What's the difference between a fictional buyer persona and an evidence-based one?

A fictional persona is assembled from a room's opinions and assumptions, with no claim in it that could be checked against real data. An evidence-based persona is built from what buyers actually said, asked, and searched for, and every line traces back to a specific source.

What sources should a buyer persona actually be built from?

Sales call notes and transcripts, support tickets, on-site search and content behavior, win-loss interviews, and renewal or expansion conversations — records generated by buyers acting like buyers, not by internal stakeholders describing buyers from memory.

How do you know if a line in a persona document is fiction?

Ask whether next quarter's data could contradict it. If a statement is so generic that no evidence could ever disconfirm it ("cares about ROI"), it's fiction. If it's specific and sourced enough that new data could prove it wrong, it's a real claim.

How often should a buyer persona be updated?

On a schedule tied to when new evidence arrives — new call notes, tickets, and loss reasons — not once a year in a workshop. A persona that never changes usually means nobody's checking it against anything.

Can a single persona represent a whole buying committee?

Not accurately. Most B2B purchases involve more than one person with different concerns, so a single persona is a compression at best. Treating it as one decision-maker instead of one voice in a committee reintroduces the fiction problem evidence was meant to fix.

Buyer personas: evidence or fiction

Guide · frameworks · 5 min read · last verified 2026-07-21

Reviewed before publication Editorial board Independent commercial review

In shortA persona built in a workshop from memory is fiction with a stock photo. One built from what buyers actually asked and searched for is evidence — and the test is whether a line in it could turn out wrong.

There are two things called a "buyer persona," and they produce almost opposite value. One is assembled in a conference room from opinion and analogy — "I think our buyer is like this" — dressed up with a name, an age range, a stock photo, and a bulleted list of "goals and frustrations" nobody sourced from an actual conversation. The other is assembled from a record of what real buyers actually asked, searched, read, and said no to. The first is fiction with a headshot. The second is evidence. They look identical in a slide template, which is exactly the problem.

How the fictional persona gets made

The workshop version has a predictable shape: a room of internal stakeholders — product, marketing, sales — spends two hours generating attributes from memory and instinct. Someone proposes a job title. Someone else adds "data-driven, time-poor, skeptical of vendors" because that description fits almost any B2B buyer and therefore commits to nothing. A stock photo gets attached because a face makes the persona feel real. The document ships, gets referenced in a positioning deck twice, and then nobody checks it against a single actual deal.

The tell isn't that the room got it wrong — memory and instinct aren't worthless, and people close to the market often have real intuition. The tell is that nothing in the resulting document can be checked against anything. There's no claim in it that a specific piece of evidence could contradict. A persona that can't be wrong isn't a model of your buyer; it's a description of what your team already believed walking in.

What evidence-based persona construction actually looks like

The evidence version starts from records that already exist and were generated by buyers acting like buyers, not by your team talking about buyers:

Sales call notes and transcripts — the actual questions buyers ask, in their own words, not the questions you assumed they'd ask.
Support and onboarding tickets — what new and existing customers get stuck on, which tells you what they don't already understand or expect.
Search queries and on-site content behavior — what people look for before they ever talk to sales, if you have that data instrument in place.
Win-loss interviews — not "why did you buy," which invites a polite retrospective story, but "what were you doing the week before you started looking," which surfaces the actual trigger.
Renewal and expansion conversations — what a buyer says once they've lived with the product, which is a different and often more honest signal than what they said while evaluating it.

If you already use a market-intelligence tool to track what buyers ask and read in public and semi-public channels, that's another legitimate input — as long as every line stays traceable back to actual buyer language, not a summary someone wrote about the buyer. None of this is a survey or a study — it's your own operational record. The discipline is refusing to add anything to the persona that isn't traceable back to one of those sources. If a "frustration" in the persona doc can't be pointed at a support ticket, a call note, or a lost-deal reason, it doesn't go in.

Worked example: turning raw signal into a persona claim

Say you pull the last quarter of sales call notes for a product line and notice a pattern: in 14 of 20 first calls, the buyer's first substantive question is some version of "how does this handle our existing approval workflow," asked before questions about price or features. Separately, 6 of the 9 deals lost to "no decision" in the same quarter have a loss note mentioning an internal approval or procurement blocker.

That's a real, checkable claim: this buyer's binding constraint in the first conversation is integration with an existing approval process, not price or feature comparison, and it's a meaningful factor in stalled deals. You can build a persona section around it — "opens with process-fit questions before commercial questions" — and you can also disprove it next quarter if the pattern doesn't hold in the next batch of call notes. That falsifiability is the entire difference from the workshop version. The count of 14 of 20 isn't a statistic borrowed from anywhere; it's a tally of your own call notes, and its only job is to tell you whether the pattern is strong enough to act on or just a few loud calls you happened to remember.

The test: can this persona be wrong?

Before adding any line to a persona document, ask whether next quarter's data could contradict it. "Cares about ROI" fails the test — nearly every B2B buyer can be described that way, and no evidence could ever disconfirm it. "Asks about approval-workflow integration before price in the first call, in roughly two-thirds of first calls" passes — it's specific, sourced, and could easily turn out to be false next quarter. A persona document made entirely of statements that pass this test is short, useful, and gets revised. One made of statements that fail it is long, comfortable, and never changes, which is usually the sign it was never really about the buyer.

Where personas still fall short

Even an evidence-built persona describes patterns in who talks to you and how — it can miss buyers who never made it into a call, ticket, or search log because they self-selected out before you ever engaged them, which is its own kind of signal worth chasing separately. And a persona, however well sourced, is still a compression of a buying committee that usually includes more than one person with more than one set of concerns — treating "the persona" as a single decision-maker is its own way of turning evidence back into fiction.

The discipline that keeps a persona honest is the same discipline that keeps it current: check it against new evidence on a schedule — new call notes, new tickets, new loss reasons — and be willing to delete a line the moment it stops holding up. That's slower than writing a persona once in a workshop. It's also the only version that tells you anything true.