Read the ratio
If mean = 50 and SD = 5, CV = .10 or 10%. If another ratio-scale variable has mean 200 and SD 20, it also has CV 10%. The measure supports relative comparison across positive ratio scales even when original units differ.
That comparison is useful only when the measurement processes and populations are comparable. A sample CV is itself uncertain, especially with small n or skewed data. Report the mean, SD, n, and interval or repeated-run evidence when a process decision depends on the value.
Standard deviation interpretation explains the numerator. The CV does not replace the mean, SD, sample size, or distribution plot.
When CV breaks down
CV is unstable when the mean is near zero because a small denominator produces a large ratio. It is generally inappropriate for interval scales with arbitrary zero points, such as Celsius temperature, because shifting the origin changes the CV. Negative means also make ordinary percentage wording difficult to interpret.
Skew, outliers, censoring, and mixtures can make mean and SD poor building blocks. Use a domain-appropriate robust relative spread measure if needed.
For example, mean = 2 and SD = 1 gives CV 50%; shifting an interval-scale variable by 100 units changes the CV even though all pairwise differences stay the same. This illustrates why a meaningful zero is a requirement, not a technical footnote.
Thresholds are discipline conventions
Laboratory precision limits, manufacturing tolerances, and ecological variability can use different CV targets. A source from one process does not establish a universal 10%, 20%, or 30% rule. The provenance guide shows how any printable cutoff needs a named context and source.
A defensible reporting sentence
“The mean was 50 units, SD 5 units, and CV 10%. Because the variable had a meaningful zero and the mean was well above zero, CV was used for relative spread; acceptability was judged against the stated process requirement.”