Skip to main content

Percentile

A statistical measure indicating the percentage of scores that fall below a value

A percentile rank indicates the proportion of scores in a distribution that a particular value exceeds. Being at the 85th percentile means your score is higher than 85% of all scores in the reference population, providing intuitive context for individual performance.

Understanding Percentile Ranks

Percentiles divide a distribution into 100 equal parts. The 50th percentile (median) splits the population in half. Unlike raw scores, percentiles immediately communicate relative standing - a reaction time at the 90th percentile means you responded faster than 90% of test-takers. Percentiles are non-linear; the difference between the 50th and 60th percentile in raw score is typically smaller than between the 90th and 95th.

Percentiles vs. Other Metrics

Percentiles differ from percentages (which measure proportion correct) and from standard scores like z-scores or IQ scales. While z-scores assume normal distribution and allow arithmetic operations, percentiles work with any distribution shape and are more intuitive for general audiences. However, percentiles cannot be averaged or directly compared across different tests without shared norms.

Percentiles in Cognitive Benchmarking

Benchmark platforms use percentiles to contextualize individual results against population norms. Your reaction time of 215 ms gains meaning when expressed as the 72nd percentile. Reference populations matter - percentiles calculated against all users differ from those restricted by age group or experience level. Larger sample sizes produce more stable and reliable percentile estimates.

Raw scores and percentiles - reading the same result two different ways
AspectRaw score (milliseconds, WPM)Percentile
What it expressesThe measured absolute value itselfRelative standing within a population
What you learnHow fast you areWhere you sit within the same group
Assumption about the distributionNone in particularNo normal distribution assumed (any distribution shape works)
Handling a skewed distributionThe mean can drift from what you feelStill works appropriately when the distribution is skewed
Keeping comparisons fairDifferences in test conditions ride along untouchedSplit by age group or experience level before comparing
What it cannot doRelative standing is hard to read offBeing averaged, or compared directly across tests without shared norms

The 50th percentile is the median, and the 90th percentile or above puts you in the top 10%. The same 10-point gap covers a larger raw-score difference in the higher bands.