Percentile Meaning: Definition, Formula & How to Calculate

The percentile meaning in statistics is this: the pth percentile is the value in a sorted data set at or below which p percent of the observations fall. A score at the 85th percentile is higher than 85 percent of all values in the data set — that single relationship is the definition of percentile in full. The same four-step formula that defines a percentile also produces the median, the quartiles, and the deciles, and it is exactly what runs behind the calculator below, Excel’s PERCENTILE.INC, and R’s default quantile() function. This guide covers the meaning of percentile precisely, the calculation steps, a fully worked numeric example, and the everyday places percentiles show up, from test scores to growth charts to income brackets.

Key takeaways

Point	Details
Percentile meaning, in one line	Pₚ is the value at or below which p percent of a sorted data set falls — a rank within the distribution, not a raw score.
The formula is four repeatable steps	Sort the data, compute a virtual position, split it into a floor and a fraction, then interpolate between the two nearest values.
A percentile is not a percentage	Scoring at the 70th percentile means you beat 70 percent of other scores; it says nothing about how many questions you answered correctly.
The median and quartiles are special cases	The 50th percentile is the median, and the 25th/75th percentiles are Q1/Q3 — same formula, different value of p.
Tools can disagree slightly	R, Excel, and NumPy don't all default to the same interpolation method, so a small data set can return a different percentile depending on the software.

Quick Checklist: How to Calculate a Percentile

Quick checklist: calculating a percentile

• Sort the data first Percentile positions only make sense on values arranged from smallest to largest.
• Compute the virtual position position = (p / 100) × (n − 1), using the target percentile p and the count n.
• Split the position into floor and fraction j = floor(position) is the lower index, and f = position − j is the fraction between neighbors.
• Interpolate between the two nearest values If f = 0, the answer is the value at index j; otherwise, slide f of the way from the value at j to the value at j + 1.
• Note which method you used Excel, R, and NumPy offer more than one percentile algorithm, and small data sets can show a visibly different result between them.

Run through these steps in order every time — skipping or reordering any one of them is a common source of a wrong answer.

Calculate a Percentile

Paste your own list of numbers into the calculator below, set the target percentile, and it returns that value along with Q1, the median, Q3, and the IQR in one pass.

Calculator

Percentile Calculator

Values

Percentile

Compute

Enter values and compute the result.

What Is the Definition of a Percentile?

A percentile is a threshold value within a sorted distribution. The pth percentile, written Pₚ, is the point such that p percent of the data falls at or below it. The NIST/SEMATECH e-Handbook of Statistical Methods states the definition formally: the pth percentile is “a value, Y₍ₚ₎, such that at most (100p)% of the measurements are less than this value and at most 100(1−p)% are greater” — and it adds that “the 50th percentile is called the median.” That definition of percentile holds across every application: if the 70th percentile of exam scores in a class is 82, then 70 percent of students scored 82 or lower.

Note what the percentile definition does not say: it does not say you answered 70 percent of questions correctly. A percentile is a rank, not a raw proportion, and that distinction is worth locking in early because it trips up most beginners.

Percentiles Versus Percentages

A percentage score of 70 means you answered 70 of 100 questions correctly. A percentile rank of 70 means you scored higher than 70 percent of test-takers. On an easy exam where most students scored well, answering 70 percent of questions correctly might put you below the median — at the 40th percentile. On a hard exam, that same raw percentage might place you in the top quarter. The percentile captures competitive position; the percentage captures the absolute score, and the two numbers move independently of each other.

Why Raw Scores Need Context

A raw number becomes meaningful only once you know its distribution. The phrase “she scored 640” says nothing about performance unless you also know the score range and how everyone else performed. The phrase “she scored at the 94th percentile” is immediately interpretable: only 6 percent of test-takers did better, regardless of what the underlying scale looked like.

