Most buyer personas are fiction with a stock photo attached
You know the document. “Marketing Mary, 34, drinks oat milk lattes, loves yoga, reads industry blogs.” It was built in a two-hour workshop, nobody in the room had spoken to a customer that quarter, and the file has not been opened since.
A buyer persona is only worth building if it changes a decision: which channel gets budget, which audience you target, which angle a writer takes in a brief. This guide shows how to create a buyer persona from data you already own: recorded sales calls, GA4 reports, CRM fields, support tickets and a short customer survey. Then it shows how to plug that persona into the three places it actually earns its keep.
What is a buyer persona (and what it is not)
A buyer persona is a research-based profile of a specific type of person who buys from you, covering their trigger to start looking, their decision criteria, their objections, the channels they trust and the language they use to describe the problem.
It is not:
- An ideal customer profile (ICP). An ICP describes the account or company that is a good fit (industry, size, revenue, tech stack). A persona describes the human inside it.
- A demographic sheet. Age and gender rarely explain why someone bought.
- A customer journey map. The journey is the sequence of steps. The persona is the character walking through them.
Rule of thumb: if a line in your persona cannot change a targeting setting, a channel budget or a piece of copy, delete it.
The data sources to use before you write a single word
Every field in your persona should trace back to a source. Here is where to pull from, ranked by signal quality. Create and edit personas is a useful companion to this.
| Source | What it tells you | Effort |
|---|---|---|
| Recorded sales calls and demos | Trigger events, objections, competitors named, exact wording | Low (the data already exists) |
| Customer interviews (6 to 10) | Why they switched, what almost stopped them, who else was involved | Medium |
| Customer survey (8 to 10 questions) | Scale: how common is each trigger, criterion and channel | Medium |
| GA4 (user attributes, tech, acquisition, paths) | Directional demographics, device, geography, entry points, converting paths | Low |
| CRM closed-won and closed-lost fields | Job titles, company size, deal length, loss reasons | Low |
| Support tickets and live chat logs | Friction points, misunderstood features, onboarding fears | Low |
| Review sites, Reddit threads, community forums | Unfiltered comparisons and the vocabulary buyers use with peers | Low |
| Search Console and keyword tools | The literal phrases people type before they know your brand | Low |
How to create a buyer persona in 6 steps
Step 1: Decide which decision the persona has to inform
Start at the end. Write one sentence at the top of the doc: “This persona exists so we can decide X.” Common versions:
- “So we can choose between LinkedIn Ads and paid search for our 2027 budget.”
- “So we can write 12 content briefs that target the objections costing us deals.”
- “So we can fix the 68% drop-off on the pricing page.”
This single line kills the urge to add favourite podcasts and coffee habits. It also tells you how many personas you need. Two or three is normal for most businesses. If your product is bought by one role in one type of company, one persona is enough. adobe.com walks through the specifics.
Step 2: Pull the quantitative baseline (30 to 60 minutes)
Before talking to anyone, get the numbers on screen so your interviews confirm or contradict something.
In GA4:
- Reports > User > User attributes > Demographic details for age, gender, interests, country and language. Add your key conversion as the metric so you see who converts, not just who visits.
- Reports > User > Tech > Tech details for device, browser and screen resolution. If 70% of converting sessions are desktop but your traffic is 65% mobile, that is a persona insight and a UX bug at once.
- Reports > Acquisition > Traffic acquisition and the Landing page report, segmented by converters, to see which entry points attract buyers rather than browsers.
- Explore > Path exploration starting from your conversion event, going backwards, to see the real pre-purchase sequence.
- Build a comparison of “users who triggered the purchase or lead event” versus all users and read every report through that lens.
GA4 caveat you should write into the doc: demographic and interest data comes from Google signals and only covers a subset of signed-in, consenting users. Reports also apply data thresholding, so small segments may be hidden. Treat GA4 demographics as directional, never as proof. The CRM is your ground truth for job titles and company size.
In your CRM: export the last 50 to 100 closed-won deals and the last 50 closed-lost. Count job titles, company sizes, industries, source, average days to close and the stated loss reason. The gap between won and lost is often the most useful persona insight you will get all year.
Step 3: Mine sales calls and support tickets for language
If your team records calls with a tool like Gong, Fathom or Fireflies, you are sitting on a qualitative research archive. Pick 10 recent won deals and 10 lost ones and tag every transcript against five buckets:
- Trigger: what happened in their business right before they contacted you
- Current workaround: spreadsheet, agency, competitor, doing nothing
- Evaluation criteria: what they compared and in what order
- Objection: what they pushed back on, and at which stage
- Verbatim phrases: the exact words used for the problem
Keep a separate tab of verbatim quotes. These become ad headlines, email subject lines and H2s. Customers rarely use your product category name. They say “we keep losing track of who promised what” rather than “we need a workflow orchestration layer.”
