Statistics
How To Calculate Coefficient Of Variation
The coefficient of variation measures spread relative to the size of the mean, so it lets you compare variability across data sets with very different units or scales. It is the standard deviation divided by the mean, normally reported as a percentage. This page explains the formula, shows how the squared CV and the relative standard error follow from it, and works through examples so you can use the measure with confidence.
Quick Answer
CV = standard deviation / mean
- CV
- Coefficient of variation, the relative spread
- sd
- Standard deviation of the data
- mean
- Arithmetic mean of the data
- n
- Number of observations in the sample
Divide the standard deviation by the mean to get the raw coefficient of variation. With a standard deviation of 15 and a mean of 60 the raw value is 0.25, which is displayed as 25 percent. Squaring it gives a dimensionless variance ratio of 0.0625, dividing by the square root of the 25 observations gives a relative standard error of 0.05, and four times the standard deviation gives an approximate range of 60.0.
What Is Coefficient Of Variation?
The coefficient of variation, usually shortened to CV, is the standard deviation divided by the mean. It measures spread relative to the size of the average and is normally shown as a percentage. With a standard deviation of 15 and a mean of 60 the raw ratio is 0.25, which a renderer displays as 25 percent. Because both parts carry the same units, those units cancel and the result is dimensionless. This is what makes the CV useful for comparing variability between quantities that could not otherwise be placed side by side.
A standard deviation of 15 means very different things depending on whether the mean is 60 or 6,000. In the first case the spread is a quarter of the average; in the second it is a mere quarter of a percent. The CV strips away the scale and reports spread as a fraction of the centre, so a small laboratory instrument and a large industrial process can be compared fairly. Absolute spread answers how many units values typically deviate, while relative spread answers whether that deviation matters given the size of the typical value.
The idea is also called the relative standard deviation, and the two names describe exactly the same calculation. Dividing by the mean grew out of the recognition that many natural and economic quantities vary in proportion to their level, so their spread grows as they grow. Expressing variability as a fraction of the mean makes it scale-free and easy to communicate. A CV of 25 percent says that the typical deviation is a quarter of the mean, regardless of whether the underlying measurements are grams, seconds or pounds.
The direction of change is worth stating clearly. Holding the mean fixed at 60 and raising the standard deviation from 15 to 24 lifts the raw CV from 0.25 to 0.4, because the data is more dispersed. Holding the standard deviation at 15 and doubling the mean to 120 halves the CV to 0.125, because the same absolute spread is now smaller relative to a larger centre. Confusing these directions is easy when reading a report, so check whether it was the spread or the mean that moved.
Squaring the CV gives a dimensionless variance ratio, the variance divided by the square of the mean. For the default inputs the squared CV is 0.0625, which is simply 0.25 multiplied by itself. This quantity is convenient in algebra because variances add and combine more cleanly than standard deviations, so the squared form often appears in derivations. It carries no units and no percentage sign, being a pure ratio of two squared quantities.
Dividing the CV by the square root of the sample size gives a relative standard error, which measures how precisely the CV has been estimated. With 25 observations the square root is 5, so the standard error is 0.25 divided by 5, or 0.05. Increasing the sample size shrinks this error: a hundred observations would halve it to 0.025. This is the reason analysts quote a sample size alongside a CV, since a figure based on a handful of readings is far less trustworthy than one based on many.
The approximate range is four times the standard deviation, giving 60.0 for a standard deviation of 15. This is a rough span that, for a roughly bell-shaped distribution, contains most observations: about two standard deviations on either side of the mean. It is a rule of thumb rather than an exact bound, and it depends on the distribution being reasonably symmetric. For skewed data the range can be a poor guide, but it is a fast way to sense the overall width of a data set.
The CV is not defined when the mean is zero, since division by zero has no meaning, and it behaves badly when the mean is close to zero, where tiny changes produce enormous swings. It also loses meaning for data that can cross zero, such as temperature in degrees Celsius or profit and loss figures, because positive and negative values cancel in the mean. A high CV is not automatically bad: in some processes variability is expected, while in others even a few percent is unacceptable. The number needs context before it can be judged.
The calculator stores the raw ratio in a helper named cvRaw and reuses it for the primary output and the squared CV. Because the primary output carries a percentage format, the check value is kept as the raw decimal 0.25 and the interface displays 25 percent; the expression never multiplies by 100. The standard error divides cvRaw by the square root of the sample size, and the approximate range scales the standard deviation by four. With the default inputs of 15, 60 and 25 the four outputs are 0.25, 0.0625, 0.05 and 60.0.
