Somewhere in your company there is almost certainly a slide with a stock-photo headshot, a made-up name like "Marketing Mary," an age, a job title, and a tidy list of "goals and frustrations." It was made with care, presented once, and has been quietly ignored ever since. The problem isn't that personas are useless. It's that most personas describe who the customer is and skip the only thing that actually predicts behaviour: the mindset they bring to the decision.

The quick version

  • A persona is a fictional, evidence-based character that stands in for a real group of users, so a team can design for a person instead of a spreadsheet.
  • A persona is only as good as the research under it. A persona built from your team's assumptions is a guess wearing a face.
  • A mindset describes the inner thinking, emotions and principles someone brings to a goal, and it predicts choices far better than age, income or job title.
  • The move: keep the persona as a memory aid, but segment your real decisions by mindset, what people are trying to achieve and how they think about it, not by demographic.

The idea in depth: a persona is a character, built from evidence

The persona, as a design tool, has a clear origin. The interaction designer Alan Cooper developed the technique through his consulting practice in the early 1990s and introduced it publicly in his 1998 book The Inmates Are Running the Asylum. His insight was almost embarrassingly simple: software teams design for an "elastic user", a vague, shape-shifting everyone who conveniently wants whatever the team already planned to build. Replace that elastic user with a single, specific, named pretend-person, and the team is forced to make real trade-offs. You can't please "all users." You can decide what one named character needs.

What follows from that is a discipline, not a deliverable. Before a roadmap argument, name the specific person you're designing for and ask whether the feature serves them, not the abstract market. A persona's whole job is to end the "well, some users might want…" conversation that justifies everything and decides nothing.

But here's the part teams skip. The persona has to be built from research, or it's theatre. The Nielsen Norman Group sorts personas into three honest types in Page Laubheimer's 2020 guide, "3 Persona Types: Lightweight, Qualitative, and Statistical." Proto-personas are quick sketches of what the team already assumes, with no new research. Qualitative personas come from a handful of real interviews or field studies (roughly 5–30 users). Statistical personas add survey data from hundreds of people and cluster them mathematically. NN/g's verdict is unambiguous: for most teams the qualitative approach is the right balance of effort and value, and a persona "must always be rooted in a qualitative understanding of users."

flowchart TB
  R(["Real research:
interviews, field studies, behaviour"]) --> P("Persona
one named character")
  A("Team assumptions
no new research") -.->|"proto-persona
= a guess"| P
  P --> D(["A sharper design decision:
does this serve THIS person?"])
A persona is a research artefact, not a creative-writing one. Skip the research and you get a guess with a stock photo. Leaders Loop

Where personas break, and why mindsets fix it

Here is the honest limitation, the part the slide-deck never admits. The classic persona leans on demographics, age, gender, income, job title, and demographics are weak predictors of choice. Two 42-year-old fathers with the same salary can want opposite things from the same product. The marketing-research firm Gradient Metrics, summarising the practitioner consensus, puts it plainly: demographic segmentation "assumes that individuals within the same demographic group have similar preferences and behaviors, which is often not the case," whereas mindset segmentation uncovers the motivations and attitudes that actually drive behaviour. (Treat that as a strong practitioner claim, not a single peer-reviewed law, it's the working assumption of the field, not settled science.)

This is where the researcher Indi Young reframes the whole exercise. Young, who wrote Mental Models (Rosenfeld Media, 2008), argues for what she calls thinking styles: "deeply-researched, demographics-free mindsets" that describe the inner thinking, emotional reactions and guiding principles people bring to a particular purpose. The crucial twist is that a mindset is not a person. It's tied to a purpose, and the same human switches between them as the purpose changes, a "just get it done, I don't care how" mindset booking a quick flight, then a "research everything, I refuse to be ripped off" mindset buying a car. Demographics can't capture that. The situation and the goal can.

A persona answers "who is my customer?" A mindset answers "what is going through their head right now?" Only one of those reliably predicts the next click.

The practical correction is to stop letting the demographic do the work the mindset should do. When you write or refresh a persona, demote age and income to background colour and promote the part that matters: what is this person trying to achieve, what do they fear getting wrong, and what principle won't they bend on? Those three answers travel across age, gender and postcode, and they tell a designer or a salesperson what to actually do. (For the formal toolkit on slicing a market this way, see segmentation (demographic, behavioural, needs-based).)

