Percentile Calculator

There is more than one accepted method. Linear interpolation between closest ranks is the most common; nearest rank always returns an actual data value. Both are shown here because both are in use.

Because they implement different definitions — inclusive and exclusive variants exist and disagree on small datasets. Neither is wrong; they answer slightly different questions.

The value below which 90% of the data falls. In performance monitoring it is used because the mean hides tail behaviour, and the tail is what users actually notice.

The percentage of values below a given one. If 60% of the data is below your value, you are at the 60th percentile rank.

Yes, by every method. It is the one percentile where the definitions coincide, which is part of why the median is so widely used.

Under the exclusive method the extreme percentiles are undefined for small samples, which is why it returns nothing there. The inclusive method caps at the maximum value.

A nine-point scale derived from percentiles, used in some educational testing. It groups percentile ranges into bands, which reduces false precision in test reporting.

As percentiles for age and sex. A child on the 25th percentile for height is taller than a quarter of their peers, and tracking along a percentile matters more than which one.

Because the mean hides tail behaviour, and the tail is what users notice. A page with a good mean and a terrible 99th percentile is slow for one visitor in a hundred, every time.

The general term — percentiles divide into 100, quartiles into 4, deciles into 10. They are all quantiles at different resolutions.

No, and doing it is a common error. Averaging the 95th percentiles of two servers does not give the 95th percentile of the combined traffic, which requires the underlying data.

Enough that the extreme percentiles have data behind them. A 99th percentile from 50 observations is essentially the maximum, which is not what it is meant to describe.