What's the difference between p50, median, and Q2?

All three are the same value — the middle of a sorted dataset. 'Median' is the everyday term; 'p50' is the percentile name; 'Q2' is the second quartile. Used interchangeably.

Why does my percentile differ from another tool's?

There are at least 9 standardised percentile-calculation methods (R's quantile() supports 9 types). The tool uses linear interpolation between order statistics (R type 7, the Excel and NumPy default). Small differences for small datasets, identical for large ones.

What's an outlier per the Tukey fence rule?

Anything more than 1.5 × IQR below Q1 or above Q3. It's a conservative, distribution-free outlier definition that doesn't assume normality. Useful for boxplot whiskers and flagging anomalies in long-tailed data.

quartiles · iqr · outliers · linear interp · local

Compute any percentile

Drop in a dataset and get the percentile of your choice plus quartiles, IQR, and Tukey outlier fences. All math runs locally.

Percentile Calculator

values

P75

target

IQR

9.5

spread

Outliers

tukey

input · dataset

DatasetTarget P: 75

output · stats

P75

11/15 below (73.3%)

6.5

IQR

9.5

Outliers (Tukey)

fence: [-7.75, 30.25]

Quick Start

How percentiles work

Sort the data, multiply (p/100) by (n−1) to get an index, linearly interpolate between neighbors.

01
Paste numbers
Any separator works — commas, spaces, newlines, semicolons.
Inputs are treated as UTF-8 throughout the pipeline and validated client-side as you type — nothing dispatches until the value is well-formed.
inpututf-8client-sidevalidated
n · text15 · set
```
3, 7, 8, …
```
loaded · parsedclient-side
02
Slide target
Pick any percentile 0–100.
Each toggle re-computes through a memoised hook, so flipping options never resets your input and the UI stays snappy on large inputs.
poptionslivememoisedtoggles
p · memoised75 · live
```
P75
```
options set · memoisedlive
03
Read quartiles
Q1, Q2 (median), Q3, IQR shown alongside.
Randomness — where used — comes from crypto.getRandomValues, never Math.random. Heavy work sits in useMemo.
quartcomputeinstantlocaldeterministic
iqr · instantspread · computed
```
q3 − q1
```
computed · instantinstant
04
Spot outliers
Tukey fences: Q1 − 1.5·IQR, Q3 + 1.5·IQR.
Copy writes through navigator.clipboard and downloads use a Blob — zero network traffic. Same input plus same options always yield the same bytes.
outliercopydownloadzero-networkreproducible
ok · copiedready · copied
```
99 is out
```
shipped · ready to shipsuccess

Pro tips

Pro tips for everyday utilities

Reach for crypto.getRandomValues

Cryptographically random for IDs, tokens, salts. Math.random is for animation and games, not security.

Sort IDs lexicographically

UUID v4 has no time ordering. Use ULID or UUID v7 if you need sorted IDs.

Mock data with seeds

Faker-style libraries can be seeded so tests are reproducible. Embed the seed in the fixture file.

Reference cards beat memory

Cheatsheets for Vim, Git, Docker, Linux signals are faster than man pages when you’re mid-flow.

Browser APIs vary

Touch events, gamepads, and clipboard differ across vendors. Feature-detect before use.

Persist in the right place

localStorage for user preferences, sessionStorage for transient state, IndexedDB for structured data.

Use Cases

When percentiles matter

Latency, growth, grading.

Latency budgets

p95 / p99 for SRE dashboards. Random values come from crypto.getRandomValues — cryptographically secure for tokens, salts, and session IDs. Math.random is deterministic enough for animations but predictable enough to be a security bug.

CSPRNG-backed

Grade curves

Where does a score land in the class? Pair with hashids, ULID, or UUID v7 when you need both random and sortable — useful for primary keys, public-facing IDs, and event log ordering without a sequence column.

reproducible

Income brackets

Median + IQR for salary research. Output is deterministic given the same seed, so fixtures and tests stay reproducible across CI runs and contributor machines. Embed the seed in the fixture file for full replay.

snappy

AB test

Compare conversion percentiles. Heavy work runs through useMemo so the UI stays snappy on large inputs. Same input + same options yields identical bytes — the kind of stability snapshot tests rely on.

feature-detected

Lab values

Reference ranges in clinical data. Storage tools persist to the right backend: localStorage for user preferences, sessionStorage for transient state, IndexedDB for structured data. Pick the right shelf and quota worries disappear.

deterministic

Outlier removal

Drop Tukey outliers before mean/SD. Browser feature detection: clipboard, fullscreen, gamepad, and Web Bluetooth APIs gracefully fall back when unavailable. No surprise white screens on Safari or older Firefox.

production-ready

FAQ

Frequently asked questions

Which percentile method?

Linear interpolation between order statistics — same as numpy's default 'linear'.

What's IQR?

Interquartile range, Q3 − Q1. Captures the middle 50% spread, robust to outliers.

Tukey fences?

Values outside [Q1 − 1.5·IQR, Q3 + 1.5·IQR] are flagged as outliers — a robust quick test.

How many points can I paste?

Tens of thousands work fine in the browser. Larger sets may slow the sort.

Is anything sent to a server?

No. All stats are computed in your browser.

Suggest a tool

Something missing?

Tell us what you need. We ship the top requests every two weeks.

47 in queue·8 shipped this month

Request queued. Thanks!