Do the same tagging pass on 50 support tickets and the last 30 chat transcripts from your sales pages. Pre-sale chat questions are objections in disguise.
Step 4: Interview and survey real customers
Six to ten 25-minute interviews will surface almost every recurring pattern. Recruit a mix: recent buyers, long-term customers, and two or three people who evaluated you and chose someone else. Offer a small incentive and be clear it is research, not a sales call.
Then validate at scale with a short survey (8 to 10 questions maximum) sent to customers who bought 30 to 90 days ago, while memory is fresh.
The 32 questions to ask
Trigger and context (the most underused questions)
- What was happening in your work the week you decided to look for a solution?
- What were you doing before? Why did that stop working?
- Was there a deadline or an event forcing the timing?
- What would have happened if you had done nothing?
Search and discovery
- What did you literally type into Google first?
- Where else did you look: a peer, a community, a marketplace, an AI assistant?
- Which sites or people did you already trust on this topic?
- How did you first hear of us? (open text, not a dropdown)
Evaluation
- Which other options made your shortlist?
- What made you eliminate the ones you dropped?
- What were the top three things you compared?
- What information was hardest to find?
- What nearly stopped you from buying?
- What convinced you in the end?
- How long did the whole process take?
The buying committee
- Who else was involved in the decision?
- Who had to approve the spend?
- Who would have been most affected if it went badly?
- What did you have to show or say internally to get a yes?
- Did anyone object? On what grounds?
Job to be done and outcome
- How do you describe what we do to a colleague?
- What does success look like 90 days from now?
- How is your performance measured in this area?
- What part of the work do you most want to stop doing?
Objections and risk
- What was your biggest worry before signing?
- What did you assume about price before you saw ours?
- What did you need to see to trust us?
- Has anything disappointed you since?
Channels and media habits
- Which newsletters or publications do you actually read?
- Which communities, groups or events do you take part in?
- Whose opinion in this industry do you trust?
- How do you prefer vendors to contact you, and how often?
Interview technique that matters: ask about behaviour in the past, not preferences in the future. “What did you do last time?” produces facts. “Would you click an ad like this?” produces polite guesses.
Step 5: Segment, then write the persona sheet
Group your findings by trigger and job to be done, not by demographics. If two people have the same trigger and the same decision criteria, they are one persona even if one is 28 and the other is 52. If two people share an age bracket but one is replacing a competitor and the other is buying a category for the first time, they are two personas with two very different content needs.
Use this template. Every field has a marketing use.
| Field | Source | What it drives |
|---|---|---|
| Label and role | CRM titles | LinkedIn job title targeting |
| Firmographics or household profile | CRM, GA4 | Audience filters, exclusion lists |
| Trigger event | Interviews, calls | Ad hooks, campaign timing |
| Job to be done | Interviews | Positioning, landing page headline |
| Top 3 decision criteria | Survey | Comparison pages, feature order |
| Top 3 objections | Sales calls, lost deals | FAQ blocks, retargeting creative |
| Buying committee | Interviews | Enablement assets, ABM lists |
| Search phrases used | Search Console, interviews | Keyword map, ad copy |
| Trusted sources and channels | Survey | Channel mix, sponsorships, PR |
| Disqualifiers | Closed-lost analysis | Negative keywords, audience exclusions |
| Three verbatim quotes | Transcripts | Headlines, subject lines, hooks |
A filled-in example
Persona: “Stretched Ops Lead”
- Role: Operations Manager or Head of Operations, companies of 40 to 200 people (61% of closed-won deals in the last 12 months)
- Trigger: Headcount grew past roughly 50 and the shared spreadsheet broke, or a visible error reached a client
- Job to be done: Stop being the single point of failure for who is doing what
- Decision criteria: 1) Setup under two weeks, 2) team will actually adopt it, 3) transparent pricing
- Objections: “My team will not use another tool”, “We tried this before and abandoned it”, “I cannot get budget approved this quarter”
- Committee: Finance approves above a threshold, two team leads veto in practice
- Search phrases: “track team tasks without spreadsheets”, “[competitor] alternative”, “simple ops software small team”
- Trusted sources: two industry newsletters, one Slack community, peer recommendation above all
- Disqualifier: Solo operators and agencies under 10 people churn at 4x the average rate
- Quote: “I did not need more features. I needed to stop answering the same question eleven times a day.”
Step 6: Convert the persona into channel, targeting and content decisions
This is the step nearly every persona guide skips, and the reason most personas die in a shared drive.
