Why most personas die in the doc that created them

Pivoting is the default state of an early-stage startup. After a decade of working in them I'd argue that most of what gets called "product strategy" is really just a long string of small course corrections. A feature evolves, a price changes, an edge case turns out to be the main use case. Each of those is healthy. Each of them also quietly invalidates the personas the team built last quarter, and nobody updates them, because nobody has time, so the team carries on referencing customers who don't quite exist anymore. Then one day someone in a roadmap meeting cites "Marketing Mary" and you realise the gap between her and your actual users has been widening for months.

It's not a story I have found, it is just what happens by default.

Janna Bastow at ProdPad has been saying this for years. Her line is the one I keep coming back to: "Time and time again, user personas are created to only be stored in some GDrive folder and never used. At best they're presented to the wider team, turned into beautifully printed artifacts and stuck on a wall, never to be touched again. The 'never to be touched again' is the key issue here."[¹]

And if you want a number on it, Forrester's 2023 marketing survey is bleak. 63% of product and portfolio marketing leaders rate their persona-building skills as above average or best in class. Only a quarter of them actually use those personas in messaging. Just 28% use them in content.[²] So most of the people building personas think they're doing it well, and almost none of them are doing anything with the result. That's the whole picture, really.

The question I'd want to ask, then, is whether the problem is personas as a concept, or the artefacts we keep building to hold them.

The three ways they die

There's nothing mysterious about how this happens. It's usually one of three things, sometimes all three at once.

The first is that personas tend to get built once. Six-week project, research phase, synthesis phase, presentation, artefact, done. The project ends and the artefact starts ageing immediately. Anna Bulanova flagged this back in 2016 in a piece for Mind the Product and it has only got truer since: "Distributing profiles via email might seem the most efficient... solution, but that is not enough to keep personas alive. Developers get dozens of emails a day and information that is not continually on display is often forgotten."[³] Six months later, the persona is describing a customer base that has already moved on. The language in it isn't the language the team hears in calls anymore. Nothing about it has been refreshed. Of course people stop using it.

The second is that personas often get built around the wrong things. I want to be careful here because I don't think demographic personas are useless, they're often exactly the right tool for a specific job. Favourite brands genuinely matter when you're working on brand design. Favourite apps matter for mobile UX. The problem isn't that those dimensions exist, it's that teams pick one dimension to flatten the customer into and then carry that flattened version around as if it were the whole person. The persona built for the brand workshop turns up in the product roadmap meeting six months later, where her favourite coffee shop is helping precisely no one make a decision.

The third is the one that hurts most, which is that personas usually get cut loose from the evidence that made them. The doc says "Sarah cares about efficiency above all else." Fine. But which Sarah? Said when? Said in response to what question? Once you can't trace a claim back to its source the persona just becomes folklore. You can argue with it but you can't check it, and once you can argue with it, you'll ignore it the moment it gets in the way of whatever you wanted to do anyway.

Personas as a state, not a document

The fix isn't to throw personas out. Teams that have gone that route mostly end up worse off, because they've traded a flawed shared language for no shared language at all. The fix is to stop thinking of personas as documents.

Teresa Torres has been making this argument for a long time and she's pretty much built a career on it, having taught discovery to about 17,000 PMs in 100 countries at this point.[⁴] Her core idea is that discovery should be continuous rather than phased: "weekly touch points with customers by the team building the product, where they conduct small research activities in pursuit of a desired outcome."[⁵] Take her seriously for a moment and the implication is uncomfortable, because if customer understanding has to refresh weekly to stay accurate, then a persona that updates quarterly (at best, and usually not even that) isn't slightly wrong. It's the wrong format, full stop.

So what does the right format look like?

Personas that survive seem to do three things differently from personas that die. The first is that they stay connected to live evidence. A persona that just claims "Sarah hates new tools" is a guess. A persona that claims it and points to the four discovery calls, eleven support tickets and survey free-text response where that resistance actually showed up, recently enough to still be true, is something else. The second one settles arguments because there's nothing to argue with. The first one starts them.

The second thing they do is use the customer's language rather than yours. Bulanova's team at Ocado made this change explicit and I think it's one of the most underrated insights in the original article: "We do not say 'As a user I want… so that…' any more. Neither Robert or Luke or Albert would ever say anything this way. In my opinion, artificial phrases make the story feel less real, hence less important. So, we started quoting the exact words the users would use."[⁶] Read a persona built from real quotes alongside a persona built from your team's paraphrased summary and the difference is obvious. One feels like a real person you might disagree with. The other feels like a marketing document, because that's what it is.

