--- title: "Percentile" description: "A percentile is a position, not a proportion. A percentile of 70 means the person scored at or above 70 per cent of the norm group; it does not mean they answered 70 per cent of the questions correctly. Percentiles are also compressed in the middle of a distribution and stretched at the extremes." canonical: https://www.assessall.com/guides/glossary/p/percentile updated: 2026-09-25 source: AssessAll --- # Percentile _Also known as: Percentile rank._ A percentile is a position, not a proportion. A percentile of 70 means the person scored at or above 70 per cent of the norm group; it does not mean they answered 70 per cent of the questions correctly. Percentiles are also compressed in the middle of a distribution and stretched at the extremes. ## The compression problem, which nobody warns readers about Because most people cluster near the middle of a normal distribution, a small difference in raw score produces a large difference in percentile in the middle, and a large difference in raw score produces almost none at the tails. The practical effect is that the gap between the 45th and 55th percentile is trivial in the underlying attribute, while the gap between the 95th and 99th is substantial. A shortlist that cuts at a middle percentile is slicing through the densest part of the distribution, where the instrument separates people least. ## The band around a mid-range percentile is 55 points wide That figure is worth working through, because it is the number this page previously gestured at without giving. Take a scale with a standard deviation of 10 and a reliability of 0.85. The [standard error of measurement](https://www.assessall.com/guides/glossary/s/standard-error-of-measurement) is 10 × √(1 − 0.85) = 3.87 scale points, so the 95 per cent interval around any single score is ±1.96 × 3.87 = ±7.6 points. Convert that band into percentiles and the compression described above does the rest. A candidate sitting exactly at the 50th percentile has a 95 per cent interval running from the **22nd to the 78th percentile** — a band 55 percentile points wide. The same ±7.6 points placed at the 95th percentile runs from the 81st to the 99th: a band of **18 points**. Identical measurement error, three times the apparent precision at the top of the range. Two consequences, and both are about where you cut. A shortlist drawn at a mid-range percentile is drawing a line through the part of the scale where the number is least trustworthy, and a report that prints a mid-range percentile to the nearest whole number is implying a precision of roughly one point where the honest figure is fifty-five. Percentiles are at their most informative in the tails, which is the opposite of how they are usually used. This is the error around one person's percentile. Comparing two people's percentiles needs the wider [standard error of the difference](https://www.assessall.com/guides/glossary/s/standard-error-of-the-difference), which on the same scale is 5.48 points rather than 3.87. ## How to read one responsibly Read the norm group first and the number second. Then read the number as a band rather than a point — and the band is far wider than readers assume. On a scale with a standard deviation of 10 and a reliability of 0.85, the 95 per cent interval around one score is plus or minus 7.6 scale points. Placed at the 50th percentile, that interval runs from the 22nd percentile to the 78th. And never average percentiles. Percentile ranks are ordinal, so their mean has no defined meaning; convert to a standard score, average that, and convert back if a composite is needed. --- Source: https://www.assessall.com/guides/glossary/p/percentile