looking for the civilization underneath.
Recurring themes. One person confused by the export feature is an anecdote. Thirty people, three months, different words, same wall — that's a persona frustration.
Who hits which wall. New users file different tickets than veterans. Admins ask different questions than end users. Those clusters sketch your segments before you've done a single interview.
The words themselves. You call it "the admin console." They call it "that settings page." Their vocabulary belongs in your persona verbatim — it's what makes the finished thing ring true in good persona examples.
Extend the dig beyond your inbox: app-store reviews, G2 reviews of your competitors, Reddit threads about the problem you solve, your sales team's call notes. Five-star reviews are especially rich — they tell you why people love a product, which is the job they're hiring it to do.
Method 5
Field Observation
People can't describe water. They're swimming in it.
Your user won't tell you they flip between your app and a spreadsheet forty times a day, because to them that's just how work works. They won't mention the sticky notes ringing the monitor, the Slack interruption every three minutes, or the fact that your app only lives on their second screen. They don't see it anymore. You have to.
The method is called contextual inquiry: watch a user do their real work in their real environment, and ask questions as an apprentice, not an examiner. Nielsen Norman Group's guide to contextual inquiry covers the mechanics; the posture matters more than the protocol. Watch. Don't coach. When something surprises you, ask "why did you do it that way?" and then be quiet.
Three sessions will show you things twenty interviews miss. Can't travel? Ask for a screen share: "Do your normal work for twenty minutes while I watch." You lose the desk, keep the workarounds.
Hunt for two things above all: workarounds — the duct-tape solutions users built where your product falls short, each one an unmet need with a prototype attached — and context switches — every exit to another tool, which maps the real ecosystem your persona lives in.
How Much Is Enough
Sample Sizes and Saturation
Saturation is when the stories start to rhyme.
Around the fifth or sixth interview, something happens. Different people, different words, same underlying story. That's saturation — the point where new interviews stop changing what you know. It isn't just folklore. In a widely cited study, Guest, Bunce, and Johnson coded sixty in-depth interviews and found the large majority of themes appeared within the first six, with saturation arriving by around twelve.
Treat that as a planning number, not a law. NN/g's review of interview sample sizes notes there's no academic consensus, and that interviews often need more participants than the famous five-user usability test because you're chasing broader questions. Practical rules that hold up:
Five to eight interviews per persona segment. If interview seven still surprises you, keep going — or you may be talking to two segments, not one.
100+ survey responses before you quote a percentage in a persona.
Three observation sessions minimum. One is an anecdote. Three is a pattern.
And the floor: one. One real conversation this week beats twelve perfect ones next quarter. Don't let the ideal sample size become the reason you have no sample at all.
From Notes to Persona
Synthesis: Turning Research Into a Persona
Research isn't finished until it changes a decision.
A pile of transcripts is not a persona. Synthesis is the move from "what we heard" to "what we know" to "what goes on the card." Cluster your observations by theme — sticky notes, a spreadsheet, whatever works — and keep only the patterns that showed up more than twice. Then map each pattern to a specific persona field. Like this:
Three rules for the synthesis step:
Don't average. If half your users are experts and half are beginners, you don't have one "intermediate" persona. You have two personas. Averages erase the very patterns you dug for.
Let the data win. Good research should overturn at least one thing you believed. If every finding confirms the roadmap, look harder.
Label your confidence. Fields backed by twelve interviews are evidence. Fields you filled from instinct are hypotheses — fine to include, but mark them, the way a proto-persona marks all of its assumptions.
Then shape the patterns into a document, following the step-by-step persona creation process or our free persona template. Or skip the blank page entirely:
Build a persona right now — free, no signup.
Paste in what your research found. Get a structured, shareable persona in minutes.
Open the persona creator →Questions
Persona Research FAQ
How many user interviews do you need to build a persona?+
Plan for five to eight per persona segment. Research by Guest, Bunce, and Johnson found most themes surface within six interviews and saturation arrives by around twelve. Three honest interviews beat zero perfect ones — add more only while new interviews keep surprising you.
What is the best research method for creating user personas?+
User interviews. Nothing else delivers motivation, context, and real language in one sitting. Then triangulate: analytics and support tickets check what people actually do against what they said; surveys test how widely the pattern holds.
Can you create personas without doing research?+
You can draft a proto-persona from team assumptions — a legitimate starting point if you label every field as a hypothesis. Just don't let it harden into folklore. Schedule the first interviews the same week.
What is the difference between qualitative and quantitative persona research?+
Qualitative methods (interviews, observation) explain why people behave as they do — small samples, studied deeply. Quantitative methods (surveys, analytics) measure how many share a behavior — large samples, measured shallowly. Personas need both: one writes the story, the other proves it's common enough to build for.
How long does persona research take?+
A useful first pass takes about a week: a few hours mining tickets and reviews, an afternoon in analytics, five short interviews. A thorough pass — a dozen interviews, a validation survey, a few observation sessions — takes two to four weeks part-time.
Here's the whole method in one sentence: watch what people do, ask why they do it, count how many share the pattern, and write down only what you can trace to a real human.
Start this week. Read fifty tickets. Book one interview — the questions are ready. Then build a persona right now — free, no signup — and give your team a user they can finally agree on. The people you serve are out there, working around your product's rough edges at 9pm on a Sunday. Go meet them.