Two services can have the same spread in delivery times while operating at very different average speeds. Comparing standard deviations alone answers one question. Comparing that spread with each service’s mean answers another.
The NIST definition of the coefficient of variation divides standard deviation by the mean. For positive measurements on an appropriate scale, the result is often expressed as a percentage.
Try the calculation with a concrete example
Imagine Service A has a mean delivery time of 10 minutes and a standard deviation of 2 minutes. Its coefficient is 0.2, or 20%. Service B has a mean of 40 minutes and a standard deviation of 2 minutes: 0.05, or 5%.
These invented figures illustrate relative spread. They do not show that Service B is preferable. Someone needing a quick delivery may care more about the much shorter average, while a tightly timed process may care about the actual spread in minutes.
The scale has to make sense
A meaningful zero is important. Dividing the standard deviation of Celsius temperatures by their mean is not a dependable relative-temperature comparison: changing to another temperature scale changes the result. A mean close to zero can also make the ratio extremely large or unstable.
Before interpreting a published coefficient, check the variable, units, mean and definition used. Do not assume “relative variability” means the same calculation in every report. NIST distinguishes its coefficient command from a relative-standard-deviation convention that uses the absolute mean.
Keep the original units beside the ratio in your notes. “20%, with a mean of 10 minutes and a standard deviation of 2 minutes” lets a reader understand both scale and spread.
The coefficient clarifies a suitable comparison. It becomes less useful when a small percentage is treated as a score of overall quality, with the underlying measurements hidden.

