Your data
Paste a table straight from Excel or Google Sheets, or drop a CSV file. Nothing is uploaded anywhere — the whole calculation runs inside this browser tab.
CSV, TSV or plain text. Files saved in windows-1251 are detected and re-read automatically, so Cyrillic headers do not turn into garbage.
Parsing options
Columns
Result
Save and share
The drawing library is downloaded only when you press the button, and only on this page.
This page has no server side at all. The table you paste is parsed by JavaScript inside your own browser, the chart is drawn on a canvas element on your machine, and the export file is assembled locally. Nothing is uploaded, stored or written to any log. You can disconnect from the network after the page has loaded and everything will still work — which is the simplest way to verify the claim yourself.
Facts and limits of this method
| IQR rule | Outside Q1 − 1.5·IQR or Q3 + 1.5·IQR |
|---|---|
| Extreme threshold | Outside 3·IQR — flagged separately as extreme |
| Z-score rule | Absolute z above 3, computed with the sample standard deviation |
| Robustness | IQR is unaffected by the outliers themselves; the z-score is not |
| Output per row | Value, z-score and which rule or rules flagged it |
| Minimum rows | 4 |
When it misleads you
- Neither rule knows whether a value is wrong. A genuine record-breaking sale and a misplaced decimal point look identical to both. Detection is where the investigation starts, not where it ends.
- The z-score rule is self-defeating on small samples. With 20 rows the maximum possible z-score is about 4.2, and one severe outlier inflates the standard deviation enough to hide itself — a phenomenon called masking.
- The 1.5 multiplier flags roughly 0.7% of observations even in perfectly normal data. On a 10,000-row column expect around seventy points beyond the fences with nothing wrong at all.
- Both rules assume a single population. If your column mixes retail and wholesale invoices, the wholesale rows will be flagged as outliers when the real problem is that two groups were combined.
How it is calculated
The IQR rule builds fences at Q1 − 1.5·IQR and Q3 + 1.5·IQR. Because quartiles are positional, the fences do not move when an outlier gets more extreme — a value ten times too large is caught just as reliably as one twice too large.
The z-score rule measures each value in standard deviations from the mean and flags anything beyond three. Both of those ingredients are themselves distorted by outliers, which is exactly the weakness the IQR rule does not share, and the reason both are run rather than one.
A third threshold at 3·IQR marks extreme outliers separately. The distinction is practical: a value just past the fence often turns out to be legitimate, while one past three interquartile ranges is usually either an error or a different kind of event entirely.
Every flagged row is listed with its position in the file so you can go back and look at it. The export contains all rows with their z-scores and a flag column, so the decision about what to do with them stays yours rather than being applied silently.
Questions and answers
Should I delete the outliers?
Not automatically. Deleting values because they are inconvenient is how analyses go wrong. Investigate first: an error gets corrected, a genuine extreme event gets reported, and a mixed population gets split into separate groups.
The two rules disagree. Which do I trust?
Prefer the IQR rule when they conflict, because it is not distorted by the outliers themselves. The z-score rule is more useful as a measure of how extreme a value is than as a detector.
Why did nothing get flagged when I can see an obvious outlier?
Probably masking: with few rows, one extreme value inflates the standard deviation enough to keep its own z-score below 3. Check whether the IQR rule caught it — it usually does.
What counts as too many outliers?
More than about 2–3% of your rows suggests the rules are not detecting anomalies but describing a skewed distribution. Check normality; on skewed data the fences flag a whole tail.
Is my data uploaded?
No. Both rules run entirely in your browser.