Formula
CV = sd / mean
Formula page →The standard deviation divided by the mean. With sd equal to 15 and mean equal to 60 the raw ratio is 0.25, displayed as 25 percent.
| Symbol | Meaning | Unit | Notes |
|---|---|---|---|
| sd | Standard deviation | same as data | Absolute spread in the original units; raising it raises the CV. |
| mean | Arithmetic mean | same as data | The centre of the data; raising it lowers the CV for fixed spread. |
CV^2 = (sd / mean)^2
Formula page →The variance divided by the square of the mean, a dimensionless ratio. For the defaults it is 0.0625.
| Symbol | Meaning | Unit | Notes |
|---|---|---|---|
| CV | Coefficient of variation | dimensionless | Squaring the raw ratio removes the percentage scale, giving a pure number. |
| CV^2 | Variance ratio | dimensionless | Used in algebra because variances combine more cleanly than standard deviations. |
SE = CV / sqrt(n)
Formula page →The CV divided by the square root of the sample size. With CV of 0.25 and n of 25 the result is 0.05.
| Symbol | Meaning | Unit | Notes |
|---|---|---|---|
| SE | Relative standard error | dimensionless | Shrinks as the sample grows, reflecting a more precise estimate. |
| n | Sample size | count | Number of observations; the square root is the relevant divisor. |
How To Calculate Coefficient Of Variation
- 1
Find the standard deviation and the mean
The coefficient of variation needs two summaries of the same data: its standard deviation and its arithmetic mean. In the worked case these are supplied directly as 15 and 60. If you start from raw observations, compute the mean first and then the standard deviation, choosing the sample or population version consistently. Make sure both figures describe the same set and the same units.
- 2
Divide the standard deviation by the mean
Compute the raw ratio sd divided by mean. With 15 and 60 the result is 0.25. This is the coefficient of variation before any percentage is applied; the interface displays it as 25 percent because the output carries a percentage format, while the underlying number stays 0.25. Never insert a multiplication by 100 into the expression, or the displayed figure would be a hundred times too large.
- 3
Square the CV for the variance ratio
Multiply the raw CV by itself to get the dimensionless variance ratio. With a CV of 0.25 the squared value is 0.0625. This companion figure is useful in further algebra, since variances combine additively. It is a pure number with no units, and it will always be smaller than the CV whenever the CV is less than one.
- 4
Divide by the square root of the sample size
For the relative standard error, divide the raw CV by the square root of the number of observations. With 25 observations the square root is 5, and 0.25 divided by 5 gives 0.05. A larger sample makes this figure smaller, showing that the estimated CV is more precise. Always report the sample size next to a CV so readers can judge how much weight to give it.
- 5
Estimate the approximate range and check direction
Multiply the standard deviation by four to get a rough span for the bulk of the data: 4 times 15 is 60.0. Then sanity-check the direction of the main result. A larger standard deviation with the same mean must raise the CV, while a larger mean with the same spread must lower it. If your figure moves the other way, the two inputs have probably been swapped.
Examples
Example 1: The worked default — sd 15, mean 60, 25 observations
- standardDeviation
- 15
- meanVal
- 60
- sampleSize
- 25
| Step | Calculation | Result |
|---|---|---|
| Raw ratio | 15 / 60 | 0.25 |
| Squared CV | 0.25^2 | 0.0625 |
| Relative standard error | 0.25 / sqrt(25) | 0.05 |
| Approximate range | 15 * 4 | 60.0 |
| Coefficient of variation | 0.25, shown as a percentage | 0.25 (25%) |
Result: The coefficient of variation is 0.25, displayed as 25 percent. The squared CV is 0.0625, the relative standard error is 0.05 and the approximate range is 60.0.
Example 2: A smaller spread — sd 9, mean 45, 16 observations
- standardDeviation
- 9
- meanVal
- 45
- sampleSize
- 16
| Step | Calculation | Result |
|---|---|---|
| Raw ratio | 9 / 45 | 0.2 |
| Squared CV | 0.2 * 0.2 | 0.04 |
| Relative standard error | 0.2 / sqrt(16) | 0.05 |
| Coefficient of variation | 0.2, shown as a percentage | 0.2 (20%) |
Result: The coefficient of variation is 0.2, displayed as 20 percent. The squared CV is 0.04, the relative standard error is 0.05 and the approximate range is 36.0.
