Free web tool
Detect outliers in a numeric CSV column using IQR or z-score and extract flagged rows.
A value is far from the rest—but unusual does not automatically mean wrong. Use IQR or z-score to create a shortlist of rows worth reviewing, not a list of rows to delete.
The IQR method uses Q1 and Q3, then flags values outside Q1−1.5×IQR and Q3+1.5×IQR. It is often a useful first pass when the distribution is not perfectly symmetric.
With very small samples, quartiles themselves are unstable. Five values and fifty thousand values should not be given the same statistical confidence, even if the same IQR rule is applied.
The z-score path measures distance from the mean in standard-deviation units. This implementation uses sample standard deviation and flags values at or beyond the threshold you choose.
If all usable values are identical, standard deviation is zero and the z-score path flags nothing. Lowering the threshold cannot create variation that is not present in the data.
Blank or non-numeric cells do not count as normal or outlier values. The numeric-row metric tells you how many rows actually participated in the calculation.
A million-dollar order in a dataset full of $10k orders may be a perfectly valid enterprise deal. Statistical rarity and data-quality failure are different things. That distinction is the reason the tool exports candidates instead of deleting anything.
The result is a review list, not the full CSV with an extra flag column. That makes it handy when your goal is to hand a short set of suspicious records to someone for inspection.