A team can spend nine months building the wrong product with total competence. The reviews pass, the code is clean, the design is immaculate. Nothing in the process raises a hand, because nothing in it was ever pointed at a real person.
This is not carelessness. Work on a product for six months and you lose the ability to see it fresh. You know where the settings live. You know what the icons mean, and which of three similar buttons is the right one. You cannot un-know any of it. And that knowledge rewrites every judgement you make about what counts as obvious.
"Obvious" comes to mean "obvious to me, now" — and the person it is obvious to has spent half a year inside the product.
So the team invents a stand-in user, built out of the people in the room. This person is patient. They are technically literate, undistracted, and reading every word on screen. Your real user is on a phone, halfway through something else, opening the product for the first time in three weeks. And when they get stuck, they almost never tell you. They hesitate. They take a wrong turn, invent a workaround, or leave. Nobody files a report. So silence gets read as approval, and the team stays confident long after the design stopped working.
Where the term comes from
"User-centered system design" was the title of a 1986 book edited by Donald Norman and Stephen Draper. Norman went on to head a three-person group at Apple in the early nineties called the User Experience Architect's office — the first time "user experience" appeared in a job title.
Don Norman, jnd.org
What it actually means
Strip away the vocabulary and one rule is left. When a product question comes up, what people do settles it, not what the room thinks. Personas, journey maps and usability labs are just tools for getting there. You do not need any of them to follow the rule.
Compare two sentences. "I think this is confusing." And: "Four of the five people we watched stopped here." Those are not strong and weak versions of the same thing. Only one of them ends the argument.
Four things it is not
- Asking people what they want. They are poor predictors of their own behaviour, and reliably polite when asked directly.
- A phase that runs before design and then closes.
- A synonym for minimalism, or for any particular visual style.
- Being nice to users. The user-centered decision is often the one that removes a feature somebody asked for.
Why a wrong assumption gets harder to fix every week
In a wireframe, a wrong assumption costs a conversation. In production it costs a data migration, a support queue, and the awkward meeting where someone admits a shipped feature was a mistake.
But the money is not the real problem. Early decisions turn into structure. A navigation model picked in week two becomes the frame six features hang off. Question it in month five and you are not changing a screen. You are unpicking everything built on top of it. Which is why research pays best at the exact moment it feels too early to bother. That is the last point where changing your mind is still cheap.
85%
of a design's usability problems are surfaced by just five participants — Jakob Nielsen, 2000, building on earlier research with Tom Landauer.
How much testing is enough
One person on their own shows you nearly a third of the problems. Every extra participant finds less than the last, because they keep walking into the same walls.
The part people skip is what Nielsen actually told them to do. Not "test with five people." Take the budget you would spend on one fifteen-person study, and run three studies of five instead. Fix what you find between rounds.
One large test tells you what is wrong once, expensively, and usually too late to act on. Three small ones let you find a problem, fix it, and check whether the fix worked.
What it looks like in practice
Pick the task where money or commitment changes hands. Find five people who look like your users. Hand them the task, then say nothing.
The silence is the whole method. Every hint you give is a hint your real users will never get, and the urge to help is very hard to resist.
Watch for hesitation, not failure. Outright failure is rare. What you want is the half-second stall, the cursor going in circles, the label somebody reads twice. When three different people pause in the same spot, you have found something real.
Try this
Five people. One task. One afternoon.
The expensive version of research is the one most teams imagine, and therefore never start. "We do not have budget for user testing" is almost always a misunderstanding of the price.
The research you already own
- Support tickets, grouped by theme rather than by ticket.
- The queries people type into your product's own search box — a list of things they expected to find and did not.
- The step in a flow where sessions end.
- The fields left blank on every submission.
Someone exporting to a spreadsheet to do a job your product already does is showing you exactly where their thinking and yours parted ways. No survey answer will ever be that honest.
Where it gets complicated
Five people find common problems. They will not find rare ones, and a bug that hits two percent of users can still be the one that loses their data.
Expert tools flip the goal entirely. Designing for a first-timer can wreck a product that someone uses for eight hours a day, because that person wants density and keyboard shortcuts, not gentle onboarding.
In B2B, the buyer is often not the user. Both are real, and what they want genuinely pulls in different directions.
And in a brand new category, watching people has a ceiling. They can tell you what annoys them about what already exists. They cannot describe something they have never seen. Research is very good at finding problems and quite bad at inventing answers. That gap is where design judgement earns its money.
Three things are going on. Teams cannot see their own products clearly. They swap in an imaginary user without noticing. And every week a wrong assumption survives, it gets harder to pull out.
No methodology fixes that. What fixes it is putting real behaviour in front of decisions while those decisions are still cheap to change.
So here is one thing to do this week. Take the flow in your product where the most is at stake. Find five people who do not work with you. Watch them try it, and do not help. Whatever stings the most is the thing to fix first.