When a coefficient of variation is misleading: a mean near zero
Check the denominator and measurement scale before interpreting standard deviation as a percentage of the mean.
Method
The coefficient of variation compares standard deviation with the mean. A percentage form multiplies that ratio by 100. Its usefulness depends on the measurement: a meaningful zero is needed for a relative comparison. Lengths or positive masses can satisfy that condition; Celsius readings do not. Check whether your standard deviation describes a sample or a complete population and keep that choice consistent across groups.
Worked example
With mean 100 and standard deviation 5, the percentage is 5 ÷ 100 × 100 = 5%. With mean 0.1 and the same standard deviation, it becomes 5 ÷ 0.1 × 100 = 5,000%. The large second number mainly reflects the tiny denominator. Moving that mean to 0.2 halves the percentage to 2,500%, even though the standard deviation is unchanged. At mean zero, division is undefined. A result for a negative mean also needs careful interpretation; a non-negative absolute-mean convention does not fix an unsuitable measurement scale.
Checks and limits
Use Coefficient of variation for a relative comparison only after checking those assumptions. Inspect the mean, standard deviation and original observations together. A near-zero mean, mixed positive and negative measurements, or an arbitrary zero makes a stand-alone percentage unreliable. Converting metres into centimetres preserves the ratio, while adding an offset to a temperature scale does not. No universal percentage divides every dataset into good and bad variability, and this ratio is not a confidence interval.
Related calculators
Source: NIST — coefficient of variation