Coefficient of Variation Calculator
Compute the coefficient of variation (CV = σ / mean × 100%) for a single dataset, or compare two datasets on different scales to see which is relatively more variable.
Coefficient of variation explained, step by step
CV expresses the standard deviation as a percentage of the mean, so it stays comparable across datasets on completely different scales. Each card below walks through one piece of the workflow this calculator runs.
1. What CV measures — spread relative to the mean
The coefficient of variation is the standard deviation expressed as a percentage of the mean. Because dividing by the mean cancels the units, CV is a unit-free measure of relative spread — two datasets on completely different scales can be compared directly.
2. Why CV wins for cross-scale comparison
Standard deviation carries the units of the data, so raw SD comparisons across different scales are meaningless. CV strips the units and puts every dataset on a common percentage footing — the one with the higher CV is the more variable relative to its own average.
3. Sample (÷ n − 1) vs population (÷ n) SD
The CV toggle switches which standard deviation goes in the numerator. Use sample SD (Bessel's correction, ÷ n − 1) for the usual case where the data is a sample from a larger population; use population SD (÷ n) only when the data really is the entire population.
4. When CV isn't meaningful
CV is only sensible on ratio-scale data that is strictly positive: mass, income, length, time, counts. It breaks down when the mean is near zero, the data crosses zero, or the scale has no true zero (Celsius, IQ, pH, calendar years).
Rough interpretation bands
| CV range | Label | Typical setting |
|---|---|---|
| < 10% | Low variability | Lab measurements, calibrated instruments. |
| 10% – 30% | Moderate variability | Business KPIs, most survey data. |
| > 30% | High variability | Financial returns, biological counts, network latency. |
These bands are conventions, not laws. Always compare to what is normal for your field.
Common mistakes
- Comparing SDs across different units. Convert to CV first, then compare.
- Using CV when the mean can be zero or negative. The formula divides by the mean; that step must make sense.
- Mixing sample and population SD. Pick one convention for all datasets you're comparing.
- Reading CV as an absolute quality score. A CV of 15% is "high" in a physics lab and "low" for daily stock returns.
Features of this calculator
- Single-dataset mode — mean, standard deviation and CV as a percentage
- Compare-two-datasets mode — side-by-side CV even when means and units differ
- Population (÷ N) or sample (÷ N − 1) standard deviation toggle
- Bar chart comparing CV values with a shared percentage axis
- Full step-by-step working: mean, squared deviations, SD and the CV formula
- Automatic classification — low / moderate / high variability
- Copy the summary or download the whole result panel as an image
Frequently asked questions
+What counts as a 'high' coefficient of variation?
Common rules of thumb: below 10% is low, 10–30% is moderate, above 30% is high. But it is field-dependent — a 5% CV is unacceptable for some lab assays and impressive for equity returns.
+When is CV NOT meaningful?
When the mean is near zero, the data crosses zero, or the variable isn't on a ratio scale (e.g. temperature in Celsius, IQ, pH). In those cases dividing by the mean doesn't give a real measure of relative spread.
+Why is CV shown as a percentage?
Multiplying by 100 turns σ/x̄ into a readable percentage — "CV = 12%" is easier to grasp and easier to communicate than "0.12".
+Which datasets should I use CV to compare?
Any two (or more) datasets measured on positive ratio scales — different units, different magnitudes, doesn't matter. The whole point of CV is that it makes those comparisons fair.