Coefficient of Variation Calculator

What Does a Coefficient of Variation Calculator Do?

A coefficient of variation calculator measures how large the spread of a data set is compared with its average. It takes your numbers, works out the mean and the standard deviation, and divides one by the other to give the CV, usually written as a percentage. A CV of 15% means the typical value sits about 15% of the mean away from the mean.

That relative scale is the whole point. A standard deviation of 6 means very little for house prices and a great deal for body temperature. The CV strips the units away, so you can compare the consistency of two data sets that are measured in different units or that have very different averages.

This calculator accepts a raw list of numbers or, if you already have them, just a mean and a standard deviation. It can treat your data as a sample or a whole population, compare two data sets side by side, and show every step, so a homework answer can be checked line by line. If you need the mean first, the mean, median and mode calculator finds it, and the standard deviation calculator gives the spread.

What Is the Coefficient of Variation Formula?

The coefficient of variation is the standard deviation divided by the absolute value of the mean, multiplied by 100 to express it as a percentage.

Here s is the standard deviation and x̄ is the mean. For a whole population the symbols are σ and μ, so CV = σ / |μ| × 100%. The two versions differ only in how the standard deviation is computed: the sample version divides the sum of squared deviations by n − 1, the population version by n.

The CV is also called the relative standard deviation (RSD), and in laboratory work it is often quoted as %RSD or %CV. They are the same number. Because it is a ratio of two quantities in the same units, the units cancel and the result has no unit.

Rearranged, the formula gives a handy conversion: standard deviation = CV × mean. If a product line has a CV of 4% and an average weight of 500 grams, the standard deviation is about 20 grams.

How Do You Use This Coefficient of Variation Calculator?

Choose how you want to enter the information, then type. Results update as you go.

1. Pick the input type. “Raw data” takes a list of numbers separated by commas, spaces, tabs, semicolons or new lines, and a column pasted from Excel works as is. “Mean and standard deviation” takes the two summary figures directly.
2. Choose sample or population. Sample (the default) uses n − 1 in the standard deviation. Population uses n. In summary mode this choice does not matter, because the standard deviation you type is already final.
3. Add a second data set if you want to compare consistency. The calculator then says which set is more variable in relative terms.
4. Set the decimal places from 2 to 8. Only the display is rounded; the arithmetic keeps full precision.

The quick buttons load examples, including a two-store comparison and a mean-and-SD case. Press Clear to return to the starting example.

How Do You Calculate the Coefficient of Variation by Hand?

Find the mean, find the standard deviation, divide, and multiply by 100. Take the five values 4, 8, 6, 10 and 12.

If the five values are the entire population, step 4 divides by 5 instead, which gives a variance of 8, a standard deviation of 2.8284 and a CV of 35.36%. A quick check on step 2: the signed deviations must add to zero. If they do not, the mean is wrong.

Example 1: Which of Two Stores Has More Variable Sales?

Store A sold 42, 45, 47, 50, 52, 55, 58 and 60 units over eight weeks. Store B sold 120, 135, 150, 160, 185, 190, 210 and 230. Both look uneven, but B sells at a much larger scale, so comparing standard deviations alone would be unfair.

Store B’s standard deviation is almost six times larger, but its average is more than three times larger as well. Once the scale is removed, the gap shrinks to a ratio of about 1.77: store B’s weekly sales are relatively more variable, with a typical week 21.9% away from its average against 12.4% for store A.

For store A, the squared deviations from the mean of 51.125 add up to 280.875. Dividing by 7 gives a variance of 40.125, the square root of which is 6.3344, and 6.3344 / 51.125 = 0.1239. The variance calculator runs the same steps on their own if you only need the variance.

Example 2: How Do You Compare Variability When the Units Differ?

This is the job the CV was made for. Six adults have heights of 158, 163, 170, 172, 175 and 181 cm and weights of 52, 60, 68, 74, 80 and 95 kg. Which measurement varies more?

A standard deviation of 8.28 centimetres and one of 15.20 kilograms cannot be compared directly, because the units are different. Their CVs can: weight varies roughly four times as much as height in relative terms, which matches everyday experience.

If you only have the summary figures, switch the calculator to “Mean and standard deviation.” A mean of 170 with an SD of 8 gives 4.71%, and a mean of 70 with an SD of 12 gives 17.14%.

Example 3: What Is the CV Used for in a Lab (Percent RSD)?

In analytical chemistry and quality control, the CV is the standard way to express precision. Six replicate measurements of the same sample give 98.2, 99.1, 97.8, 98.6, 99.4 and 98.0. The mean is 98.5167 and the sample standard deviation is 0.6338, so the CV (the %RSD) is 0.64%.

A small CV here means the method repeats well. It says nothing about whether the results are correct. A scale that always reads 5 grams too high can have an excellent CV and still be wrong, because the CV measures spread, not accuracy.

Labs and regulators usually set their own acceptable limits for a given method, and those limits depend on the concentration and the technique. Check the requirement that applies to your work rather than relying on a general rule of thumb.

Should You Use the Sample or the Population Coefficient of Variation?

Use the sample version when your numbers are a subset of a larger group, which is the usual case. Use the population version only when you have measured every member of the group you care about.

The two answers get close as n grows, but for small data sets they differ noticeably:

The sample standard deviation is itself a slightly biased estimate, and so is the CV built from it. For small samples from roughly normal data, some textbooks multiply the sample CV by (1 + 1/(4n)). For store A that turns 12.39% into 12.78%. The calculator shows this adjusted figure as an extra, and it should not be used when your course expects the plain formula.

When Is the Coefficient of Variation Misleading?

