Coefficient of variation vs standard deviation
How the CV and the standard deviation differ, why a unitless ratio compares data sets a raw spread cannot, and when each one is the right tool.
What each number measures
The standard deviation measures spread in the same units as your data. If you weigh parcels in kilograms, the standard deviation comes back in kilograms and tells you the typical distance of a value from the mean. The coefficient of variation takes that same standard deviation and divides it by the mean, which cancels the units and leaves a pure ratio. So the standard deviation answers how much values vary, while the CV answers how much they vary relative to their average size.
Why the CV travels across units
Because it is unitless, the CV lets you line up data sets that a raw standard deviation cannot. Compare the eight number example, which has a mean of 5 and a sample standard deviation of about 2.13809, giving a CV of roughly 0.427618 or 42.76%. A second data set measured in entirely different units could have a far larger standard deviation yet a smaller CV, and that comparison only makes sense once the units are gone. This is why fields from finance to laboratory science quote the CV, often called the relative standard deviation, to judge consistency across instruments, assays or portfolios.
Sample versus population in the denominator
Both measures come in a sample and a population form, and the only difference is the denominator when averaging the squared deviations. The sample version divides by n minus 1, a correction that compensates for estimating the mean from the same data, and it needs at least two values. The population version divides by n and treats your numbers as the whole group. For the eight number example the sample CV is 42.7618% while the population CV is exactly 40%, because the population standard deviation is 2 rather than 2.13809. Pick the form that matches whether your data is a sample or the full population.
When the CV misleads
The CV shines on ratio-scale data with a meaningful zero, such as weights, times, counts or concentrations. It breaks down when the mean is near zero, because dividing by a tiny mean sends the ratio toward infinity, and the tool refuses a mean of exactly zero for that reason. It is also awkward for data that can go negative, like temperatures in Celsius or profit and loss figures, where a small or sign-changing mean makes the ratio unstable. In those cases the plain standard deviation, or a different measure of spread, tells a clearer story.