Personas: What They Are, How to Build One That Is Useful

Updated September 2026 · Written and maintained by the Progression Agency strategy team

A persona is a short, evidence-based description of one type of person you are trying to reach, written so a team can make decisions with it. Most persona documents fail that test badly: they contain a stock photograph, an invented name, a fictional biography and no decision anybody could actually take differently as a result. This page covers what belongs in a persona, where the evidence comes from, how character personas in fiction differ from buyer personas in marketing, and the cases where the honest answer is not to build one.

The CPO title

An abbreviation that means several different jobs.

CPO title meaning depends entirely on the industry. In technology it is usually Chief Product Officer, owning product strategy and the product organization. In procurement-heavy industries it is Chief Procurement Officer. In organizations with a large workforce it can be Chief People Officer, and in regulated sectors Chief Privacy Officer.

The reason it matters in marketing is targeting: a B2B campaign aimed at CPOs will reach four different buyer types unless the targeting is qualified by industry and by the responsibilities named in the profile. Buying a list of CPO titles without that qualification is one of the more common ways B2B budget is wasted.

Persona: meaning, definition, and the characteristics that make one usable

The persona definition used in marketing is narrower than the word’s general meaning. In ordinary usage a persona is the character or face someone presents to the world — the word comes from the Latin for a theatrical mask, and Jung later used it for the social self a person constructs. In marketing the meaning is a specific research artefact: a documented composite of a real customer segment, built from interviews and behavioral data, used to make decisions about who you are writing and building for. The def most teams work from in practice is shorter still — a named, written summary of one type of buyer. The persona characteristics that make it usable are what separate a working document from a poster: the job or situation that creates the need, the trigger that starts the search, the criteria used to compare options, the objections raised before purchase, and where the person actually looks for information. Characteristics that feel vivid but change no decision — a stock photograph, a first name, a favorite coffee order, an age to the year — are the ones to leave out. If a detail would not change a headline, a feature priority or a channel choice, it is decoration.

What is a persona?

A persona is a short description of one type of person you are trying to reach, compiled from evidence about real people and written so that a team can make decisions with it. The defining characteristic is not the format — it is that every attribute in it would change something somebody does.

That definition rules out most persona documents in circulation. A page containing a stock photograph, an invented name, an age, a job title and a paragraph of imagined biography is not a persona in any useful sense, because nobody writing an advertisement, a page or a sales email would do anything differently having read it.

What a persona is for

Making decisions consistent across people who cannot all talk to customers. A copywriter, a designer, a media buyer and a salesperson should reach compatible conclusions about what to say and where to say it, and the persona is the shared document that makes that possible.

What a persona is not

It is not a market segment, which is a group defined for measurement rather than for empathy. It is not an ideal customer profile, which describes the firms worth selling to rather than the people inside them. And it is not a substitute for talking to customers — it is a summary of having done so.

What are character personas?

In fiction, screenwriting, game design and role-play, a character persona is an invented profile — voice, background, motivation, mannerisms — used to keep a character consistent across scenes or sessions. It is authored rather than researched, and that is the fundamental difference from a marketing persona.

The confusion between the two is not merely semantic and it causes real damage. A marketing team that builds buyer personas the way a novelist builds character personas — by imagining a coherent person — produces something that reads convincingly and describes nobody. The method that is correct in one discipline is precisely the failure mode in the other.

When invention is the right method

When you are creating something rather than describing something. A character in a story, a voice for a brand, a persona for an assistant or a game NPC — all of these are authored, and consistency rather than accuracy is the standard they are judged against.

When evidence is the only acceptable method

When decisions about real money will be made from the document. If a persona will determine what you advertise, where you spend, or what a page says, it must be compiled from real people or it will be confidently wrong in a way nobody catches for a year.

What belongs in a buyer persona?

Five things: the trigger that makes somebody start looking, what they compare you against, the objection that would stop them, where they look for information, and who else has to approve the decision. Add the words they actually use for the problem, and you have a persona somebody can work from.

The scorecard applies a single test to each line: name the decision this attribute would change. The trigger changes what a headline says. The comparison set changes what a page must address. The objection changes what the FAQ covers. A favorite coffee changes nothing, and by that test most persona documents are two thirds deletable.

The quadrant chart shows why bad personas happen. Demographics are trivially easy to establish and change almost nothing; triggers and objections are harder to establish and change almost everything. Documents drift toward the easy attributes unless somebody enforces the test.

