How to use this coefficient of variation calculator
Enter your dataset above, choose Sample or Population, and press Calculate. The CV is reported as a percentage, next to the mean and standard deviation it was built from.
Spread you can compare across units
A standard deviation of 2, is that a lot? For bottle fills averaging 500 ml, trivial. For diamond weights averaging 3 carats, enormous. The coefficient of variation settles the question by expressing spread as a percentage of the mean, stripping the units away entirely. Now a bottling line (in millilitres) and a bakery (in grams) can be compared on one scale: whose process is relatively steadier?
A filling machine's eight bottles (ml): 498, 502, 500, 501, 499, 503, 497, 500. Mean exactly 500, sample SD exactly 2, so CV = 0.4%, a very consistent machine. If the bakery's flour scoops had a CV of 6%, the bottles are fifteen times steadier in relative terms, a comparison the raw SDs (2 ml vs 30 g) could never have made.
Rules and limits
- CV only makes sense for ratio data with a real zero (weights, prices, times). Never use it for temperatures in °C. The "zero" is arbitrary and the CV becomes nonsense.
- When the mean is near zero, the CV explodes; treat values above ~30% as "this summary is struggling" rather than a precise figure.
- Rule-of-thumb bands: under 10% low, 10–20% moderate, 20–30% high relative variation.
Frequently asked questions
What is a good coefficient of variation?
Context decides, but a common guide is: under 10% low variation, 10–20% moderate, 20–30% high, above 30% very high. Precision manufacturing often demands under 1%.
Why use CV instead of standard deviation?
The standard deviation carries units and scale; the CV is a pure percentage, so it can compare variability between datasets measured in different units or at very different sizes.
Can CV be used for temperature?
Not in Celsius or Fahrenheit, whose zero points are arbitrary. In Kelvin, a true ratio scale, it is legitimate.