Free · No signup · Runs in your browser

Detect & removeCSV outliers

Pick a numeric column, choose a method, and flag, isolate, or remove the values that don't fit. 100% client-side — nothing is uploaded.

01 · How it works

Three steps, then done.

Outlier detection turns a column of numbers into a clean signal by separating the typical values from the extremes. Load your CSV, choose the column to analyze, and pick how aggressively to detect.

1

Load your CSV and pick a column

Drop in a file or paste your data. Choose the numeric column you want to scan — prices, durations, scores, sensor readings, anything measurable. Non-numeric and blank cells are ignored when computing the statistics.

2

Choose a detection method

Use 2 or 3 standard deviations (z-score) for roughly normal data, or the IQR 1.5x rule for skewed distributions where a few extremes would otherwise distort the mean. Three std dev and IQR are more conservative; two std dev flags more rows.

3

Flag, isolate, or remove

Add an is_outlier column to mark every row, keep only the outliers to inspect them, or remove them to get a clean dataset. Export the result as CSV — all computed locally, with a count of how many outliers were found.

02 · Why ours

Why detect outliers in your CSVbefore you analyze

A handful of extreme values can wreck an average, blow out a chart axis, or hide the real shape of your data. Catching them early keeps every downstream calculation honest.

  • 01

    Clean averages and charts

    One mistyped order of 99999 or a stuck sensor reading drags the mean and stretches axes. Removing or flagging outliers gives you summaries and visualizations that reflect the typical case.

  • 02

    Catch data-entry errors

    Outliers are often bugs, not signal — an extra zero, a swapped unit, a placeholder value. Isolating the extremes is the fastest way to spot what needs fixing at the source.

  • 03

    Method that fits your data

    Normal-ish data suits z-score thresholds; skewed or heavy-tailed data suits the IQR fences, which don't assume symmetry. Switching methods takes one click, so you can compare what each flags.

  • 04

    Private by design

    Detection runs entirely in your browser with no upload, no account, and no server. Financial, medical, or proprietary numbers stay on your machine the whole time.

"Outliers are often bugs, not signal — an extra zero or a stuck sensor. Flagging the extremes is the fastest way to find what to fix."
On why outlier detection matters
03 · FAQ

outliers questions.

What's the difference between the z-score and IQR methods?
Z-score (standard deviation) flags values that sit more than 2 or 3 std devs from the mean — a good fit when your data is roughly bell-shaped. The IQR 1.5x rule flags values below Q1 − 1.5×IQR or above Q3 + 1.5×IQR, using quartiles instead of the mean, so it's more robust for skewed data where a few extremes would otherwise inflate the standard deviation.
Two std dev is more sensitive and flags more rows (about 5% of normally distributed data), useful when you want to catch anything unusual. Three std dev is conservative and flags only the most extreme ~0.3%, useful when you only want to remove clear errors. Start with 2 and tighten to 3 if it flags too much.
They're skipped when computing the mean, standard deviation, or quartiles, so they never count as outliers and never distort the thresholds. Their rows are still kept in the output (and marked is_outlier=false when flagging).
When you choose the flag action, every row gets an is_outlier value of true or false based on the selected column and method. The original columns are left untouched, so you can sort or filter on the flag afterward. If your CSV already has an is_outlier column, the tool adds a numbered variant to avoid a collision.
No. The entire computation runs client-side in your browser using JavaScript. No file is sent to a server, there's no account, and nothing is stored — closing the tab discards everything.
Related tools
CSV Profiler
Get per-column stats — min, max, mean, distribution — to see where outliers might be hiding before you filter.
Open →
CSV Sort & Filter
Sort by the flagged column or filter rows by value to inspect the extremes you just detected.
Open →
CSV Validator
Check your data against type and range rules to catch the data-entry errors that produce outliers.
Open →
CSV to Pivot
Summarize and aggregate your cleaned data once the distorting outliers are removed.
Open →
CSV Dedupe
Remove duplicate rows — another common cleanup step before statistical analysis.
Open →