# Buyer Persona: Build It from Evidence, Not Assumptions Build a useful buyer persona from interviews, customer quotes, and observed behavior with a practical, evidence-based method and template. - Canonical URL: https://www.harmate.com/en/blog/buyer-persona-build-it-from-evidence-not-assumptions - Author: Harmate Team - Published: 2026-08-24 - Updated: 2026-08-24T01:52:59.884635+00:00 - Language: en ## Content # Buyer Persona: Build It from Evidence, Not Assumptions A useful buyer persona is not a creative biography of your “ideal customer.” It is a revisable synthesis of situations, goals, barriers, alternatives, and decision criteria observed among real people. A name and photo may help a team remember the profile. They prove nothing. The persona becomes useful when every important characteristic can be traced back to interviews, open-ended responses, observed behavior, or other identifiable evidence. Start with the template below. The rest of this guide explains how to fill it without turning a few intuitions into false certainty. ## An evidence-based buyer persona template Leave a field blank when you do not yet have enough evidence. | Field | Question to answer | Evidence to keep | |---|---|---| | Decision | What decision should this persona support? | Team brief and explicit hypothesis | | Triggering situation | When does the need become concrete? | Accounts of recent experiences | | Desired outcome | What is the person actually trying to achieve? | Repeated wording and real examples | | Barriers | What slows, blocks, or stops progress? | Incidents, objections, and trade-offs | | Current alternatives | What do people do instead today? | Tools, processes, and workarounds mentioned | | Choice criteria | What moves the decision forward or stops it? | Comparisons, trade-offs, and stated reasons | | Internal variation | When does this profile behave differently? | Counterexamples and divergent situations | | Source quotes | Which exact words should remain visible? | Contextualized, reviewable excerpts | | Limits | Who and which situations are not covered? | Recruitment gaps, missing data, uncertainty | | Review date | When should the profile face new evidence? | Version and owner of the next review | This structure is less decorative than a template filled with age, hobbies, and favorite brands. It is also more actionable: every field serves a decision and can be challenged. ## What a buyer persona is — and what it cannot prove A persona represents a group of people who share characteristics relevant to a marketing, product, or sales decision. French public business agency [Bpifrance Création](https://bpifrance-creation.fr/encyclopedie/letude-marche/outils-danalyse/creer-persona-marketing-3-etapes) recommends grounding personas in surveys, interviews, research, and real data before compiling the profile. The main benefit is compression: a persona turns a complex research corpus into something a team can remember and use. A Delphi study with experts identified communication and focus among the benefits personas can bring to design processes ([Miaskiewicz and Kozar, 2011](https://doi.org/10.1016/j.destud.2011.03.003)). But a polished card does not become representative because it looks credible. It does not prove: - how large the segment is; - how common a need is across the market; - willingness to pay; - that one characteristic caused a behavior; - that any real person combines every attribute on the page. A well-known methodological critique argues that highly specific personas can be difficult to connect to an identifiable population or falsify with data ([Chapman and Milham, 2006](https://doi.org/10.1177/154193120605000503)). This does not make every persona useless. It means the level and boundaries of certainty should remain visible. ## Why personas quickly turn into stereotypes The problem often begins with the template. An “age” box demands an age even when age has no observed effect on the decision. A “personality” box encourages the team to infer stable traits from a few impressions. A precise stock photo then gives visual coherence to a combination the evidence may never have shown. Four mistakes are especially common. ### Filling every field Missing evidence becomes a plausible detail. The document looks complete but mixes facts, hypotheses, and decoration. ### Treating demographics as mechanisms “Age 35–45, urban, senior manager” describes a possible population. It does not explain why someone switches tools, delays a purchase, or chooses an offer. Demographics belong in the persona only when they change the decision at hand and the evidence supports that relationship. ### Turning one quote into a collective opinion A striking quote can expose a mechanism or provide accurate language. It does not measure frequency. Keep it as situated evidence, not a slogan attributed to an entire segment. ### Smoothing away contradictions Two people may seek the same result under opposite constraints. Averaging them can produce a persona that accurately describes neither. A recent scoping review of personas in co-design reported both their value for keeping attention on consumers and a risk of biased or less authentic profiles when construction is subjective ([scoping review, 2025](https://pmc.ncbi.nlm.nih.gov/articles/PMC12054091/)). The review concerns healthcare; it illustrates a methodological tension rather than providing a universal marketing estimate. ## How to create a buyer persona you can verify ### 1. Write the decision before the portrait A persona created “to know our customers better” has no boundary. Choose a concrete decision instead: - select the main message for a landing page; - prioritize an objection to address in the offer; - decide which onboarding journey to design first; - define the criteria for a concept test. The decision determines which data matters. Preferred information channels might matter for a campaign and be irrelevant to redesigning an expense approval process. ### 2. Collect situations, not adjectives Ask about recent experience: the last time the problem happened, what the person tried, what failed, the consequence, and the final trade-off. “Are you innovative?” invites a flattering identity. “Tell me about the last time you changed tools” produces a sequence the team can examine. Our [semi-structured interview guide](/en/blog/semi-structured-interview-guide-method-steps-and-15-examples) covers framing, follow-up questions, and common wording mistakes. For a broader research sequence, see how to [listen before you measure](/en/blog/how-to-do-useful-market-research-listen-before-you-measure). ### 3. Code what changes the decision Review the corpus through operational dimensions: trigger, desired outcome, barrier, alternative, choice criterion, consequence, and context. When an important response does not fit the current framework, add or revise a category instead of forcing it into the nearest theme. Thematic analysis is interpretive; themes do not mechanically fall out of text. Braun and Clarke describe a process of familiarization, coding, theme