The average that hides everyone

In the 1950s the United States Air Force had a problem it could not explain. Pilots were losing control of aircraft that were, by every measurement, correctly built. The cockpits had been designed around the average pilot, using body measurements collected a decade earlier.

A young researcher named Gilbert Daniels decided to check how many pilots actually were average. He took ten measurements that mattered for cockpit design — height, chest circumference, arm length and so on — and defined 'average' generously, as the middle thirty per cent of the range for each one. Then he measured four thousand and sixty-three pilots.

Not one of them was average on all ten. Not one. Even using only three of the ten measurements, fewer than three and a half per cent fitted.

The cockpit had been designed for a person who did not exist. Every pilot was being asked to adapt to a shape that matched nobody, and the mismatch showed up as accidents.

The solution was not a better average. It was adjustment: seats that moved, pedals that slid, straps that lengthened. The fix cost money and the manufacturers resisted it, but accident rates fell once it arrived.

The statistical point is simple and widely ignored. An average is a summary of a group, not a description of a member. It is a useful number when you are comparing groups, and a dangerous one when you are designing for an individual. Any time a system is built for the average user, it is worth asking Daniels's question: how many real people does this actually fit?