How this instrument works
The coefficient of variation, CV, takes the standard deviation and expresses it as a percentage of the mean: CV = (s ⁄ x̄) × 100%. On its own, a standard deviation is unitless-blind to context — is 5 units of spread a lot? That depends entirely on what you are measuring and around what average. CV strips that ambiguity out by asking a relative question instead: how big is the spread compared to the typical value itself?
That relative framing is what makes CV useful for comparing two things that a raw standard deviation cannot fairly compare — two investments with different average returns, two biological measurements on different scales, two production runs with different target weights. An investment returning an average of 8% with a standard deviation of 4 percentage points (CV = 50%) is proportionally far more volatile than one returning an average of 20% with the same 4-point standard deviation (CV = 20%), even though the raw spread number is identical in both cases.
CV is most meaningful for data measured on a ratio scale with a meaningful, nonzero mean — things like returns, weights, concentrations, or reaction times. It becomes unstable or nonsensical when the mean is close to zero, since dividing by a near-zero number inflates CV wildly even for modest spread, and it is undefined outright when the mean is exactly zero.
- Paste or type your numbers into the Data set field — commas, spaces, or new lines all work as separators.
- The instrument computes the sample standard deviation and the mean from the same data automatically.
- Read the result in Coefficient of variation (%) — it expresses spread as a percentage of the mean, so two data sets on different scales can be compared side by side.
- Lower CV means tighter, more consistent data relative to its own average; higher CV means proportionally more scattered data, regardless of the units involved.
- The instrument needs at least 2 values and a nonzero mean — CV is undefined when the average of your data is exactly zero.
Worked example — the eight quiz scores
Paste the data set {2, 4, 4, 4, 5, 5, 7, 9} — n = 8 — into Data set. The mean is exactly 5, and the sample standard deviation works out to s ≈ 2.13809. Coefficient of variation (%) reads 42.7618%, computed as (2.13809 ⁄ 5) × 100.
That 42.76% says the spread in these scores is a bit under half the size of the average score itself — a genuinely scattered set relative to its own mean. Compare that to a class where every score sat within a point or two of a mean of 50: even with a similar raw standard deviation, the CV would be under 5%, because the same absolute spread matters far less against a much larger average.
Questions
What does the coefficient of variation actually tell me?
It tells you how large the spread in your data is relative to its own average, expressed as a percentage. A CV of 10% means the typical deviation from the mean is about a tenth of the mean's size; a CV of 100% means the spread is roughly as large as the average value itself — a strong sign of highly variable data.
Why use CV instead of just comparing standard deviations directly?
Because raw standard deviations are not comparable across data sets with different means or units. A standard deviation of $500 is enormous for an average transaction of $100 and negligible for one of $1,000,000. CV rescales both to the same relative footing, which is exactly what you need when comparing investment risk, biological variability, or manufacturing consistency across different products or scales.
Is coefficient of variation the same thing as %RSD?
Mathematically, yes — both are (s ⁄ mean) × 100%, the identical formula. Coefficient of variation is the name used in general statistics, finance, and biology; relative standard deviation (%RSD) is the name the analytical chemistry and pharmaceutical world uses for the same calculation, usually applied to repeated instrument readings rather than a population of individuals. Same number, two field-specific names.
Why is CV undefined or misleading when the mean is near zero?
Because you are dividing by the mean, so as the mean approaches zero, CV blows up toward infinity even for a modest, unremarkable standard deviation — and at exactly zero it is undefined outright. CV is only meaningful for data with a genuinely nonzero, ratio-scale mean, such as returns, weights, or concentrations; it is a poor fit for data like temperature in Celsius, where zero is arbitrary rather than a true absence of the quantity.
What counts as a 'high' or 'low' CV?
There is no single universal cutoff — it depends entirely on the field and what is being measured. In manufacturing quality control a CV under 5% might be considered tight and acceptable, while in financial returns a CV under 20% might already count as relatively low volatility. Compare CV within your own field's typical range rather than applying a borrowed threshold from a different domain.
Should I use population or sample standard deviation in the CV formula?
This instrument uses the sample standard deviation (dividing by n−1), the conventional choice when your data is a sample rather than a complete population — the more common real-world case. If your numbers genuinely are the entire population rather than a sample of it, a population-standard-deviation version of CV would be marginally smaller, though the difference shrinks quickly as n grows.