Related Terms Built on the Same Definition

Several familiar statistics are just percentiles at a specific value of p:

• Median — the 50th percentile, the value that splits the data set exactly in half.
• Quartile — the 25th (Q1), 50th (Q2), and 75th (Q3) percentiles, which divide the data into four equal parts.
• Decile — every 10th percentile; the 3rd decile, for instance, is the 30th percentile.
• Quintile — every 20th percentile; the 1st quintile is the 20th percentile.

Once you can define percentile and apply the formula below, you already know how to find any of these — each one is the same calculation at a different p.

How Do You Calculate a Percentile Step by Step?

The most widely used method for how to calculate percentile values — and the one behind R’s default quantile(), Excel’s PERCENTILE.INC, and the calculator above — is linear interpolation, formally Hyndman-Fan Type 7. It works on any sorted numeric data set and answers how to calculate percentiles for any p from 0 to 100.

Step 1: Sort the Data in Ascending Order

Arrange all n values from smallest to largest. Every later step depends on this order being correct.

Step 2: Compute the Virtual Position

position = (p / 100) × (n − 1)

Here p is the target percentile (0 to 100) and n is the count of values. The n − 1 scales the position to a zero-based index into the sorted array.

Step 3: Split the Position Into a Floor and a Fraction

j = floor(position)    ← lower index
k = ceil(position)     ← upper index
f = position − j       ← fraction, between 0 (inclusive) and 1 (exclusive)

Step 4: Interpolate Between the Two Neighboring Values

If f = 0 (the position lands exactly on an integer), the percentile is sortedData[j]. Otherwise:

Pₚ = sortedData[j] + f × (sortedData[k] − sortedData[j])

This slides f of the way from the lower neighbor to the upper one. When f = 0.5, the result lands exactly halfway between them.

For a data set of 10 values, there are not 100 distinct positions — only 10 actual data points. The interpolation formula creates a continuous mapping from percentile to value, so P₄₅ gives a well-defined answer even when no data point sits precisely there. The open-access textbook OpenStax, Introductory Statistics — §2.3 Measures of the Location of the Data walks through this same procedure with additional practice data sets.

A Worked Example: Finding the 90th Percentile

Data set of ten exam scores, already sorted:

5, 10, 15, 20, 25, 30, 35, 40, 45, 50

n = 10. Find P₉₀.

Step 2 — virtual position:

position = (90 / 100) × (10 − 1) = 0.90 × 9 = 8.1

Step 3 — floor and fraction:

j = floor(8.1) = 8
k = ceil(8.1)  = 9
f = 8.1 − 8    = 0.1

(Index 8 in the sorted array is the 9th value, 45; index 9 is the 10th value, 50.)

Step 4 — interpolate:

P₉₀ = sortedData[8] + 0.1 × (sortedData[9] − sortedData[8])
    = 45 + 0.1 × (50 − 45)
    = 45 + 0.5
    = 45.5

The 90th percentile is 45.5. Ninety percent of the scores — 9 out of 10 — fall at or below 45.5; the only score above that threshold is 50. The result sits exactly 10 percent of the way from 45 to 50, matching the 0.1 fraction from Step 3.

Enter the same ten values into the calculator above and set the percentile to 90 to confirm the result without redoing the arithmetic by hand.

What Percentile Rank Does a Value You Already Have Correspond To?

Everything above answers “what value sits at the pth percentile?” The reverse question — “what percentile rank does this value I already have correspond to?” — can give more than one defensible answer, because more than one counting convention is in common use, and they don’t all agree.

Take the value 45 in the same ten-score data set used throughout this guide. Four conventions, four different answers:

• At or below (the definition this page opened with): 9 of the 10 scores — 5, 10, 15, 20, 25, 30, 35, 40, and 45 itself — are at or below 45, so by a plain count, 45 sits at the 90th percentile rank.
• Strictly below: only 8 of the 10 scores are strictly less than 45 (45 itself isn’t counted), which gives an 80th percentile rank instead.
• The Type-7 formula, run in reverse: solving position = (p / 100) × (n − 1) for p, using 45’s zero-based index of 8, gives p = (8 / 9) × 100 ≈ 88.9. This is the direction consistent with the forward formula used everywhere else on this page, though neither this section’s formula nor the calculator above actually computes it for you — both only go from p to a value, not from a value back to p.
• NIST’s rank-based convention: the NIST/SEMATECH e-Handbook defines the percentile corresponding to the i-th ranked point as i / (n + 1); for 45 at rank i = 9 of n = 10, that’s (9 / 11) × 100 ≈ 81.8.

