I have sat through plenty of project discussions about “the user”. We talk about what the user wants, how the user behaves and how quickly the user should be able to complete a task.

Before long, that user becomes surprisingly specific. They can see the screen, operate a mouse, read at the expected speed and remember the information from the previous step. They have a modern device, a reliable connection and enough uninterrupted time to finish the process.

It is an impressively detailed person, considering they do not exist.

How average became normal

Fernando Estrada Franco explores the history of this idea in his article, The Invention of Normal.

The word “normal” originally belonged to geometry. It described something perpendicular, drawing on the Latin norma, or carpenter’s square. According to philosopher Ian Hacking, it acquired its familiar modern meaning during the nineteenth century.

The Belgian astronomer and statistician Adolphe Quetelet then helped connect normality with the mathematical average.

Astronomers used averages to reconcile slightly different measurements of the same object. The star existed, so variation between observations could reasonably be attributed to measurement error.

Quetelet applied the same mathematics to people. In 1835, he introduced l’homme moyen, the average man. He later used the chest measurements of 5,738 Scottish soldiers to demonstrate how physical characteristics could form a bell-shaped distribution.

The important step was not simply calculating an average. Quetelet treated that average as an ideal, combining physical measurements with ideas about moral and social behaviour.

A statistical description had become a model of what a person should be. When the average becomes the ideal, real people become deviations from it.

Francis Galton took the argument in a more troubling direction. Quetelet placed the ideal at the centre of the curve, while Galton wanted to rank human qualities and encourage those he considered desirable. In 1883, he gave that project a name: eugenics.

This history matters because measurement is not automatically neutral. Data can help us understand variation, identify barriers and improve services, but it can also turn difference into deficiency and preference into supposed scientific fact.

The problem is not the average itself. The problem begins when an average stops describing a group and starts judging an individual.

The cockpit that fitted nobody

The most practical example in Fernando’s article comes from the United States Air Force.

Cockpits were still being designed using body measurements taken from male pilots in 1926. The assumption was that designing around average dimensions would create equipment suitable for most pilots.

By the early 1950s, the Air Force was undertaking much more extensive anthropometric research. This is the measurement of human bodies to inform the design of equipment, clothing and working environments.

Gilbert Daniels examined measurements from 4,063 serving Air Force personnel. For his 1952 technical report, The “Average Man”?, he selected ten dimensions and defined an average range for each.

Not one person was average across all ten.

The measurements were not wrong. Combining multiple averages had simply produced a theoretical person who was unlikely to exist.

This finding became part of a wider shift towards more adjustable cockpit design. Seats, pedals, controls and restraints could accommodate variation instead of requiring every pilot to match a fixed configuration.

Fernando captures the principle perfectly:

“They did not correct the pilots. They corrected the seat.”

That might be one of the clearest explanations of accessible design I have encountered. The pilots were not the design problem. The fixed environment was.

We are still building fixed cockpits

Digital services may look very different from a twentieth-century aircraft, but the underlying mistake is familiar.

We design around an assumed user and describe everybody else as an exception. The assumed user can see the verification code displayed on screen, hear instructions on a telephone call and complete a form before the timer expires.

They can drag an item with a mouse, remember a password and understand an error message written by the development team. When someone cannot do those things, we reach for an accommodation.

Perhaps there is a separate process. Perhaps customer support can complete the task on their behalf. Perhaps an accessibility feature will be considered in a future release.

The system stays fixed while the person is expected to find a way around it.

As someone who uses VoiceOver and other assistive technology every day, I recognise this pattern. The barrier is rarely described as a design decision. It is presented as an unfortunate mismatch between the user and a service that apparently works for everyone else.

A fixed cockpit is still a fixed cockpit, however polished the controls look.

Accessibility is the adjustable seat

Accessibility starts with the recognition that human variation is ordinary.

People see, hear, move, understand, remember and communicate in different ways. Those abilities can also change with age, illness, injury, environment and circumstance. There is no permanent, fully capable default user against whom everybody else should be measured.

