Coefficient of Variation Calculator
Calculate the coefficient of variation (CV) two ways: paste raw numbers for a full breakdown, or enter a mean and standard deviation directly if you only have summary statistics. Compare two data sets side by side, apply the unbiased correction for small samples, or reverse-solve for the standard deviation needed to hit a target CV.
Formula
Worked example
For 12, 15, 14, 18, 11, 16: mean = 14.33, sample SD = 2.582, so CV = (2.582 / 14.33) x 100 = 18.01%. With the unbiased correction (n = 6): CV_corrected = 18.01% x (1 + 1/24) = 18.76%.
What the coefficient of variation tells you
The coefficient of variation, sometimes called the relative standard deviation (RSD), rescales the standard deviation by the mean and reports the result as a percentage. Because it is a ratio of two quantities with the same units, the units cancel and the CV becomes a pure dimensionless number. That property makes it ideal for comparing the consistency of data sets that would otherwise be impossible to line up: the variability of monthly rainfall against the variability of daily stock returns, or the precision of two laboratory instruments that measure on completely different scales. A production process that averages 500 kg with a standard deviation of 25 kg has the same CV as one averaging 5 kg with a standard deviation of 0.25 kg - both are 5% relative variation.
Three input modes: raw data, summary statistics, and reverse-solve
This calculator works in three modes. In raw-data mode, you paste a comma-separated list and get the full breakdown: mean, standard deviation, variance, range, median, and CV. In summary-statistics mode, you type in a mean and standard deviation you already have (for instance, from a published paper or a spreadsheet) and the calculator converts them to a CV instantly. In reverse-solve mode, you specify the target CV and the expected mean, and the calculator tells you the maximum standard deviation your data can have to stay within that CV - useful when designing acceptance criteria for a manufacturing or lab process. Switch between modes with the Input mode selector at the top.
Population versus sample, and the unbiased correction
If your numbers represent every member of the group you care about, use the population basis, which divides squared deviations by N. If they are only a sample drawn from a larger population, use the sample basis (divides by N minus one) to correct for the tendency of samples to underestimate spread. For very small samples there is an additional positive bias in the CV itself: the formula systematically underestimates the true population CV. Omni Calculator and Kenney and Keeping (1951) recommend multiplying the raw CV by the factor (1 + 1/4n) to correct for this bias. Toggle that correction on when you have fewer than about 30 observations and need a more conservative estimate.
Comparing two datasets and reading the result in context
To compare relative variability across two groups - even if they are measured in different units or have very different magnitudes - enter the second dataset in the Dataset B field. The comparison table shows count, mean, SD, variance, min, max and CV for both groups side by side. The dataset with the lower CV is relatively more consistent regardless of scale. There is no single universal threshold for a good or bad CV: analytical chemistry may require a CV under a few percent, while in ecology or economics a CV of 30% or more can be normal. As a rough rule of thumb, many analysts treat a CV below 15% as low variability, 15 to 30% as moderate, and above 30% as high, but always interpret the figure against the accepted norms of your own field.
When the coefficient of variation should not be used
The CV has two important limitations. First, it is undefined when the mean equals zero and becomes unstable when the mean is very close to zero: the ratio blows up. Second, it should only be used with ratio-scale data that has a meaningful, non-arbitrary zero point. Temperature in Celsius or Fahrenheit fails this test: shifting the scale (converting 20 degrees Celsius to 293 Kelvin) changes the CV, which makes the result meaningless for comparison. Absolute temperature in Kelvin, mass, length, time, and concentration all have true zeros and are suitable for CV analysis.
Coefficient of variation benchmarks by field
| Field or context | Typical CV range | Interpretation |
|---|---|---|
| Analytical chemistry (instrument precision) | Below 2% | Excellent repeatability required |
| Clinical lab assays (in-lab repeatability) | 2% - 5% | Acceptable precision |
| Clinical lab assays (between-lab) | 5% - 15% | Normal inter-lab variation |
| Manufacturing process control (Six Sigma) | Below 10% | Process in control |
| Financial returns (low-volatility portfolio) | 10% - 30% | Moderate relative risk |
| Biological or ecological field data | 20% - 50% | High natural variability |
| Survey or questionnaire data | Above 30% | High heterogeneity in responses |
Rough guidelines only. What counts as acceptable varies widely within each field.
Frequently asked questions
What is the difference between CV and standard deviation?
The standard deviation measures spread in the original units of the data, while the coefficient of variation expresses that same spread relative to the mean as a unitless percentage. Use the standard deviation when the units and scale of the data are fixed and meaningful. Use the CV when you need to compare variability across datasets with different scales, units, or mean values.
Should I use the population or sample formula?
Use the population formula (divide by N) only when your data includes every member of the group you are studying. Use the sample formula (divide by N minus one) when your data is a subset drawn from a larger population. In practice, most real-world data is a sample, so the sample basis is the more common choice.
What is the unbiased CV correction and when should I use it?
The standard CV formula (SD divided by mean, times 100) slightly underestimates the true population CV when the sample is small, because small samples tend to underestimate the true standard deviation. Multiplying the raw CV by the factor (1 + 1/4n) - where n is the sample size - corrects for this bias. It matters most when n is below about 20 to 30. For larger samples the correction is negligible (less than 1% difference with n = 100).
Why is the coefficient of variation sometimes undefined?
Because CV divides by the mean, it is undefined when the mean equals zero and becomes unreliable when the mean is close to zero. It is also inappropriate for interval-scale data such as temperature in Celsius or Fahrenheit, where the position of zero is arbitrary: shifting the scale changes the CV and makes comparison meaningless.
How do I use reverse-solve mode?
Select "Reverse-solve" from the Input mode dropdown, then enter the target CV percentage and the expected mean. The calculator rearranges the formula (SD = CV / 100 x |mean|) to find the maximum standard deviation compatible with your target. This is useful in quality control and lab method validation, where you need to set a specification limit on variability.
Can I compare two datasets with different units using this calculator?
Yes. Enter the first dataset in Dataset A and the second in Dataset B. The comparison table appears below the result and shows all key statistics side by side. Because CV is dimensionless, you can compare a dataset measured in kilograms against one measured in dollars or milliseconds. The dataset with the smaller CV is relatively more consistent regardless of units.