At or below:        9 / 10 × 100  = 90
Strictly below:      8 / 10 × 100  = 80
Type-7, reversed:    (8 / 9) × 100 ≈ 88.9
NIST, i / (n + 1):   (9 / 11) × 100 ≈ 81.8

None of these four numbers is wrong — each follows a legitimate, documented definition of what a percentile rank counts. By the “at or below” definition this page opened with, 45 genuinely is at the 90th percentile rank of this data set; that is not a mistake, it’s just a different (and simpler) question than the Type-7 interpolation this page’s formula and calculator use to go the other direction. When a specific percentile-rank figure needs to be reproducible, state which convention produced it.

How Do Percentiles Relate to Quartiles, Deciles, and the Median?

Quartiles, deciles, and the median are not separate statistics — they are percentiles at specific values of p, computed with the identical four-step formula above. The median is the 50th percentile; Q1 and Q3 are the 25th and 75th percentiles; a decile is any multiple of the 10th percentile.

Running the same ten-score data set through Steps 2–4 at p = 25, 50, and 75 gives Q1 = 16.25, the median = 27.5, and Q3 = 38.75 — the same numbers a direct quartile calculation would produce, because it’s the same formula applied three more times. The interquartile range guide covers what happens once you subtract Q1 from Q3 (the IQR), why that difference resists outliers, and how it anchors the 1.5×IQR rule used in the outlier detection guide; both go well beyond what a single percentile value tells you on its own.

Where Are Percentiles Used in Real Life?

The percentile meaning translates across many fields because ranking a value within its distribution is a near-universal need.

Standardized testing. The SAT, GRE, and most professional licensing exams report scores alongside a percentile rank. The College Board’s guide to understanding SAT Suite scores notes that a score report includes “percentiles that compare your scores to other students who took the same test” — admissions offices lean on that ranking because raw scores from different test forms and years aren’t directly comparable.

Pediatric growth charts. The CDC’s clinical growth charts plot height, weight, and head circumference against percentile curves calculated from national reference populations by age and sex. A child at the 60th percentile for weight is heavier than 60 percent of children the same age; tracking whether that percentile stays stable across visits is more informative than any single raw measurement.

Income and wealth distribution. Economic reporting regularly cites the “median income” (the 50th percentile) and thresholds like “the 90th-percentile earner” or “the top one percent” (above the 99th percentile). These labels convert a complex income distribution into a comparable, single number across regions and time periods.

Manufacturing process capability. When process data don’t follow a normal distribution, NIST/SEMATECH’s process capability guidance replaces the usual ±3-standard-deviation spread with two percentiles of the actual data instead — the 99.865th percentile minus the 0.135th percentile — which the handbook says “is justified to mimic the coverage of ±3 standard deviations for the normal distribution.” That percentile-based spread then feeds directly into nonnormal capability indices such as Ĉₙₚₖ and Ĉₙₚₘ, in place of the standard-deviation-based versions used for normal data.

Financial risk management. Value at Risk (VaR) is defined directly as a percentile of a loss distribution over a set holding period: a 99th-percentile VaR estimate means losses are expected to exceed that figure only 1 percent of the time. Risk teams and regulators specify both the percentile and the holding period they’re using, because a one-day 99th-percentile VaR and a ten-day 99th-percentile VaR describe different things, and the two numbers aren’t interchangeable.

How Do You Calculate Percentiles in Excel, R, and Python?

Each tool implements the percentile formula with its own defaults and quirks, and they don’t all agree on small data sets.

