Coefficient of Variation Calculator
Compute CV = (SD / mean) × 100% from a raw dataset or from mean + SD, compare two datasets head-to-head, or invert CV back to SD — with sample/population SD convention clearly labeled and validity guardrails built in.
CV from a dataset
CV from mean and standard deviation
Compare two datasets — which is more consistent?
Reverse: from CV and mean back to SD
The formula
CV = ( σ / μ ) × 100%Divide the standard deviation by the mean and multiply by 100 to express it as a percentage. For the dataset 2, 4, 4, 4, 5, 5, 7, 9: mean μ = 5, population SD σ = 2, so CV = (2 / 5) × 100% = 40.00%. Using the sample SD (÷ n−1) instead: σ = 2.1381, CV = 42.76%. Both numbers are shown by this tool with the convention labeled.
The reverse formula: σ = (CV / 100) × μ. With CV = 15% and mean = 200, σ = 30.
When CV is valid — and when it is not
| Situation | Verdict | Why |
|---|---|---|
| Positive mean, ratio-scale data (length, weight, money, counts) | Valid | True zero point; relative dispersion is meaningful |
| Comparing datasets with different means or units | Valid — ideal use | CV is dimensionless, so scales cancel out |
| Mean close to zero | Misleading | Tiny mean changes swing the CV wildly |
| Mean exactly zero | Undefined | Division by zero — this tool refuses and explains |
| Negative mean | Caution | This tool uses |mean| and flags it explicitly |
| Interval scales (temperature in °C / °F) | Misleading | Arbitrary zero makes the ratio meaningless |
Scope boundary
This tool answers relative dispersion questions: "how variable is this data compared to its own average?" (CV = σ/μ × 100%), "which of two datasets is more consistent?", and "what SD corresponds to this CV?". It does not do:
- Standard error / inference about the mean — "how precise is my sample mean?" (SEM = s/√n) and confidence intervals are the job of the standard error of the mean calculator (#1352), not this one.
- Pooling variances across groups — combining SDs from several groups is the job of the pooled standard deviation calculator, not this one.
- Plain descriptive SD without CV — the average and standard deviation calculator covers basic SD; this tool's purpose is the relative measure and the comparisons it enables.
Worked example
Quality control measures two production lines. Line A: 100, 102, 98, 101, 99 (mean 100, SD 1.4142). Line B: 10, 12, 8, 11, 9 (mean 10, SD 1.4142). Both have the same standard deviation — yet CV(A) = 1.41% and CV(B) = 14.14%. The relative measure reveals that line A is far more consistent, which the raw SD cannot tell you.
Frequently asked questions
What is the coefficient of variation?
The coefficient of variation (CV) is a standardized measure of dispersion: CV = (standard deviation / mean) x 100%. It tells you how large the spread of your data is relative to its average, expressed as a percentage. Because it is dimensionless, you can compare the variability of datasets with different means or different units — something raw standard deviation cannot do.
What is the formula for the coefficient of variation?
CV = (sigma / mu) x 100%, where sigma is the standard deviation and mu is the mean. Worked example with the dataset 2, 4, 4, 4, 5, 5, 7, 9: the mean is 5, the population standard deviation is 2, so CV = (2 / 5) x 100% = 40.00%. This calculator shows each intermediate step.
Should I use the sample or population standard deviation for CV?
It depends on what your data are. Use the population SD (divide by n) when the dataset is the entire group of interest, and the sample SD (divide by n-1) when the data are a sample of a larger population. This tool reports both and labels which one it uses. For the dataset 2, 4, 4, 4, 5, 5, 7, 9 the population CV is 40.00% while the sample CV is 42.76% (sample SD = 2.1381).
When is the coefficient of variation valid?
CV is valid when the mean is positive and the data are on a ratio scale — a scale with a true zero, such as length, weight, money, counts, or time durations. Under those conditions CV is a stable, unit-free measure of relative variability that can be compared across datasets of very different magnitudes.
