Designing for one person is harder than for a million

Why a single user removes every statistical excuse.

Designing for one person is harder than for a million because a product built for a huge, varied population can succeed by satisfying the comfortable middle of a statistical distribution while quietly failing at its edges without anyone much noticing, whereas a product built for exactly one specific person has no average to hide behind, since that one person either finds the design usable or does not, with no crowd of other users around to dilute the failure into an acceptable statistic.

Norman's The Design of Everyday Things is mostly about doors, and it ruins doors permanently. I have not walked through an unfamiliar door the same way since reading it, since every handle, plate and hinge now announces, before I have consciously decided to notice, whether it was designed for the person actually using it or merely for a population that person happened to belong to.

Introduction and overview

A product designed for a large population can be judged a success by a statistical standard, comfortable for the great majority, tolerable for most of the rest, and genuinely unusable for a small remaining minority the designer may never even meet directly. That minority's dissatisfaction is real, but at scale it is absorbed the way any large system absorbs a small, predictable loss, refunds issued, complaints logged, a slightly lower satisfaction score averaged in among a great many higher ones, none of it threatening the product's overall success in any way a spreadsheet would flag as urgent. The designer of that mass-market product can, in a sense, afford to be a little wrong about a meaningful slice of the people who will eventually use it, because the ones the design does serve well are numerous enough to carry the average, and being wrong about a tenth of a million users still leaves nine hundred thousand people the product genuinely works for. A product designed for one specific named person offers no equivalent cushion, since there is no larger population for that one person's dissatisfaction to be diluted into, and a design that fails them once is, for the only user who will ever actually use it, simply a design that has failed.

The custom-bicycle-frame comparison

A shop selling bicycles in three or four standard frame sizes serves the overwhelming majority of riders acceptably well, since most people's proportions cluster close enough to the middle of the range that a size chosen one notch up or down from an exact fit is still comfortable enough for ordinary riding. A frame built from scratch to one specific rider's exact inseam, reach and arm length has no such margin to hide behind, since there is only the one rider it was built for, and a reach that ends up a centimetre too long will announce itself in that rider's shoulders on every single ride rather than disappearing into a chart of mostly satisfied customers somewhere else entirely. There is nowhere for that discomfort to go, no larger group of comfortable riders to be quietly averaged in against, no follow-up model a season later aimed at fixing the complaint for the next batch of buyers, only the one rider, riding the one frame, feeling the one mistake on every outing until it is corrected directly.

Why a population absorbs failure and an individual cannot

A statistical design target is, in effect, a decision about how much of the intended population is worth serving well against how much effort that requires, and choosing to cover, say, the middle ninety percent of an intended population by design is also, quietly, a choice to leave the remaining tenth without adequate provision, an acceptable trade at the scale of a whole market and not remotely an acceptable trade at the scale of one person who happens to fall on the wrong side of that particular line. This is precisely why designing for one person, a specific assistive device, a piece of equipment built for one named operator, a tool fitted to one particular pair of hands, forces a completely different discipline, since every dimension chosen has to be checked against the one real person it will actually serve rather than against a curve describing a population that person is not, in any useful sense, a representative sample of.

One figure worth keeping in mind

A mass-market product's design target commonly aims to serve something like the middle ninety percent of its intended population, a range that still leaves a genuine tenth of that population underserved by design, a proportion small enough to be treated as an acceptable rounding error at the scale of a market and yet entirely capable of describing the single specific person a one-off design is actually meant for.

What this changes in practice

Designing for one person means gathering the actual measurements, preferences and constraints of that specific individual before a single dimension is fixed, rather than starting from a population average and hoping the one real user happens to sit close enough to it, since the entire safety margin a statistical design quietly relies on simply does not exist when there is only one person the outcome will ever be checked against. It also means treating feedback from that one person as complete information rather than as a single noisy data point to be smoothed against a larger sample, since for a one-off design, that one person's report genuinely is the entire population the product has to satisfy. A complaint that would be dismissed as an outlier on a mass-market product, one voice among many, is instead the single voice a one-person design was built to answer, and treating it with anything less than full seriousness defeats the entire purpose of designing individually in the first place.

Where this stops being true

None of this means population-level statistics are useless once a genuinely individual design is underway, since anthropometric data and broad usability findings remain a sound starting point for a first draft even of a one-person design, a place to begin measuring from rather than a substitute for actually measuring the person in front of the designer. The distinction that matters is not whether population data gets consulted at all, but whether it is ever allowed to stand in for the one measurement that a one-person design can always, and cheaply, simply go and take directly.