Excel:  =PERCENTILE.INC(A1:A10, 0.9)
        → uses the same linear-interpolation method as this page's formula

R:      scores = c(5, 10, 15, 20, 25, 30, 35, 40, 45, 50)
        quantile(scores, 0.9)              # type = 7 by default → 45.5

Python: import numpy as np
        scores = [5, 10, 15, 20, 25, 30, 35, 40, 45, 50]
        np.percentile(scores, 90)          # method='linear' by default → 45.5

Microsoft’s PERCENTILE.INC documentation confirms the function returns the k-th percentile “inclusive” of 0 and 1, matching the linear-interpolation formula used throughout this guide. R’s quantile() documentation lists nine distinct sample-quantile algorithms, selected with its type argument, and defaults to type = 7 — the same method. NumPy’s percentile() function offers a matching set of interpolation methods under its method parameter, with linear as the default, again producing the same 45.5 on this data set.

Excel also ships a second function, PERCENTILE.EXC, which Microsoft’s documentation states takes a k strictly between 0 and 1, interpolates when k isn’t an exact multiple of 1/(n + 1), and returns a #NUM! error “if it cannot interpolate for the percentile… specified.” In practice that means k only resolves to a value when 1/(n + 1) ≤ k ≤ n/(n + 1) — at the 90th percentile (k = 0.9), that requires n ≥ 9, so the ten-score data set used throughout this guide just clears the bar, but a five-value data set would return #NUM! at that same percentile. The same n ≥ 9 floor applies at the 10th percentile (k = 0.1), since the valid range is symmetric around the middle of the data set. PERCENTILE.EXC does not generally return the same value as PERCENTILE.INC on the same input.

What Mistakes Do People Make When Interpreting Percentiles?

• Confusing percentile with percent correct. A score at the 75th percentile does not mean you answered 75 percent of questions correctly — it means you scored higher than 75 percent of other test-takers. You could answer 60 percent of questions correctly and still land at the 90th percentile on a genuinely hard exam.
• Assuming the 50th percentile equals the mean. The 50th percentile is the median. In a symmetric distribution, the median and the mean coincide; in a right-skewed distribution — income is the classic case — the median sits well below the mean, because a small number of very high values pull the mean upward.
• Expecting every tool to agree. As the Excel/R/Python comparison above shows, the interpolation method matters, and PERCENTILE.INC and PERCENTILE.EXC are not interchangeable. Always state which method produced a reported percentile when the exact value matters.
• Treating percentile gaps as uniform in value. The raw-value gap between the 10th and 20th percentile is not necessarily equal to the gap between the 80th and 90th percentile, even though both cover ten percentile points. In a right-skewed distribution, the gap between the 90th and 99th percentiles can be many times larger than the gap between the 1st and 10th — percentile points measure rank spacing, not value spacing.
• Reporting a percentile with no sample size. A 90th-percentile value computed from 10 observations and the same percentile computed from 10,000 observations carry very different amounts of statistical certainty, even when the number itself looks identical. With only 10 observations, each data point effectively controls 10 percentage points of the distribution, so a single unusual measurement can shift an extreme percentile noticeably; with 10,000 observations, the same percentile is far more stable.
• Assuming there’s one universally correct interpolation method. The NIST/SEMATECH handbook itself says no single accepted way exists to interpolate between data points for a percentile that doesn’t land exactly on an observation, and it points to Hyndman and Fan’s 1996 review of nine different methods used in practice. Picking a method is a documented choice, not a universal constant.
• Reporting a percentile rank without naming the convention. As the percentile-rank section above shows, the same value can correspond to a 90th, 88.9th, 81.8th, or 80th percentile rank depending on whether you count at-or-below, strictly-below, the reversed Type-7 formula, or NIST’s i/(n + 1) rule. State which one you used whenever the exact figure matters.

Statohub’s Take on Percentiles

Statohub’s position is that the percentile meaning is one of the most underused tools for communicating a number clearly. A raw figure — a score, a measurement, a dollar amount — says little on its own; a percentile instantly tells a reader where that figure sits relative to everyone else. The trade-off is that a percentile hides the underlying scale, so report both the raw value and the percentile together whenever the audience might need to judge the absolute size of a gap, not just its rank.