When is the coefficient of variation misleading?
CV becomes unstable when the mean is close to zero (small changes in the mean swing the CV wildly), is undefined when the mean is exactly zero (division by zero — this tool refuses and explains why), and is meaningless on interval scales with an arbitrary zero, like temperature in Celsius or Fahrenheit. For a negative mean, this tool reports CV relative to the absolute value of the mean with an explicit caution flag.
What is the difference between standard deviation and coefficient of variation?
Standard deviation is an absolute measure of spread; the coefficient of variation is relative to the mean. Compare two datasets with the same SD of 5: dataset A has mean 100, so CV = 5.00%; dataset B has mean 10, so CV = 50.00%. The SD alone makes them look equally variable, but the CV reveals that B is ten times more variable relative to its size.
What is a good coefficient of variation value?
As a rough rule of thumb, a CV of 10% or less indicates low variability, 10% to 30% is moderate, and above 30% is high — but the right threshold is field-dependent. Precision manufacturing may demand a CV under 5%, while biological measurements routinely run 20% to 30%. Treat these bands as a starting interpretation, not a universal verdict.
How do I calculate the coefficient of variation in Excel or Google Sheets?
For population CV enter =STDEV.P(B2:B9)/AVERAGE(B2:B9) and format the cell as a percentage; for the dataset 2, 4, 4, 4, 5, 5, 7, 9 this gives 40.00%. For sample CV use =STDEV.S(B2:B9)/AVERAGE(B2:B9), which gives 42.76% on the same data. This calculator reproduces both numbers exactly.
Can the coefficient of variation be greater than 100%?
Yes. A CV above 100% simply means the standard deviation is larger than the mean — common in highly skewed data, counts with many zeros, or volatile series. Example: a dataset with mean 4 and standard deviation 5 has CV = (5 / 4) x 100% = 125.00%.
Is the coefficient of variation the same as relative standard deviation (RSD)?
Essentially yes. Relative standard deviation is SD divided by the mean; the coefficient of variation is the same ratio expressed as a percentage. CV = RSD x 100%. In analytical chemistry, RSD is the more common name; in statistics and quality control, CV dominates.
Why is the coefficient of variation undefined when the mean is zero?
Because CV requires dividing by the mean, and division by zero is undefined. For example the dataset 1, -1, 2, -2 has mean exactly 0, so no CV exists — this tool reports that instead of a nonsense number. Even a mean merely close to zero makes the CV numerically unstable, so treat such results with suspicion.
How do I compare the consistency of two datasets with CV?
Compute the CV of each dataset and take the smaller one as more consistent. Example: dataset A (100, 102, 98, 101, 99) has mean 100, population SD 1.4142, CV = 1.41%; dataset B (10, 12, 8, 11, 9) has mean 10, population SD 1.4142, CV = 14.14%. Although both have the same SD, A is the more consistent dataset. The comparator tab of this tool runs this calculation for you.
How do I find the standard deviation from CV and mean?
Rearrange the formula: SD = (CV / 100) x mean. Example: if CV = 15% and the mean is 200, then SD = (15 / 100) x 200 = 30. The reverse tab of this tool performs this inversion; it requires a non-zero mean.
Does the coefficient of variation depend on units?
No. CV is dimensionless: the units of the standard deviation cancel against the units of the mean. A dataset with mean 5 kg and standard deviation 2 kg has CV = 40.00%, and the same data measured in grams gives the identical 40.00%. Multiplying every value by a constant leaves the CV unchanged.
Can I use CV with negative values in my data?
Only with care, and only when the mean itself stays positive. This tool always divides by the absolute value of the mean and flags a negative mean explicitly. Example: the dataset -2, -4, -6 has mean -4 and population SD 1.6330, so the reported CV is (1.6330 / 4) x 100% = 40.82%, flagged as computed on |mean|. If your data mix positives and negatives around zero, prefer a different dispersion measure.
Related statistics tools
Free tool by NRTechStudio