Channel selection
| Persona signal | Channel decision |
|---|---|
| Trigger is sudden and painful, buyers search immediately | Weight paid search and bottom-funnel SEO (comparison, alternative, pricing pages) |
| Nobody searches because they do not know the category exists | Weight paid social, newsletter sponsorships, communities, video |
| Peer recommendation dominates the survey | Fund referral programme, review site presence, community participation over display |
| Long committee-based cycles | Email nurture plus shareable internal business-case assets |
| Converters are overwhelmingly desktop in GA4 | Cap mobile bids or use mobile for capture, desktop for conversion |
Ad targeting
- Job titles and seniority from CRM analysis go straight into LinkedIn audience filters, not the titles you wish you sold to.
- Disqualifiers become exclusions: company size bands, industries and negative keywords. This usually saves more money than any bid tweak.
- Objections become the retargeting sequence: one objection per creative.
- Verbatim quotes become primary text and headline tests, straight from the transcript tab.
- Trigger events become timing signals (hiring surges, funding, seasonality, regulation dates).
Content briefs
Add four persona-derived lines to the top of every brief:
- Persona and stage: Stretched Ops Lead, evaluation stage
- The question this page answers: in the customer’s own words
- The objection it must defuse: “my team will not adopt another tool”
- The vocabulary rule: phrases to use, jargon to avoid
A writer who receives this produces a usable draft. A writer who receives “Marketing Mary, 34, likes yoga” produces filler.
How many buyer personas should you build?
Between one and four for most companies. Each persona should represent a meaningful slice of revenue and require a genuinely different message. If two personas would get the same landing page and the same ads, merge them. Also build one negative persona: the profile that buys, complains, churns and drains support. Knowing who to exclude improves paid efficiency faster than anything else on this list.
Common mistakes to avoid
- Building from a workshop instead of data. Whiteboard sessions capture internal assumptions, not customer reality.
- Demographics as the backbone. Age and gender are filters, not explanations.
- Interviewing only happy customers. Lost deals and churned accounts hold the objections costing you revenue.
- Letting AI invent the persona. Generators are useful for structuring and summarising your transcripts and survey exports. Asking a model to imagine a customer just returns the internet’s average, which is what your competitors also got.
- Never refreshing it. Review quarterly against closed-won data, re-interview annually. Markets, pricing and competitors move.
- Ten pages nobody reads. One page per persona. If it does not fit, it is not a persona, it is a research report.
A realistic timeline
| Week | Work |
|---|---|
| 1 | Define the decision, pull GA4 and CRM data, recruit interviewees |
| 2 | Tag 20 sales calls and 50 support tickets, launch the survey |
| 3 | Run interviews, close the survey, cluster findings by trigger |
| 4 | Write one-page personas, rewrite ad targeting and the next 10 content briefs |
FAQ
What is a buyer persona example?
A useful example reads like this: “Stretched Ops Lead, Operations Manager at a 40 to 200 person company, starts looking when headcount growth breaks the shared spreadsheet, needs setup done in under two weeks, main objection is team adoption, trusts peer recommendations and two industry newsletters, searches ‘track team tasks without spreadsheets’.” Notice that none of the decisive information is demographic.
What are the four types of buyer personas?
One common model groups buyers by decision style: competitive (wants speed, proof and superiority), spontaneous (decides fast, needs low friction), methodical (needs detail, specs and documentation) and humanistic (needs trust, relationships and social proof). Some teams instead use lifecycle types: the new buyer, the switcher, the renewer and the influencer. Either model helps you vary format and tone, but neither replaces research into your own customers.
Can I create a buyer persona with no customers yet?
Yes, with weaker evidence. Interview 10 people in your target role even if they are not customers, read review sites for competing products, study community threads and analyse search demand for the problem. Label it a hypothesis and rewrite it after your first 20 sales.
How is a buyer persona different from an ICP?
The ICP defines which companies or households are worth pursuing. The persona defines the individual who evaluates, objects and signs. In B2B you need both: the ICP sets the account list, the persona sets the message.
How often should buyer personas be updated?
Do a light quarterly check against closed-won and closed-lost data, and a full refresh once a year or whenever you change pricing, launch a major product line or enter a new market.
Do AI persona generators work?
They work well as an assistant and badly as a researcher. Feed a model your interview transcripts, survey exports and call notes and ask it to cluster the patterns. Do not ask it to invent a customer you have never spoken to.
The test that matters
Once the persona is written, run it past one question: what did we change this week because of it? If the answer is a new set of negative keywords, a rewritten landing page headline, one channel defunded and three content briefs rebuilt around real objections, the persona is working. If the answer is “we added it to the onboarding deck”, start again at step one.