Example 3: A larger sample — sd 20, mean 80, 100 observations
- standardDeviation
- 20
- meanVal
- 80
- sampleSize
- 100
| Step | Calculation | Result |
|---|---|---|
| Raw ratio | 20 / 80 | 0.25 |
| Squared CV | 0.25^2 | 0.0625 |
| Relative standard error | 0.25 / sqrt(100) | 0.025 |
| Coefficient of variation | 0.25, shown as a percentage | 0.25 (25%) |
Result: The coefficient of variation stays at 0.25, displayed as 25 percent, because spread and mean scaled together. The squared CV is 0.0625, the relative standard error falls to 0.025 and the approximate range is 80.0.
Calculator
Coefficient of variation
25.00%
- Squared CV
- 0.0625
- Relative standard error
- 0.05
- Approximate range
- 60
Values update as you type. This calculator covers the single scenario its formula assumes — see Common Mistakes for what it leaves out.
Prefer a full-width tool? Open the Coefficient Of Variation calculator page.
Common Mistakes
Multiplying by 100 inside the expression
The output already carries a percentage format, so the renderer applies the scaling. Writing a multiplication by 100 in the expression makes the displayed figure a hundred times too large, turning 25 percent into 2,500 percent. Keep the raw decimal such as 0.25 in the formula and let the format handle the display.
Dividing by the standard deviation instead of the mean
The ratio runs from spread to centre, not the other way round. The reciprocal answers a different question and can exceed one even for very consistent data. Check the direction with a simple case: if the spread is 15 and the mean is 60, a quarter is sensible, whereas four is not.
Applying the CV when the mean is zero or near zero
A zero mean makes the division undefined, and a mean close to zero makes the ratio explode with tiny changes. Data that can cross zero, such as temperature in Celsius or profit and loss, is especially troublesome because the mean may cancel toward nothing. In those cases report the standard deviation on its own.
Ignoring the sample size
A CV computed from a handful of readings is far less reliable than one from hundreds, yet the two can look identical. Compute the relative standard error and quote the sample count so readers can judge precision. With 25 observations the standard error is 0.05; with 100 it falls to 0.025 for the same CV.
Assuming a higher CV is always worse
The CV only describes relative variability, not quality. Some biological and financial series are naturally variable, and a high figure can be entirely expected. Conversely a low CV can hide a systematic bias that a measure of spread will not detect. Interpret the number against what is normal for the field before drawing a conclusion.
Mixing sample and population standard deviations
The CV inherits whatever convention was used for the spread, so a sample standard deviation divided by n minus one and a population standard deviation divided by n give different CVs for the same data. State which convention was used and keep it consistent when comparing groups, especially when the samples are small.
FAQ
What is the coefficient of variation?
It is the standard deviation divided by the mean, a measure of spread relative to the size of the average. It is normally shown as a percentage and is dimensionless, so it can compare variability across data sets with different units. With a standard deviation of 15 and a mean of 60 the raw value is 0.25, displayed as 25 percent.
How do I calculate the coefficient of variation?
Take the standard deviation of your data, take the mean of the same data, and divide the first by the second. With 15 and 60 you get 0.25, which the interface shows as 25 percent. The expression only produces the raw decimal; the percentage formatting is applied separately, so never multiply by 100 yourself.
Why use the CV instead of the standard deviation?
Because the standard deviation is tied to the units and scale of the data, so it cannot fairly compare a small process with a large one. The CV removes the scale by dividing by the mean. A spread of 15 is a quarter of a mean of 60 but only a small fraction of a mean of 6,000, and the CV captures that difference directly.
What is the squared coefficient of variation?
It is the CV multiplied by itself, equal to the variance divided by the square of the mean. For the default inputs it is 0.0625. Because variances combine more neatly than standard deviations, the squared form is often used in algebra and in deriving properties of ratios, even though the ordinary CV is easier to read.
How does sample size affect the result?
The CV itself does not change with sample size, but the precision of the estimate does. The relative standard error divides the CV by the square root of the number of observations. With 25 observations it is 0.05; with 100 it drops to 0.025. Always quote the sample count so the reliability of the figure is clear.
When should I avoid using the coefficient of variation?
Avoid it when the mean is zero or close to zero, since the division becomes unstable or undefined. It also misleads for measurements that can take negative values, where positive and negative readings may cancel in the mean. In those situations the standard deviation alone is a more honest description of spread.
References
- [1], Coefficient of variation — https://en.wikipedia.org/wiki/Coefficient_of_variation
- [2], Standard deviation — https://en.wikipedia.org/wiki/Standard_deviation
- [3], Relative standard deviation — https://en.wikipedia.org/wiki/Relative_standard_deviation
