Choosing which population percentile to design for decides who gets excluded

What it is

Body dimensions form a distribution across a population, not a single value, and a percentile marks a position on that distribution: the 5th percentile of a dimension is smaller than 95% of the population; the 95th percentile is larger than 95% of the population. Design coverage is the share of the population a design actually accommodates once a percentile range has been chosen. Which percentile to pick is not arbitrary — for the same dimension, opposite kinds of design constraints call for opposite directions.

Why it happens

The key distinction is whether a dimension is an upper-bound constraint or a lower-bound constraint. Clearance dimensions — hatch openings, seats, aisles — that must leave room for larger bodies should be designed to a high percentile (e.g., 95th percentile stature) so the largest target users can still pass through or sit down; designing to the average leaves the larger portion of the population physically blocked. Reach dimensions — control placement, pull-handle height — that must remain accessible to smaller bodies should go the other way, to a low percentile (e.g., 5th percentile arm length), so the shortest-armed target users can still reach; designing to the average leaves shorter-armed users unable to reach. The same population and the same dataset call for opposite percentile directions depending on whether the dimension is clearance or reach, and confusing the direction is the most common misuse. Coverage also compounds down when a single design must simultaneously satisfy several independent dimensions: if each of five independent dimensions covers 90% of the population on its own, the share of people who simultaneously satisfy all five is far below 90%, approximately the product of the individual coverage rates.

Studying it

The percentile values behind a coverage calculation come from large-scale anthropometric surveys (military or civilian anthropometry databases), which report dimension distributions stratified by sex and age band and give commonly used percentile points for lookup — not a single value describing a "standard body."

Where it stops holding

Single-dimension percentile selection assumes the dimensions are independent, but in reality body dimensions correlate (hand length correlates somewhat with stature). Simply picking a percentile on each dimension independently and multiplying the coverage rates together underestimates or overestimates the true joint coverage; a real multivariate distribution is needed instead of an independence assumption.

Applying it

• Before applying a value, determine whether the dimension in question is a clearance constraint or a reach constraint: clearance takes a high percentile (covering larger bodies), reach takes a low percentile (covering smaller bodies). Getting the direction wrong systematically excludes part of the target population.
• When a single design must satisfy coverage on several independent dimensions at once, don't report the single-dimension coverage number as the overall coverage — calculate joint coverage from real multivariate data, or convert a single-dimension problem into an adjustability problem with an adjustable mechanism instead.
• Verification: recruit real participants from the target population whose bodies sit near the boundary of the chosen percentile range and have them actually perform the pass-through, seating, or reach action, rather than judging on paper from a looked-up number alone.

Related

• Same group: A11.06.2 Hand length, finger width, and grip dimensions · A11.06.4 Static dimensions vs. dynamic working dimensions · A11.06.5 Regional and generational differences in anthropometric data · A11.06.6 The design-to-the-average fallacy
• Adjacent: A8.25 Reach envelope and range of motion
• Search terms: percentile · design coverage · anthropometry · accommodation