The four middle rows are frequently conflated in the same document, which is how a persona ends up containing a revenue figure, a job title and an invented hobby simultaneously. Keep them separate and each becomes useful.

Where does the evidence come from?

Customer interviews, lost-deal interviews, sales call notes, support tickets, survey responses and search query data. In roughly that order of signal quality, and in almost exactly the opposite order of how often they are used.

The most valuable row on that chart is lost-deal interviews. People who evaluated you and chose something else know more about your positioning problem than any customer does, and almost nobody asks them. Eight of those conversations will change a persona more than fifty survey responses.

How many interviews are enough?

Eight to twelve per persona is the usual point at which you stop hearing new things. If the tenth conversation is still producing surprises, you probably have more than one persona in the group and should split it.

What to ask

Six questions, all about what actually happened rather than what people would hypothetically do. Past behavior is evidence; predicted behavior is opinion, and personas built on the second are the ones that turn out to describe nobody.

Record the language exactly

Write down the words customers use for the problem, not the words your company uses for the product. That vocabulary drives your page copy and your search targeting simultaneously, and it is almost always different from internal language in ways that matter.

How do you build one, step by step?

Decide what the persona is for, gather what you already have, interview eight to twelve people including recent losses, cluster by behavior rather than demographics, write only decision-changing attributes, and then name five real customers who fit.

Cluster on behavior, not job title

Two people with the same job title can have completely different triggers, and two people with different titles can behave identically. Grouping by what made them start looking produces personas that predict something; grouping by title produces an org chart.

The five-real-people test

If you cannot name five actual customers who match a persona, you have written a hypothesis. That is not a failure — hypotheses are useful — but it must be labeled as one, or the team will treat a guess as a finding.

How many personas should you have?

Three or four is usually the ceiling. Beyond that, nobody remembers them, everybody defaults to the one they invented in their head, and the set stops being used. A single well-evidenced persona used daily beats seven documented ones nobody opens.

The last row is the practical test. If having two personas would not cause you to spend money differently, you have one persona and a note about variation within it.

The right-hand column is the diagnostic. Every one of those useless versions is technically an answer and none of them changes a decision. If a persona field could be filled in without having spoken to anybody, it has been.

When are personas a waste of time?

When you have very few customers and can simply talk to all of them, when the buying decision is genuinely uniform across your market, when nobody will read the document, or when the exercise is substituting for the customer conversations it was supposed to summarize.

The failure ordering makes the underlying problem clear: personas fail because of how they are made rather than because the idea is unsound. Built from assumption, stuffed with inert attributes, never revisited — three fixable errors account for most of the skepticism the technique attracts.

The workshop trap

A room of internal people confidently describing customers produces a document that reflects the room. It feels productive, it is fast, and it encodes the organization’s existing assumptions as though they were findings. If a persona was built without talking to a single customer, that should be stated on the document.

The wall-poster trap

A persona printed, framed and hung is finished, and a persona should never be finished. Keep it in a document somebody can edit, with a date on it and a review in the calendar.

Keeping personas current

Review at twelve months against the last year of actual customers, and rebuild rather than edit whenever the product or the market shifts materially. The most common state for a persona is three years old and quietly wrong.

The timeline is deliberately unglamorous. Nothing about persona maintenance is difficult; it requires only that somebody owns the document and that a review exists in a calendar. Both cost nothing and both are usually absent.

How personas connect to the rest of the work

A persona is only worth building if something downstream changes because of it. In practice that means page copy, search targeting, ad creative, sales enablement and product priorities — and if none of those change, the exercise was theater.

Use that table as an audit. Take an existing persona document, work down the right-hand column, and check whether any of those artefacts actually differ because of it. If none do, the persona is not in use regardless of how good it looks.

Our marketing agency overview covers where this sits in a wider plan, and branding services covers the positioning work personas feed into.

Turning a persona into search and copy decisions

Answer first: take the language field and nothing else. The phrases customers used for their problem are simultaneously your page headings, your FAQ questions and your keyword targets, and they are almost never the phrases your company uses internally.

Work this table in both directions. Building forward tells you what to publish; auditing backward from an existing site tells you whether any persona is genuinely in use, because every artefact in the right-hand column should be traceable to a field in the left.

Research, positioning and measurement talks from the platform publishers

Publicly available sessions on customer research, positioning and measurement. None of these are ours; each is credited to its channel and upload date, every identifier was checked live before publication, and each tile loads its player only when clicked.

Frequently asked questions