How the two fit together: face plus engine

Personas and mindsets aren't rivals; they're a body and an engine. The persona gives a team a shared, memorable face, so a roomful of people argue about the same customer instead of five imaginary ones. The mindset is the engine inside, the motivation that explains the choice and predicts the next one. Build a persona around a mindset and you get both: empathy a team can hold in their heads, grounded in behaviour they can verify.

The failure mode to name: the proto-persona that never gets tested. Jeff Gothelf, who popularised the lean approach, is blunt about it, start with a rough proto-persona if you must, but then go talk to customers, because, in his phrase, "a persona without data is just a guess." A guess is fine as a starting hypothesis; it's dangerous as a permanent wall fixture, because it launders the team's opinions into something that looks like fact. (How to gather that grounding data sits in voice-of-customer programs and customer needs identification & latent needs.)

flowchart LR
  M(["Mindset / thinking style
goal · fear · principle"]) --> P("Persona
a memorable face for the mindset")
  P --> T1("Product:
design for the goal")
  P --> T2("Marketing:
speak to the fear & principle")
  P --> T3("Sales:
match the buying mindset")
Mindset is the engine; the persona is the face that makes it shareable. Decisions hang off the mindset, not the headshot. Leaders Loop

A worked example

Imagine a mid-sized online accounting-software firm, call it Ledgerline, selling to small businesses. (Figures here are illustrative.) Their persona deck has two characters: "Startup Sam, 29," and "Established Eric, 51." Marketing has been targeting Sam with speed-and-modern messaging and Eric with reliability-and-support messaging, split by age. Conversion has been flat for a year.

Before another redesign, the product lead runs twelve interviews and listens for mindset instead of demographic. Two thinking styles emerge that cut clean across the age line. One group, call it the delegators, wants the software to "just handle it so I never think about tax again"; their fear is a nasty surprise from the tax office, and their unbendable principle is I will not become an accountant. The other group, the controllers, wants to see every number and reconcile it themselves; their fear is losing the thread of their own finances, and their principle is I trust what I can check. Crucially, plenty of 29-year-olds are controllers and plenty of 51-year-olds are delegators. Age predicted nothing; the mindset predicted everything about which features they used and which messages landed.

So Ledgerline rebuilds the two personas around the two mindsets, not the two ages. The delegator gets messaging about automation and "we'll flag anything that needs you"; the controller gets the detailed ledger view and an audit trail front-and-centre. Onboarding opens with one question, "Do you want to handle the numbers yourself, or have us handle them for you?", and routes accordingly. Illustratively, trial-to-paid conversion climbs from roughly 14% to 22% in two quarters. Nothing about the product's quality changed; the team simply stopped sorting people by birthday and started sorting them by what was in their heads. The line they now repeat: we weren't selling to Sam and Eric; we were selling to delegators and controllers who happened to be any age.

Frequently asked questions

Are personas dead? I keep reading that they are.

No, but lazy personas deserve to be. The criticism is almost always aimed at demographic, assumption-based personas that never touched a real customer. A persona built from research and organised around a mindset is still one of the cheapest ways to give a team a shared focus. The fix isn't to abandon personas; it's to put real thinking inside them.

What's the actual difference between a persona and a mindset?

A persona is a character, a single fictional person who represents a group, with a name and a face so the team can empathise. A mindset (or "thinking style") is the way of thinking someone brings to a goal: their motivations, fears and principles in that situation. One person can hold different mindsets in different situations, which is exactly why a mindset predicts choices better than a fixed demographic profile.

How much research do I need before a persona is "real"?

Less than you fear. Nielsen Norman Group's guidance is that a handful of qualitative interviews, roughly 5 to 30 users, is enough for most teams to build defensible qualitative personas; you only need large-scale statistical clustering if you're a very big organisation with the appetite for it. Start with a rough proto-persona by all means, but treat it as a hypothesis to test, not a finding.

Won't sorting customers by mindset just create endless segments?

It tends to do the opposite. Researchers usually find a small number, often two to seven, of distinct mindsets within a single purpose, because there are only so many ways to think about a given goal. That's far more manageable than the spray of demographic micro-segments most teams accumulate, and each one points to a clear design or message.

Where do mindsets come from, do I just make them up?

No, and this is the whole point. You find mindsets the way you find any honest insight: by interviewing real people about a real, recent decision and listening for the thinking underneath the choice, not the features they say they want. Invent them in a workshop and never test them, and you've just built a fancier proto-persona.

Related in the Toolkit

Where to go next