Detect Outliers in CSV Files Online
Detect statistical outliers in CSV files directly in your browser. Flag or remove rows where numeric values exceed a chosen number of standard deviations from the mean — no upload required.
How to detect Outliers in CSV files
- Drop your file onto the upload area. It is loaded into the in-browser engine and the first 200 rows are shown.
- Pick a threshold in standard deviations from the mean: 1.5σ, 2σ (the default), 2.5σ or 3σ.
- Pick an action: Flag outliers adds an is_outlier column, Keep outliers returns only the outlier rows, and Remove outliers returns the rest.
- Choose the columns to check. Every numeric column is selected to start with. Click a column to toggle it, or use All and None. A row counts as an outlier if any selected column is beyond the threshold.
- Click Detect Outliers. The count shows how many rows were flagged, kept or left. Review the result, then download it in the same format you uploaded.
Your file is processed locally in your browser and is never uploaded. The free limit is 50 MB per file; larger files work if your device has the memory for them.
Worked example
A food distributor logs one temperature reading from each of six freezer units. One unit has a failing compressor and reads far warmer than the rest.
Input (CSV)
unit,location,temp_c
FZ-01,Bay 1,-18.2
FZ-02,Bay 1,-18.6
FZ-03,Bay 2,-17.9
FZ-04,Bay 2,-18.4
FZ-05,Bay 3,-9.5
FZ-06,Bay 3,-18.1Settings
- Threshold: 2σ
- Action: Flag outliers
- Columns to check: temp_c (the only numeric column)
Result
| unit | location | temp_c | is_outlier |
|---|---|---|---|
| FZ-01 | Bay 1 | -18.2 | false |
| FZ-02 | Bay 1 | -18.6 | false |
| FZ-03 | Bay 2 | -17.9 | false |
| FZ-04 | Bay 2 | -18.4 | false |
| FZ-05 | Bay 3 | -9.5 | true |
| FZ-06 | Bay 3 | -18.1 | false |
The mean is −16.78 °C and the sample standard deviation is 3.58 (both rounded), so 2σ is 7.15 degrees. FZ-05 is 7.28 above the mean, a z-score of 2.04, so it is flagged. The faulty reading also inflates the mean and the deviation it is judged against. At 2.5σ nothing would be flagged in a file this small.
Working with CSV files
Every column detected as numeric when the CSV loads starts selected. That includes columns such as store_id or zip that are all digits but are not measurements. A row can be flagged because its ID is unusually large, so deselect those before you run. Numbers written with units or separators, such as "12 kg" or "1,250", load as text and cannot be checked until you clean them. Numbers in scientific notation, such as 1.2e-5, are recognised as numbers.
Empty fields load as NULL. A blank value is never an outlier, so in Flag mode it gets false unless another checked column flags the row. Remove outliers keeps blank rows, and Keep outliers leaves them out. In the downloaded CSV, is_outlier is the last column and is written as true or false.
One stray text value changes what gets checked. A single "n/a" or "-" in a readings column makes the whole column text when the CSV loads, so it drops out of the numeric list and no rows are tested against it. Clear or replace those markers first. Also keep in mind that the rest of the file is rewritten from the loaded table, so numbers appear in their parsed form: 12.50 is written as 12.5 and 007 as 7 if that column was read as numeric.
Frequently Asked Questions
Does Remove outliers drop CSV rows with blank values?
No. A blank value is treated as not an outlier, so the row stays unless another checked column flags it.
Why was a row flagged because of an ID column?
Every numeric CSV column is checked by default, including IDs and codes. Deselect columns that are not measurements before you click Detect Outliers.
Which method does the tool use?
A z-score test. A value is an outlier when it is more than the chosen number of sample standard deviations away from its column's mean. The mean and deviation come from the whole file. IQR and median-based methods are not offered.
Why are no outliers found at 3σ in my small file?
With n rows, no value can be more than (n − 1) / √n sample standard deviations from the mean. That means 2σ needs at least 6 rows, 2.5σ needs 9, and 3σ needs 11 before anything can be flagged at all. Extreme values also inflate the deviation they are measured against.
How are blank values handled?
They are ignored when the mean and deviation are computed, and a blank is never an outlier. In Flag mode the row gets false unless another column flags it. Remove outliers keeps the row, and Keep outliers leaves it out.
Related Tools
Filter CSV Files Online
Filter rows in CSV files by column value, directly in your browser. Your data stays on your device.
Aggregate CSV Files Online
Group and aggregate CSV files by any column directly in your browser. Calculate sum, average, min, max, and count for any numeric column — no upload required.
Fill Empty Values in CSV Files Online
Fill empty and null values in CSV files with a custom replacement value, directly in your browser.
Detect Outliers in Excel Files Online
Detect statistical outliers in Excel files directly in your browser. Flag or remove rows where numeric values exceed a chosen number of standard deviations from the mean — no upload required.
Detect Outliers in Parquet Files Online
Detect statistical outliers in Parquet files directly in your browser. Flag or remove rows where numeric values exceed a chosen number of standard deviations from the mean — no upload required.
Detect Outliers in JSON Files Online
Detect statistical outliers in JSON files directly in your browser. Flag or remove rows where numeric values exceed a chosen number of standard deviations from the mean — no upload required.
CSV Viewer Online
View and inspect CSV files directly in your browser. Browse rows, check column names and data types — no upload required, your data stays on your device.
Convert CSV to Parquet Online
Convert CSV files to Parquet format directly in your browser. No upload required — your data never leaves your device.