Average human dimensions describe nobody
Anthropometric data, and how to use it without designing for a fiction.
Average human dimensions describe nobody in particular because an average calculated separately for each individual measurement, height, arm reach, hand width, sitting height, is a genuine statistical fact about a population while being almost nothing about any single real person in it, since a real body's own combination of measurements rarely lands close to the average on every dimension at once, however tightly each dimension considered on its own clusters around its mean.
What is really going on
Anthropometric data is gathered by measuring a large number of real people across a set of standard dimensions and then summarising each dimension separately, an average height, an average arm reach, an average hand width, each figure entirely legitimate as a description of that one dimension across the whole measured group. The mistake, an easy and common one, is treating a person built from all of those separate averages stitched together, average height, average reach, average hand width all at once, as though such a person is a realistic stand-in for an actual human being, when in practice a real person's own height, reach and hand width are only loosely related to one another, so an unusually average number on one dimension carries very little information about where that same person will fall on any of the others. A tall person is not reliably long in the arm relative to their height, nor is average sitting height reliably paired with average hand span, and the more separate dimensions a design tries to satisfy at once, the more these small, mostly independent mismatches stack up against any single composite person ever actually existing.
The shoe-size comparison
A shoe described by length alone, a single number stamped on the box, can fit two feet of identical length yet noticeably different width or arch height, since a single averaged dimension was only ever describing one aspect of a foot's actual shape rather than the whole of it. Two customers buying the same labelled size can walk out with a shoe that fits one of them well and pinches or slips on the other, not because the sizing system made an error but because it was never designed to capture more than the one dimension it actually measures, leaving every other dimension of fit to chance. This is why a well-fitted shoe traditionally carries a separate width fitting alongside its length, narrow, standard, wide, an acknowledgement that length alone was never enough information to guarantee a genuine fit for everyone wearing the same numbered size.
Why the average pilot never existed
When a cockpit, a seat, or any piece of equipment is designed around the average of several separate body dimensions measured across a large group, and that same design is then checked against every individual in the group on every one of those dimensions simultaneously rather than one at a time, the number of people who land close to average on all of them at once is typically far smaller than instinct expects, often close to none of the group at all, even though each individual dimension considered alone clusters neatly around its own mean. The apparently solid, singular figure of an average person built this way turns out to describe a composite that essentially nobody in the measured population actually matches, a useful abstraction for certain kinds of statistics and a genuinely dangerous one to design a single fixed cockpit, seat or handle around. Cockpit designers working from exactly this kind of averaged figure in the mid-twentieth century found, once they checked real individuals against every assumed dimension at once rather than one at a time, that almost nobody in a large measured group of pilots actually matched the average pilot the seat and controls had been built around, a finding that pushed adjustable seats and controls into becoming the ordinary standard.
The number that matters here
When one large group of people is measured across even a modest handful of separate body dimensions at once, the fraction of that same group who land close to the average on every single dimension simultaneously is far smaller than instinct expects, often close to none of them at all, even though each dimension considered in isolation clusters closely around its own mean.
What this changes in practice
Designing from anthropometric data well means designing for a range on every dimension that actually matters to the task, rather than for a single central figure on each one, providing real adjustability wherever a mismatch on that dimension would cause a genuine problem, a seat that slides, a handle offered in more than one size, a control reachable across a stated range of arm lengths rather than fixed at one assumed length. A car's driver seat, steering column and mirrors adjusting independently of one another is exactly this principle built into an everyday object, since a single fixed driving position chosen for one average body would leave a real share of drivers unable to reach the pedals safely or see over the wheel. Where adjustability genuinely is not practical, designing to accommodate the extremes of the range on whichever dimension the task cannot tolerate failure on, a doorway height, an emergency handle's reach, matters far more than designing to the comfortable middle, since the middle was never the group actually at risk of being left out.
Where this stops being true
Not every design decision needs this level of individual accommodation, and some genuinely benefit from a single fixed dimension chosen sensibly rather than a fully adjustable one, a standard sheet of paper or a standard doorway width serving everyone acceptably well precisely because the cost of a mismatch on that particular dimension is low and easily tolerated by nearly anyone. The distinction that decides which approach a given dimension deserves is not how variable real people are on it, since people vary considerably on almost everything, but how costly a mismatch on that particular dimension actually turns out to be for the person who experiences it. A doorway a little too low is a minor inconvenience for a tall person ducking through it once; a driving position that cannot be reached safely is a hazard on every journey, and it is that difference in consequence, not the raw variation in the underlying population, that decides where fixed dimensions stay acceptable and where they stop being safe.