📉Outlier Detector

Find Outliers Two Ways
3-Sigma and IQR, Side by Side

Upload a CSV and compare two outlier definitions per column, with adjustable thresholds and an interactive histogram and box plot. Every flagged value shows its row number nothing is removed automatically.

✓ 3-Sigma + IQR methods✓ Adjustable thresholds✓ Histogram + box plot✓ No upload to server✓ Free forever
📂

Drop your CSV file here

CSV files only your file never leaves your browser

Numeric columns need at least 4 values to be analyzed.

How to use

How to Find Outliers in a CSV File

This tool calculates both standard outlier definitions for every numeric column and lets you explore each one visually.

1
📂

Upload your CSV

Click or drag your CSV file onto the upload area. Your file is read locally nothing is sent to any server.

2
📋

Scan the overview

Every numeric column is listed with mean, median, std dev, quartiles and outlier counts from both methods at default thresholds.

3
🔬

Drill into a column

Column Detail shows a histogram and box plot. Switch between 3-Sigma and IQR, and adjust the threshold to see the shape change live.

4
📋

Investigate, not delete

Copy the flagged row numbers and check each one in your original file. Decide case by case whether to fix, keep, or remove.

Two definitions, one tool

Two Ways to Define an Outlier And Why They Disagree

The 3-sigma (Z-score) method assumes your data is roughly bell-shaped and asks "how many standard deviations from the mean is this value?" It is simple and widely taught, but the mean and standard deviation are themselves dragged around by extreme values one massive outlier can inflate the standard deviation enough that genuinely unusual values nearby no longer look extreme by comparison.

The IQR (Tukey) method instead looks at the median and the middle 50% of your data (between Q1 and Q3). Because the median barely moves even with extreme values present, this method is far more robust for skewed distributions revenue, transaction amounts, response times where most values cluster low and a long tail of larger values is completely normal. This tool runs both, side by side, with adjustable thresholds, so you can see exactly where and why they differ for your data.

📐

3-Sigma (Z-score)

mean ± threshold × std dev. Best for roughly normal data.

📦

IQR (Tukey)

Q1/Q3 ± multiplier × IQR. Robust to skew, the box-plot standard.

⚙️

Adjustable

2.0–3.5σ or 1.5×/3× IQR see the count change live.

🔍

Compared

Both counts shown per column in the overview table.

Reading the charts

How to Read the Histogram and Box Plot

The histogram groups all values into bins taller bars mean more values fall in that range. Bins shaded red fall entirely inside the current outlier zone for your selected method and threshold; purple bins are the normal range. The box plot shows the box from Q1 to Q3 (the middle 50% of values) with a line at the median, whiskers reaching to the most extreme non-flagged values, and individual red dots for every value beyond the whiskers hover any dot to see its row number and value.

Four common causes of outliers in business data

⌨️

Data Entry Errors

An extra zero, a misplaced decimal, or a copy-paste that duplicated a digit typing 100000 instead of 10000.

🖥️

System Defaults

A sensor or API returns -1, 0 or 9999 for a missing reading instead of leaving the field blank.

✅

Genuine Extremes

A single large order, a record day, a payment reversal real values that are the signal, not the noise.

🧪

Leftover Test Data

Staging records with placeholder amounts like 12345.67 or dates from 1970 or 2099, never cleaned before export.

FAQ

Frequently Asked Questions

What outlier detection methods does this tool use? +
Two, calculated for every numeric column and shown side by side: the 3-sigma (Z-score) rule, which flags values more than a chosen number of standard deviations from the mean, and the IQR (Tukey) method, which flags values below Q1 minus a multiplier times the interquartile range, or above Q3 plus that multiple. Both thresholds are adjustable.
What is the difference between the 3-sigma method and the IQR method? +
The 3-sigma method assumes a roughly normal, symmetric distribution and uses the mean and standard deviation, both of which extreme values themselves distort a single huge outlier can inflate the standard deviation enough to mask smaller but still unusual values. The IQR method is based on the median and quartiles, which barely move with a few extreme points, making it more robust for skewed data. The two methods often agree, but disagreement is usually a sign the data is skewed rather than normally distributed.
Can I adjust the outlier detection threshold? +
Yes. For 3-Sigma, choose 2.0, 2.5, 3.0 or 3.5 standard deviations lower values flag more points. For IQR, choose the standard 1.5x multiplier ("outliers") or 3x ("extreme outliers" / "far out" points). The histogram, box plot and outlier list update immediately.
How do I read the box plot and histogram? +
The box plot shows a box from Q1 to Q3 with a line at the median, whiskers to the most extreme non-flagged values, and dots for every outlier beyond the whiskers hover a dot for its row number and value. The histogram groups values into bins; bins entirely within the outlier zone for the current method and threshold are shaded red, normal bins are purple.
Should I remove outliers from my data? +
Not automatically. An outlier might be a data entry error to correct, or a genuine extreme value that is exactly the signal you care about. This tool has no one-click removal button by design it lists every flagged value with its row number and lets you copy those numbers to investigate each one in your original file.
What causes outliers in business data? +
Four common causes: data entry errors (an extra zero, a misplaced decimal), system or sensor defaults (-1, 0 or 9999 standing in for a missing reading), genuinely extreme real values (a single large order), and leftover test or staging data with unrealistic placeholder values never cleaned before export.
How is this different from the outlier detection in the Data Quality Checker? +
The Data Quality Checker runs a fixed 3-sigma check as one of four factors feeding an overall 0-100 score, shown as a simple table. This tool is a dedicated deep-dive: both 3-sigma and IQR side by side, adjustable thresholds, and an actual histogram and box plot per column so you can see the shape of your data. Use the Data Quality Checker for an overall score; use this tool to investigate one column’s distribution.
Is my CSV file uploaded to a server? +
No. Your file is read directly in your browser via the FileReader API. All statistics, binning and outlier calculations run locally nothing is transmitted anywhere, which makes this safe for confidential data.
Related tools

More Free Data Quality Tools