The CV breaks down whenever the mean is close to zero, because dividing by a tiny number produces a huge, unstable ratio. If the mean is exactly zero it is undefined. Take the values −2, −1, 0, 1 and 2: the mean is 0, the sample standard deviation is 1.5811, and the CV cannot be computed. The calculator reports it as undefined instead of showing a meaningless number.

It is also meaningful only for data measured on a ratio scale, where zero really means “none.” Temperature in degrees Celsius is the classic trap. Take 10, 20 and 30 °C: the mean is 20, the sample SD is 10, and the CV is 50%. Convert the same readings to kelvin (283.15, 293.15, 303.15) and the SD is still 10, but the mean is 293.15, so the CV drops to 3.41%. The spread has not changed. Only the arbitrary zero moved, and the CV followed it.

Negative means are another warning sign. The formula uses the absolute value of the mean, so the CV stays positive, but a data set that mixes positive and negative values rarely has a sensible “relative” spread. Finally, the CV describes spread in the original scale. For log-normal data with a log-scale standard deviation of 0.5, the exact CV is √(e^0.25 − 1) = 53.29%, which you cannot read straight off the logged numbers.

How Do You Interpret a Coefficient of Variation?

A lower CV means the values cluster more tightly around their mean, and a higher CV means they are more scattered, relative to the average. There is no universal cutoff for “high” or “low.” A CV of 5% would be poor for a laboratory assay and remarkable for a field measurement of crop yield.

Useful ways to read it:

• Against a benchmark. Compare with the CV of similar processes, past batches or published values in your field.
• Between groups. The group with the lower CV is the more consistent one, as with the stores and measurements above.
• As a back-of-envelope SD. Multiply the CV by the mean to recover the standard deviation in original units.

The CV compares spreads. It does not test whether two CVs differ significantly. With small samples, a gap such as 12% against 22% can be nothing more than chance, and a formal comparison test is the right next step.

How Do You Calculate the Coefficient of Variation in Excel, R and Python?

Excel, Google Sheets and R have no dedicated CV function, but each makes it a one-line formula.

Excel / Sheets (sample)       =STDEV.S(A2:A9)/AVERAGE(A2:A9)
Excel / Sheets (population)   =STDEV.P(A2:A9)/AVERAGE(A2:A9)
R (sample)                    sd(x) / mean(x) * 100
Python (sample)               np.std(x, ddof=1) / np.mean(x) * 100
Python (population)           np.std(x) / np.mean(x) * 100
SciPy                         scipy.stats.variation(x, ddof=1) * 100

Format the Excel cell as a percentage to see the result as a percent. Note that scipy.stats.variation uses ddof=0 by default, which is the population version, and that it returns a fraction rather than a percentage. For the store A data, the Excel formula gives 0.1239, which is 12.39%, and SciPy with ddof=1 gives the same figure.

The calculator above prints the exact formulas for your own data in the “Same calculation in other tools” box.

What Mistakes Cause a Wrong Coefficient of Variation?

• Forgetting to multiply by 100. A CV of 0.124 and a CV of 12.4% are the same thing. Check which form your answer needs.
• Mixing sample and population. Taking the mean from one source and a population SD from another gives a figure that matches neither definition.
• Using it on interval data. Celsius, Fahrenheit and many index scores have no true zero, so the CV is not meaningful.
• Using it when the mean is near zero. The result swings wildly with tiny changes in the data.
• Comparing CVs from very small samples as if they were exact. With five or six values, the CV has wide uncertainty.
• Treating a low CV as proof of accuracy. It shows consistency only.
• Dividing by the variance instead of the SD. The CV uses the standard deviation, which has the same units as the mean.

How Does the Coefficient of Variation Compare With Other Measures of Spread?

The standard deviation and variance describe spread in absolute terms. The CV describes it relative to the mean. The IQR and the mean absolute deviation are two other options with different strengths.

If your data contain outliers, the mean and standard deviation, and therefore the CV, will be pulled around by them. In that case the five number summary calculator gives a more robust picture of spread, and the mean absolute deviation calculator offers a second, more intuitive spread measure. When some observations should count more than others, work out the mean first with the weighted mean calculator. For data that really are bell-shaped, the normal curve calculator turns a mean and standard deviation into percentiles.

Where Is the Coefficient of Variation Used in Practice?

The CV turns up wherever people care about consistency rather than size. Four common settings show how the same ratio gets used.

• Quality control and manufacturing. A machine filling 500 g bags and one filling 5 kg sacks can be compared fairly by their CVs. The one with the lower CV fills more consistently for its size.
• Laboratory science. Replicate measurements are summarized with %RSD to show how well an instrument or method repeats, as in Example 3.
• Finance. Dividing the standard deviation of returns by the average return gives risk per unit of return. It helps when two investments have very different average returns, though it fails when an average return is near zero or negative.
• Biology and agriculture. Researchers compare variability in traits of different scales, such as the body mass and the wing length of the same species, or crop yields across fields with different soils.

In each case the logic is identical: remove the units, keep the proportion, and compare. The cautions from the section on misleading CVs apply in all of them. A ratio-scale measurement and a mean well away from zero are the two conditions that make the number trustworthy.

Frequently Asked Questions

Sources: Sokal, R. R. and Rohlf, F. J. (1995), Biometry, 3rd ed.; NumPy documentation for numpy.std; SciPy documentation for scipy.stats.variation; Microsoft documentation for STDEV.S and STDEV.P.

Related Calculators

Want another view of spread or center? Try these, or browse the full list of free statistics calculators:

Five Number Summary CalculatorQuartiles, IQR, outlier fences and a box plot. Mean Absolute Deviation CalculatorAverage distance from the mean or median, with a step table. Weighted Mean CalculatorFind the center when some values count more than others. All StatsSolve CalculatorsBrowse the full directory of free statistics tools.