Percentiles and Quartiles: What They Mean and How to Calculate Them

A test score of 720 sounds impressive — until you learn nothing else about it. Is that out of 800? Out of 1600? What did everyone else score? Percentiles and quartiles answer a more useful question than the raw number ever could: where does this value rank relative to everyone else's? A score at the 95th percentile tells you something immediately meaningful, regardless of the underlying scale.

What a Percentile Means

Percentiles and quartiles sit alongside mean, median, and mode and standard deviation as the core toolkit of descriptive statistics — where those measures summarize the centre and spread of a whole data set in a single number, percentiles let you place any individual value precisely within that same distribution.

The k-th percentile is the value below which k% of the data falls. The 50th percentile is the median — half the data falls below it, half above. The 90th percentile is the value below which 90% of the data falls, leaving the top 10% above it.

This is why standardized test results are so often reported as percentiles rather than raw scores: "scored at the 85th percentile" is immediately interpretable — you outperformed 85% of test-takers — regardless of whether the underlying test was scored out of 100, 800, or 1600.

Quartiles: Percentiles at the Quarter Marks

Quartiles are simply the 25th, 50th, and 75th percentiles, given their own names because they divide a data set into four equal-sized groups:

• Q1 (first quartile) = 25th percentile
• Q2 (second quartile) = 50th percentile = the median
• Q3 (third quartile) = 75th percentile

Together with the minimum and maximum, these three values form the five-number summary — a compact description of a data set's centre, spread, and shape that's the exact basis of a box plot: the box spans Q1 to Q3, the line inside marks the median, and the whiskers typically extend to the most extreme values within 1.5×IQR of the box.

Worked Example: Finding Quartiles

Problem: Find Q1, Q2, and Q3 for the data set: 12, 15, 18, 20, 22, 25, 28, 30, 35.

Sorted (already sorted), n = 9. Median (Q2): the 5th value = 22.

Lower half (values below the median): 12, 15, 18, 20 → Q1 = median of these four = (15+18)/2 = 16.5

Upper half (values above the median): 25, 28, 30, 35 → Q3 = median of these four = (28+30)/2 = 29

IQR = Q3 − Q1 = 29 − 16.5 = 12.5

Finding a Percentile Rank: The Reverse Direction

Sometimes you have a specific value and want to know its percentile rank — the opposite direction from finding the value at a given percentile. The basic formula:

Example: in the data set above (12, 15, 18, 20, 22, 25, 28, 30, 35), what's the percentile rank of 25? Five values fall below 25 (12, 15, 18, 20, 22). Percentile rank = (5/9) × 100 ≈ 55.6th percentile.

Percentiles vs. Percentages: A Common Confusion

These sound alike but measure completely different things. A percentage is a proportion of a total — scoring 85% on a test means getting 85 out of 100 possible marks. A percentile is a rank relative to other people — scoring at the 85th percentile means outperforming 85% of other test-takers, regardless of your raw percentage score. It's entirely possible to score 60% on an unusually difficult test and still land at the 90th percentile, if most other test-takers scored even lower.

Percentiles and the Normal Distribution

For normally distributed data, percentiles connect directly to z-scores: a z-score of 0 is always the 50th percentile; a z-score of +1.645 is approximately the 95th percentile; a z-score of −1.645 is approximately the 5th percentile. This is exactly how standardized test percentiles are often calculated in practice — convert a raw score to a z-score using the population mean and standard deviation, then convert that z-score to a percentile using the standard normal distribution.

Real-World Uses

Growth charts: a paediatrician plotting a child's height or weight "at the 40th percentile for age" is using exactly this concept — the measurement relative to same-age peers, not an absolute number.

Standardized testing: the SAT, GRE, and most national exams report both a raw/scaled score and a percentile rank, precisely because the percentile is directly comparable across different test administrations and score scales.

Salary benchmarking: HR departments describe compensation as "targeting the 75th percentile for this role" — meaning the offered salary should exceed what 75% of comparable roles pay, a far more useful anchor than a raw dollar figure alone.

Network and server monitoring: engineers track "p95" or "p99" latency (the 95th or 99th percentile response time) rather than the average, because the average can hide a real problem — a fast average with a bad 99th percentile means most users have a good experience while a meaningful minority are badly underserved, exactly the kind of pattern a single mean would mask.

Common Mistakes

Confusing percentile with percentage, as covered above — a value can have a low percentage score and a high percentile rank simultaneously, and vice versa. Always be explicit about which one you're reporting.

Assuming percentiles are evenly spaced in value. Moving from the 50th to the 60th percentile doesn't cover the same raw-value distance as moving from the 90th to the 100th percentile, especially for skewed data — percentiles near the extremes of a distribution are typically much more spread out in raw terms than percentiles near the middle. This is exactly why "the top 1%" in income data is such an enormous range in dollar terms, while the middle 40-60th percentiles might span a comparatively narrow band.

Using the wrong reference population. A percentile is only meaningful relative to the specific comparison group it was calculated from. "90th percentile height" means something different depending on whether the reference group is all adults, adults of a specific sex, or a specific age band — always check which population a reported percentile is relative to before drawing conclusions from it.

A Second Worked Example: Percentile Rank With Ties

Problem: A class of 20 students takes a test. A particular student's score is higher than 15 other students' scores, and tied with 2 others. What's a reasonable percentile rank for that student?

A common convention handles ties by including half the tied group in the "below" count: percentile rank ≈ [(number strictly below) + 0.5 × (number tied)] / total × 100 = [15 + 0.5(2)] / 20 × 100 = 16/20 × 100 = 80th percentile. This "midpoint" treatment of ties is one of several accepted conventions — like quartile methods, different software may handle ties slightly differently, and this is a normal, expected source of small variation rather than an error.

Percentiles in Large Data Sets: Approximation Methods

Calculating an exact percentile requires sorting the entire data set, which becomes computationally expensive for very large data sets (millions or billions of values, common in web analytics or scientific data). In practice, approximation algorithms (like the t-digest or HdrHistogram methods used in many monitoring systems) are used to estimate percentiles — particularly the extreme ones like p99 or p99.9 — without needing to store or sort every single data point. These trade a small amount of precision for a large gain in speed and memory efficiency, and the approximation error is typically negligible for practical monitoring and reporting purposes.