Pick one interpolation method and name it whenever a percentile value needs to be reproducible — this guide uses linear interpolation throughout because it’s the shared default across R, Excel’s PERCENTILE.INC, and NumPy, which keeps a hand-computed answer, a spreadsheet formula, and a script all landing on the same number. Switching methods silently, or comparing two percentiles computed with different ones, is where most disagreements about “the right answer” actually come from.

Put Percentiles to Work With Statohub’s Tools

The calculator above handles any percentile, plus Q1, the median, and Q3, for a list you paste in directly. To round out the surrounding statistics, the standard deviation guide covers a spread measure that uses every value instead of a single rank, and the interquartile range guide shows what the 25th and 75th percentiles do once you subtract one from the other. For the data points that a percentile calculation alone can miss, the outlier detection guide explains how the 25th and 75th percentiles also set the standard fences for flagging extreme values. The Descriptive Statistics hub ties mean, median, mode, range, and percentiles together in one place, and the full calculators hub lists every tool on the site.

Sources

Sources

1. NIST/SEMATECH e-Handbook of Statistical Methods — 7.2.6.2. Percentiles NIST
2. OpenStax — 2.3 Measures of the Location of the Data, Introductory Statistics OpenStax
3. Microsoft Support — PERCENTILE.INC function Microsoft
4. Microsoft Support — PERCENTILE.EXC function Microsoft
5. NumPy — numpy.percentile NumPy
6. R Documentation — quantile (Sample Quantiles) The R Project
7. CDC — Clinical Growth Charts CDC
8. NIST/SEMATECH e-Handbook of Statistical Methods — 6.1.6. What is Process Capability? NIST
9. College Board — What Do My SAT Suite Scores Mean? College Board

FAQ

Frequently asked questions

How Do You Calculate a Percentile Step by Step?

Sort the data in ascending order, then compute a virtual position with position = (p / 100) × (n − 1). Split that position into an integer floor (j) and a fractional remainder (f). If f equals 0, the percentile is the value at index j; otherwise, interpolate by adding f × (the value at j + 1 minus the value at j) to the value at j. This procedure works for any percentile p from 0 to 100, and it is the method behind Excel, R, and NumPy's default percentile functions.

What Is the 50th Percentile?

The 50th percentile is the median — the value that splits a sorted data set into two equal halves. For the ten-score example in this guide, [5, 10, 15, 20, 25, 30, 35, 40, 45, 50], the 50th percentile is 27.5, found by interpolating halfway between the 5th value (25) and the 6th value (30). In a symmetric distribution, the median equals the mean; in a skewed distribution, the two diverge.

What Does the 90th Percentile Mean?

The 90th percentile is the value at or below which 90 percent of a data set falls, leaving only 10 percent above it. In the worked example on this page, P₉₀ equals 45.5 for the data set [5, 10, 15, 20, 25, 30, 35, 40, 45, 50]. A person at the 90th percentile of any measured characteristic has a higher value than 90 percent of the comparison group.

What Is the Difference Between a Percentile and a Percentage?

A percentage is a proportion of a whole — answering 70 of 100 questions correctly gives a 70 percent score. A percentile is a rank within a distribution — scoring at the 70th percentile means outperforming 70 percent of other test-takers. The two numbers move independently: a 70 percent raw score could land at the 40th percentile on an easy test, or the 95th percentile on a hard one.

Why Do Excel, R, and Python Sometimes Return Different Percentiles for the Same Data?

Each tool offers more than one percentile algorithm, and the defaults don't always match. Excel's PERCENTILE.INC, R's default quantile() (type 7), and NumPy's default percentile() (method='linear') agree with each other on most data sets, but Excel's separate PERCENTILE.EXC function uses a different formula and can return a visibly different value, especially on a small sample.

Recommended

• How to Find the Mean (Average): Formula & Steps
• Z-Score Calculator
• How to Find the Range in Statistics
• Variance: Definition, Formula & vs Standard Deviation