Percentile Calculator
From one dataset: any percentile, the common ones, a full table, and the rank of a specific value.
How to Use
- Paste the data
Commas, spaces or newlines all work; non-numeric entries are skipped and the note tells you how many.
- Set the percentile
Enter a number between 0 and 100, decimals allowed, to read one percentile directly; the table below stays consistent with it.
- Rank a value
Enter a value and the panel reports what share of samples it strictly exceeds and what share it reaches — the "where do I place" question.
FAQ
Percentile versus percentage?
A percentage is a share of the whole; a percentile is a position in the sorted order. Scoring 85 when 85 maps to P90 means 90% of people scored lower — the score is 85, the position is 90. In the same data these two numbers often differ a lot, especially with an uneven distribution.
Why do different tools report different percentiles?
Because there is no single convention. Nine methods are in common use: some take the ⌈n×p⌉-th value, some interpolate between neighbours, some base the calculation on (n+1) instead of n−1. With large samples the difference is negligible; below about 20 samples you can get entirely different numbers. This tool uses R-7 (linear interpolation), matching Excel's PERCENTILE.INC and numpy's default.
What does P95 response time mean?
Sort every request's duration in that window from fastest to slowest; the value at the 95% position is P95. It means 95% of requests were faster and 5% slower. For SLAs it beats the average, which gets dragged up by a few very slow calls while the many fast ones mask the long tail.
What if that value is not in my data?
Linear interpolation produces an in-between value that may not exist in the sample. The P90 of the numbers 1 to 10 is 9.1, which is not one of them. That is expected — interpolation gives a smooth position estimate.