← Back to the blog

My UX personas don't have stock photos (and that's why they work)

UX personas have a bad reputation, and it's a deserved one. Most are a made-up name, a stock-photo smile, and three demographic facts that change no design decision. They get made, printed, pinned to a wall, and never looked at again.

The problem isn't the tool. It's that it's almost always built backwards.

The question that decides whether it's useful

Before building a persona, I ask myself: what design decision is this going to change?

If I have no answer, I don't make it. A persona that changes nothing is a graphic design exercise, not a research tool.

How I build UX personas from data

In the Mexx redesign I built two personas, Lucas and Sofía, and neither started with "let's imagine a user". They started with patterns that appeared in the survey and the audit.

Lucas, 26, developer. Not a fact that "would be nice" to have. It came from several respondents mentioning they bought specific software licenses and knew what they were buying. They knew the difference between versions and compared prices elsewhere first. That pattern matters because the interface doesn't need to explain what a digital product is. It needs to be fast and frictionless: he already decided what to buy before arriving.

Sofía, 32, freelancer. Came from the opposite pattern: people who researched on their phone during the week and bought at the weekend on a computer, comparing between tabs. That pattern explains why the cart lost between devices was a real problem and not one of my hypotheses.

Neither has a photo. They have documented behaviour, with each data point's source marked. That's the difference between a persona and what Nielsen Norman calls an evidence-based persona: the first describes someone, the second explains why they do what they do.

The proof it worked

After building Lucas and Sofía, I went back to the feature list we were about to prioritize. Cart recovery via magic link — which was in the middle of the list — moved up in priority because it's exactly what Sofía needs: getting back to a cart she built on another device without having to search for the product again.

That's the moment that separates a useful persona from a decorative one: when it changes the order of a list, not when it looks nice on a slide.

What I leave out on purpose

I don't add a full name, a hometown, or made-up hobbies. Every extra fact that didn't come from real evidence is an opportunity for someone on the team to design for the invented detail instead of the real pattern. If Lucas "lives in Palermo and likes mate", that sentence helps decide nothing — it just makes the persona feel more real without being so.

Fewer decorative details make it easier to see which ones matter.

When NOT to build personas

If you have fewer than five or six interviews or responses, there's no pattern yet: there are loose anecdotes. Building a persona out of that is inventing an average from a sample that isn't big enough. Better to say "I don't know what patterns exist yet" than to fabricate a persona that feels rigorous without being so.

Personas are a summary of evidence, not a substitute for it. If you never went back to the original evidence after creating them, you probably did the wrong exercise.

They are also worth revisiting when the product changes. A persona written for an early stage stops describing anyone once the real audience shifts, and keeping it out of habit is worse than having none: it lends the authority of data to an old memory.

The full process where personas fit in — after the survey, before defining the problem — is in how I design an interface from scratch.