Percentiles, Explained
After reading this you can find the value that sits at any percentile of a data set, work out the percentile rank of a specific value, and know which of the several definitions you are using.
What a percentile is
A percentile splits your data by proportion. The 75th percentile is the value below which about 75 percent of the data falls. The median is just the 50th percentile.
Take these ten test scores: 55, 62, 68, 71, 74, 77, 80, 84, 89, 95. If someone scored 84, roughly 80 percent of the scores are at or below that, so 84 sits near the 80th percentile. Turn the question around and ask for the 75th percentile: the value comes out to 82 with linear interpolation. The calculator answers both directions. In Value at a percentile mode you give a percentile and get a value. In Rank of a value mode you give a value and get its percentile.
When to use it, and when not to
Percentiles shine when you care about position rather than distance. Growth charts, exam grading curves, salary bands, and response-time targets all use them. A web service that promises a "95th percentile latency under 200 ms" is saying that 95 percent of requests finish faster than 200 ms.
Percentiles resist outliers. Change the top score in the sample from 95 to 950 and the median stays at 75.5. The mean jumps from 75.5 to 161.5. That stability is why percentiles describe skewed data better than the mean does.
Percentiles need enough data to mean anything. With 8 values, the gap between neighboring order statistics is about 14 percentiles wide, so a "37th percentile" is really an interpolation between two points. Do not report a percentile to a precision your sample cannot support.
Avoid percentiles when you need to add or average groups. Percentiles do not combine: the 90th percentile of two merged data sets is not the average of their separate 90th percentiles.
The formula and the intuition
Sort the data into n values x_1 \le x_2 \le \dots \le x_n. For percentile p the calculator uses linear interpolation between order statistics. First find a rank position:
Here h is a fractional position on the sorted list, n is the count of values, and p is the percentile from 0 to 100. Split h into its whole part k and its fraction f, so h = k + f. The percentile value is then:
x_k is the value at whole-number rank k, and x_{k+1} is the next one up. The fraction f tells you how far to slide between them. When f = 0 you land exactly on a data point; when f = 0.5 you take the midpoint of two neighbors.
This is the definition used by many spreadsheet functions (the "inclusive" variant). It places the 0th percentile at the minimum and the 100th percentile at the maximum, which matches most people's expectations.
Worked example with the sample data
The 75th percentile of ten scores
Load the demo data 55, 62, 68, 71, 74, 77, 80, 84, 89, 95 with p = 75.
- The data is already sorted, and n = 10.
- Compute the position: h = \frac{75}{100}(10 - 1) + 1 = 0.75 \times 9 + 1 = 7.75.
- Split it: whole part k = 7, fraction f = 0.75.
- The 7th value is x_7 = 80 and the 8th is x_8 = 84.
- Interpolate: P = 80 + 0.75 \times (84 - 80) = 80 + 3 = 83.
So the 75th percentile is 83. Three quarters of the scores lie at or below 83. Note this is the upper quartile Q_3. The same method gives Q_1 = 71.75 at p = 25 and the median Q_2 = 75.5 at p = 50.
Going the other way: percentile rank
To find the percentile rank of a value x, invert the interpolation. Locate where x falls among the sorted values and read off the percentile that would land there. For a value between two data points, interpolate on position.
Ask for the rank of 80 in the sample. It is the 7th of 10 values, so its position gives p = \frac{7 - 1}{10 - 1} \times 100 = \frac{6}{9} \times 100 \approx 66.7. So 80 sits at about the 66.7th percentile. Check it: feeding p = 66.7 back into the value formula returns close to 80.
Value-at-percentile and rank-of-value are exact inverses here. If the calculator says the 75th percentile is 83, then asking for the rank of 83 returns 75. This round-trip is a quick way to sanity-check any result.
Reading and interpreting results
A percentile value shares the units of your data. The 75th percentile of scores is a score. Percentile rank is dimensionless, always between 0 and 100.
Two habits keep interpretations honest. First, state the direction: "at or below" is the usual convention, so the 90th percentile means 90 percent of values are at or below it. Second, remember the boundaries. The minimum is the 0th percentile and the maximum is the 100th under this method, so no value can rank above the largest observation.
Different software may disagree by a little. There are at least nine published definitions of a quantile. The inclusive linear method here can differ from an "exclusive" method by a few percent on small samples. On the demo data, the exclusive variant puts the 75th percentile at 84.25 instead of 83. Neither is wrong; report which one you used.
Common mistakes
- Forgetting to sort
- Every percentile formula assumes sorted data. The calculator sorts for you, but if you compute by hand on the raw order you will get nonsense.
- Confusing percentile with percentage
- Scoring in the 90th percentile is not the same as answering 90 percent of questions correctly. The first is your position among test-takers, the second is your raw score.
- Over-precision on tiny samples
- With five data points, only a handful of distinct percentile positions exist. Reporting the "63rd percentile" of five numbers implies detail that is not there.
- Averaging percentiles across groups
- You cannot pool percentiles. To get the 95th percentile of a combined data set, combine the raw data first, then compute once.
Related calculators
If you want the quartiles, median, and spread in one pass, use the Descriptive Statistics Calculator, which reports the same percentile-based quartiles alongside the mean and variance. When an extreme value is dragging your summary, the Outlier Detector applies Tukey fences, which are built directly from Q_1 and Q_3. For relationships between two variables rather than the shape of one, see the Covariance Calculator.
Frequently asked questions
What is the difference between a percentile and a quartile?
Quartiles are three specific percentiles. The first quartile is the 25th percentile, the second is the 50th (the median), and the third is the 75th. On the demo data those are 71.75, 75.5, and 83.
Why does my spreadsheet give a slightly different answer?
There are several accepted quantile definitions. This calculator uses inclusive linear interpolation, matching one common spreadsheet function. An exclusive method gives 84.25 rather than 83 for the 75th percentile of the sample. Match the method, not just the software.
Can a percentile be higher than every value in the data?
No. Under this method the 100th percentile equals the maximum, so no percentile exceeds the largest observation. The 0th percentile equals the minimum.
How much data do I need?
Enough that the percentile you report reflects real gaps between values. For a stable 95th percentile you generally want dozens of observations, not five. With small samples, treat any interior percentile as an estimate.
Is the median more reliable than the mean?
For skewed data, yes. The median (50th percentile) barely moves when one value is extreme, while the mean chases it. Replacing 95 with 950 in the sample leaves the median at 75.5 but pushes the mean from 75.5 to 161.5.
Related tools
Descriptive Statistics Calculator
Mean, median, quartiles, variance, skewness, outliers and more.
Kaplan–Meier Survival Calculator
Survival curve, life table and median survival from time-to-event data with censoring.
Outlier Detector
Grubbs' test, Tukey fences and z-scores on your data.
Time Series Lab: Decomposition, ETS & ARIMA Forecast
Paste a CSV: seasonal decomposition, ETS and ARIMA fits, forecast intervals and rolling-origin cross-validation.
Autocorrelation Calculator
ACF by lag with white-noise bands and the Ljung–Box test.
Covariance Calculator
Sample and population covariance with Pearson correlation.