The third thing, and honestly the one I find most useful in my own work, is that good personas surface the gap between what people say and what they do.

A few years ago I worked on a healthtech app. The kind of thing where sensors collect patient data and route it through to their assigned physicians, basically remote monitoring that the NHS keeps trying to scale. We did the research properly, spoke to physicians, spoke to patients, listened to both groups. When we asked patients what they wanted from the app the answer was almost universal. Less interaction. More automatic data transfer in the background. Quieter. Less intrusive. Get out of the way.

So that's what we optimised for. Months of design and engineering all pointing in the same direction, which was reduce the surface area of the app.

It was the wrong target.

What patients actually needed was the opposite of what they'd told us. The device data was already flowing fine, the sensors were doing their job. What they couldn't do, no matter how much they wanted to, was tell their physician the things the sensors weren't picking up. The vague worry. The "this reading looks fine but I want to flag it anyway" instinct. The stuff you'd say to a doctor face-to-face that just doesn't show up in a data feed. So they were working around the app constantly, finding other channels to communicate the things it wouldn't let them say.

"Less interaction" was a stated preference. The real need was somewhere completely different, sitting in the gap between what they were saying and what they were actually doing. A persona built on the stated preference would have led us further in the wrong direction. The useful thing wasn't either of the two answers, it was the distance between them.

Jodie Shaw has a related point that I think gets at the same thing from another angle: "One of the biggest myths in B2B is that people in the same role think the same way. They don't. Some CMOs want awards. Some want a pipeline. Some just want to survive the next board meeting."[⁷] A persona that smooths those differences into a single archetype is comforting, and it's also basically useless. A persona that names the variation gives the team something to actually work with.

What "alive" looks like in practice

If a persona is a living state rather than a document, what does that mean concretely?

It means the persona updates as new evidence comes in. New calls, new tickets, new survey responses, all without someone scheduling a quarterly refresh project that no one will get around to. It means anyone on the team can ask the persona a question and get an answer that's actually grounded in real conversations rather than someone's recollection of real conversations. It means confidence is visible per attribute, so when an answer is weakly evidenced or based on stale data, you know. It means new patterns get surfaced in real time, language shifts, emerging objections, new buying motions, things you'd otherwise notice six months too late.

That's the bar I've come to think personas need to clear to be worth the time they take. It's hard to clear that bar manually, which is, I think, why most teams don't.

One question

Here's the test I'd use, ignoring every framework that exists.

When was the last time someone on your team made a different decision because of a persona?

Not referenced one. Not name-checked one. Made a different decision, actually changed the outcome, chose to build something or not build something, prioritised differently, killed an experiment, rewrote a piece of copy. Because of what a persona told them.

If you can answer with a specific example from the last few weeks, your personas are alive. If you have to reach back six months, or if you can't think of anything at all, they died, and probably some time ago, and nobody told you.

The job isn't to fix the persona doc. The job is to build a practice where customer understanding stays current enough, specific enough and well-evidenced enough that decisions actually get made differently because of it. Documents can't do that. Practices can.

If you've inherited a persona doc that no one uses, or watched one die in real time, I'd love to hear how it went. The comments are open.

References

[1] Bastow, J. Why User Personas Fail (And How to Make Them Successful). ProdPad. https://www.prodpad.com/blog/user-personas/

[2] Forrester. Q2 2023 Portfolio Marketing Survey, cited in Are Your Personas Powerful Enough?, Forrester Research. https://www.forrester.com/blogs/are-your-personas-powerful-enough/

[3] Bulanova, A. How to Bring Your User Personas to Life. Mind the Product, 2016. https://www.mindtheproduct.com/how-to-bring-your-user-personas-to-life/

[4] Torres, T. About. Product Talk. https://www.producttalk.org/about/

[5] Torres, T. Continuous Discovery Habits. Product Talk. https://www.producttalk.org/continuous-discovery/

[6] Bulanova, A. How to Bring Your User Personas to Life. Mind the Product, 2016. https://www.mindtheproduct.com/how-to-bring-your-user-personas-to-life/

[7] Shaw, J. Why Most B2B Personas Are Useless. LinkedIn. https://www.linkedin.com/pulse/why-most-b2b-personas-useless-jodie-shaw/