construction, and review ([Braun and Clarke, 2006](https://doi.org/10.1191/1478088706qp063oa)). Keep a route back from every theme to the excerpts that support or challenge it. ### 4. Split profiles only when the action changes Do not create one persona per industry, age group, or company size by default. Create two profiles when an observed difference requires a different promise, product, journey, or risk treatment. Two operations managers may share the goal of closing accounts without errors but need different journeys: one needs to remove duplicate entry, while the other needs to see who has not submitted a receipt. If the product or marketing response changes, the split is useful. Otherwise, keep one persona and record the variation. ### 5. Connect every claim to evidence For each important line, keep: - the synthesized claim; - source excerpts or observations; - the types of situations in which it appeared; - counterexamples; - the permitted scope of the claim; - the collection date. This evidence file does not need to appear on the public persona card. It does need to remain available to the team using it. If you have a large corpus, use a structured method for [analyzing open-ended responses without confusing memorable with frequent](/en/blog/analyzing-200-open-responses-without-bias-a-method-for-actionable-decisions). ### 6. Test, version, and retire Review the persona with people close to the field: customer-facing teams and, when possible, members of the audience studied. Do not ask only, “Does this sound like you?” Test decisions instead. Does the message answer the documented barrier? Does the journey work in the observed situation? Which responses contradict the profile? Then date it. Offers, markets, and behavior change. A persona that can never be revised becomes internal folklore rather than a learning tool. ## Example: from ideal customer to actionable profile Imagine an expense-management software company. Its first persona reads: “Sophie, 38, organized operations manager, likes saving time, and uses LinkedIn.” The card is easy to illustrate and almost useless for a concrete decision. After interviews and open-ended responses, the team rewrites it: | Field | Evidence-based persona | |---|---| | Decision | Choose the main promise for the trial page | | Situation | Month-end close with receipts spread across several teams | | Desired outcome | See what is missing before chasing people and before the accounting deadline | | Main barrier | The manager cannot tell whether a document is missing, pending, or already sent elsewhere | | Alternative | Tracking spreadsheet, individual messages, and a shared folder | | Choice criterion | Visibility into each request without additional data entry | | Variation | Some teams prioritize accounting integration; others prioritize employee reminders | | Illustrative quote | “The problem is not filing the receipt. It is knowing which one will never arrive.” | | Limit | Exploratory sample of already equipped small businesses; no estimate for the full market | This example is fictional. It demonstrates the shift: the persona no longer claims to describe “Sophie’s” personality. It connects a messaging decision to a situation, barrier, alternative, and explicit limit. The quote helps with wording; it does not prove segment size. ## Interview questions that actually feed the persona Use open-ended questions that leave room for unexpected categories: 1. Tell me about the last time you faced this problem. 2. What did you try before arriving at your current solution? 3. When did you consider changing — or decide not to change? 4. What would have made the decision easier? 5. Who else is involved, and what matters to them? 6. What concrete consequence did the problem have? 7. In what situation would you handle it differently? Avoid leading with your offer or a list of characteristics. First collect the words, causes, exceptions, and trade-offs your template did not predict. Closed questions become useful later, when you need to measure categories that have stabilized. ## Can AI create a buyer persona? AI can help organize a corpus, suggest structure, group related wording, and identify areas that need review. It can also produce a persuasive profile in seconds from almost no evidence — which is exactly the risk. Keep three operations separate: 1. collect or import real material; 2. generate themes and hypotheses connected to sources; 3. let people decide what belongs in the persona. Ask the tool to show missing evidence, contradictions, and the excerpts it used. Do not ask it to invent likely hobbies, personality traits, or motivations for a segment. Fluent prose is not proof. ## Where Harmate can help Harmate can help marketing teams frame a hypothesis, collect or import open-ended responses, compare audiences, and preserve themes with their source quotes in a decision brief. The buyer persona can then become a deliverable derived from reviewable evidence rather than an uncontrolled generated conclusion. The platform does not recruit participants, make a small sample representative, or decide which segment to target. The team remains responsible for recruitment, the scope of its claims, and the final decision. See the workflow on [Harmate for marketing teams](/en/profiles/marketing-teams). ## Frequently asked questions ### What is the difference between a target audience and a buyer persona? A target audience describes a group the organization wants to reach. A buyer persona synthesizes observed situations, goals, barriers, and criteria within a group to support a concrete decision. It is a working tool, not a replacement for market segmentation or sizing. ### How many buyer personas should you create? Use the smallest useful number. Add a persona only when an observed difference requires another action. Three profiles that lead to the same message or journey can probably be merged. ### Does a buyer persona need an age and photo? Only when those elements help the document without inventing causality. A photo can improve recall but may also narrow the team’s mental image. Age matters when it demonstrably changes the context, need, or decision. ### How many interviews do you need? There is no universal threshold. Recruit contrasting situations, track what each new interview adds, and disclose which audiences are missing. An exploratory corpus can build hypotheses; it does not automatically measure their frequency. ### When should a buyer persona be updated? Update it when a new audience, offer change, behavior shift, or important counterexample changes the decision it supports. Add a date, version, and review trigger from the start. ## Key takeaway The best buyer persona is not the one that feels most human. It is the one that helps a team make a decision without forgetting where its claims came from. Start with the decision. Collect open situations. Group mechanisms without erasing disagreement. Link each characteristic to evidence. Then measure, test, and revise. The name makes the persona memorable. The sources make it useful.