An adjustable digital service might provide:

• Different ways to perceive information, including text, audio and captions
• Keyboard, touch, pointer, voice and assistive technology input
• Text resizing and reflow without losing content or functionality
• Enough time to read, understand and complete a task
• Clear language and helpful error recovery
• More than one route through authentication and customer support

These are not optional extras attached to the real product. They are ways of making the product fit the people who actually use it.

Good accessibility does not lower the standard. Adjustable seats did not make aircraft less rigorous or pilots less capable. They enabled more pilots to operate safely and effectively.

The same principle applies to digital systems.

The history of body mass index offers another example. The calculation began with Quetelet’s work on population measurements and was known as the Quetelet Index until Ancel Keys and his colleagues named it the body mass index, or BMI, in 1972.

BMI may be useful when studying populations, but used alone it provides a limited description of an individual. It cannot capture the full reality of somebody’s health, body or circumstances.

This is the recurring mistake: taking a population-level pattern and turning it into a complete judgement about one person.

French physician and philosopher Georges Canguilhem offered a more useful way to think about health. For him, health was not simply conformity to a fixed norm. It involved being able to respond to changing circumstances and establish new ways of functioning.

That idea translates well to accessibility. A successful service should not depend on everybody functioning in one predetermined way. It should be robust enough to support different people, methods and circumstances.

Standards are the start, not the outcome

The Web Content Accessibility Guidelines give organisations an essential framework for identifying and preventing barriers. However, a technically conformant component can still sit inside a journey that somebody cannot complete independently.

A form may have correctly labelled fields but impose an unrealistic time limit. An application may work with a keyboard but send its verification code through an inaccessible channel. A website may pass an automated test while its content remains confusing and its error messages offer no useful recovery.

Conformance matters, but it is not the same as a successful outcome.

The real test is whether people with disabilities can achieve what they came to do, with a reasonable level of effort and without surrendering their independence. That requires testing complete journeys with real people, not simply inspecting individual controls against an imagined user profile.

The average user is rarely documented explicitly. They hide inside requirements, acceptance criteria, policies and procurement decisions.

A useful design review should ask:

1. Which abilities are we assuming?
   Does the journey depend on sight, hearing, dexterity, memory or reading speed?
2. Where have we created only one route?
   Look at authentication, communication, payment, support and error recovery.
3. What happens when the user cannot conform?
   Is there a genuine alternative, or must somebody else complete the task?
4. What are our exceptions telling us?
   Repeated adjustments, abandoned journeys and support requests may reveal a structural barrier.
5. Have people with disabilities tested the complete journey?
   Component checks cannot reproduce the accumulated effort of using a real service.

These questions do not replace accessibility standards. They help turn those standards into services people can actually use.

Correct the seat

Normality encourages us to sort people into those who fit and those who require special treatment. Accessible design offers a more useful approach.

Start with human variation. Build flexibility into the requirements, architecture, interface and support model. When people struggle, examine the environment before deciding that the individual is the problem.

There is no average customer, employee or citizen waiting to use the systems we create. There are only people, each with a particular combination of abilities, preferences and circumstances.

If the design fits nobody particularly well, we should stop correcting the users and adjust the seat.

Sources

• The Invention of Normal, Fernando Estrada Franco, LinkedIn, 24 July 2026
• Sur l’homme et le développement de ses facultés, Adolphe Quetelet, 1835
• The History of Statistics, Stephen M. Stigler
• Notes on the History of Normality, Scandinavian Journal of Disability Research
• The “Average Man”?, Gilbert S. Daniels, US Air Force technical report, December 1952
• Inquiries into Human Faculty and Its Development, Francis Galton, 1883
• Adolphe Quetelet, the average man and indices of obesity
• The Normal State, Ian Hacking, Cambridge University Press
• Reasoning in Life: Values and Normativity in Georges Canguilhem