Open any product team's research folder and you'll meet her. Sarah, 34, marketing manager. She values efficiency. She reads the onboarding. She compares options on a spreadsheet before she buys. She is calm, attentive, and she has never once existed.

We build personas like Sarah because they're easy to design for. A user who weighs costs against benefits and picks the best one is a user you can reason about. You can predict her. You can A/B test against her. The trouble is that almost nobody behaves like her, including the people who wrote her.

The fiction has a name. Economists spent a century modeling homo economicus, a creature who knows what he wants and acts to get it. It made the math work. It also described no human being who has ever lived. Daniel Kahneman built a career, and won a Nobel, partly on showing how far real decisions drift from that model.

UX research inherited the fiction without noticing. Look at the questions we ask. "Would you use this feature?" "What matters most to you in a checkout flow?" "How likely are you to recommend us?" Every one of them assumes a person who can inspect his own motives and report them back accurately. That person is rare. Most of us act first and explain second. The explanation feels like the cause, but it showed up after the fact, assembled to make the behavior sound reasonable.

This is why stated preference and revealed preference keep coming apart. Survey people about what they want in a phone and battery life wins every time. Watch what they actually buy and it's the thinner one with the better screen. Ask a team which features users are begging for, and you'll get a backlog of earnest requests. Build them, and the usage stays flat.

People aren't lying. They're answering a different question than the one you think you asked. "Would you use a budgeting feature?" gets answered as "Am I the kind of person who should budget?" The yes is real. It's about identity, not behavior. You shipped to the second answer and got punished for it.

Jakob Nielsen's eye-tracking work showed years ago that users don't read pages. They scan them in a rough F-shape, catch a few words, and click the first thing that looks close enough. Steve Krug built a whole book around the instruction Don't Make Me Think. The real user satisfices. They take the first option that clears the bar, not the best one, because finding the best one costs attention they aren't willing to spend on your settings page.

Design for the optimizer and you'll keep adding the thing the optimizer would want: more options, more controls, more to weigh. The satisficer reads none of it. They see a wall and bounce.

Watch how a three-tier pricing page actually works. The middle plan sells, not because anyone audited its features, but because it sits between a cheap one that feels insufficient and an expensive one that feels indulgent. Move the prices and the "right" choice moves with them. Nothing about the product changed. The frame did, and the frame is most of the decision.

An economist would call that irrational. The customer paid more because of a number sitting next to another number. But there's nothing broken about it. It's how perception works. We don't read value off an absolute scale. We read it off whatever happens to be standing nearby.

The cost of designing for Sarah is that you optimize the variable nobody feels. You shave a step off a flow that already felt fine and leave the part that made people anxious untouched. You win the argument in the research readout and lose the behavior in the wild.

The irrational majority isn't a problem to be corrected. It's the actual material you're working in. The user who acts on feeling and explains afterward isn't a degraded copy of the rational one you wish you had. They're the only one who was ever going to show up. Design for them, and the work starts to fit the people using it.