Percentile Rank Calculator

Percentile rank: the position of a value in a distribution

The percentile rank of a value x in a dataset of size n is the share of observations less than or equal to x: P(x) = (count ≤ x) / n · 100%. Say you have 100 exam scores and 75 of them sit at or below 80. Then 80 lands at percentile 75, meaning three quarters of candidates scored at most that much. A few conventions are worth knowing. The inclusive method counts x itself; the exclusive one leaves it out (that's what Excel's PERCENTRANK.EXC does); and the interpolated approach, which approximates the CDF, hands back fractional ranks. Whichever you pick, the payoff is the same. You can place a single value against a whole population without leaning on the absolute scale.

Applications

ENEM and SAT scores report which percentile a candidate landed in. WHO growth curves for children mark the typical range with P3, P50 and P97. The same idea shows up when you compare salaries and wealth, where the question is which percentile a given income falls in, and in engineering SLAs, where P95 / P99 latency is the response time that 95% or 99% of requests stay below.

FAQ

What is the difference between percentile and quartile? Quartiles are just the percentiles at 25, 50 and 75. So every quartile is a percentile, but the reverse doesn't hold.

Can two values have the same percentile? Yes. When values repeat, the ties share a rank, and the inclusive method tends to push tied values up to the higher percentile.

Why does engineering monitor P95 instead of the average? The average hides outliers. P95 latency tells you the worst experience 95% of users actually get, which is far more meaningful for SLAs than a mean dragged down by all the fast requests.