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
| Min–max | (value − min) ÷ (max − min); result lands in 0…1 |
|---|---|
| Z-score | (value − mean) ÷ standard deviation; result centred on 0 |
| Robust | (value − median) ÷ IQR; centred on 0 and scaled by the middle half |
| Outlier sensitivity | Min–max is worst affected, z-score second, robust barely at all |
| Reversibility | All three are reversible if you keep the parameters |
| Minimum rows | 2, and the spread must be non-zero |
When it misleads you
- Min–max scaling is defined by the two most extreme observations, which makes it the most fragile of the three. One wrong maximum compresses every other value into the bottom of the range, and adding a new record beyond the old maximum changes every previously computed value.
- Normalisation never changes the shape of a distribution. A skewed column stays exactly as skewed after any of these transformations; only the labels on the axis move. If you need to change the shape, that is a logarithm, not a normalisation.
- Scaling parameters must come from the training data and be reused, not recomputed. Normalising a test set on its own statistics is a classic and quiet source of leakage in machine learning.
- The robust method fails when the interquartile range is zero, which happens when more than half the values are identical. In that case the column is closer to categorical than to continuous.
How it is calculated
Min–max maps the smallest value to 0 and the largest to 1, spreading everything else proportionally between them. It is the right choice when a bounded range matters — for progress bars, colour scales and neural network inputs — and the wrong one whenever outliers are possible.
Z-score normalisation subtracts the mean and divides by the standard deviation, producing a column centred on zero with a standard deviation of one. It has no bounds, so extreme values stay visibly extreme instead of being squeezed against a ceiling, which is usually what you want for statistical work.
The robust method replaces the mean with the median and the standard deviation with the interquartile range. Both replacements are positional rather than magnitude-based, so a single enormous value shifts nothing — this is the method to use when you know outliers are present and legitimate.
All three are computed on the same column and shown side by side. Comparing them is itself informative: when the z-score and the robust values disagree substantially for a row, that row is being pulled by the outliers in the column.
Questions and answers
Which method should I use?
Min–max when you need a bounded 0–1 range and trust your extremes. Z-score for general statistical work. Robust when outliers are present and you want them not to distort everything else.
Does normalising make my data normal?
No — the names are unrelated. Normalisation rescales; it does not reshape. A skewed column stays equally skewed afterwards, which the normality check will confirm.
Why do my z-scores exceed 1?
Because z-scores are not bounded. A standard deviation of one means typical values fall near ±1, but extremes go well beyond. If you need a strict 0–1 range, that is min–max.
Can I reverse the transformation?
Yes, provided you keep the parameters — the min and max, or the mean and standard deviation, or the median and IQR. They are not stored in the exported file, so record them separately.
Is my data uploaded?
No. All three transformations run in your browser.