Coefficient of Variation Calculator

The coefficient of variation (CV) expresses the standard deviation as a fraction of the mean, turning spread into a unitless, scale-free number. That is what makes it the right tool for comparing variability between things measured on completely different scales — a $100 stock against a $20 stock, or reaction times against test scores.

Population or Sample Standard Deviation?

Sample CV (SD divides by n − 1): your data is a subset standing in for something larger — the usual case for measurements, assays, and returns.

Population CV (SD divides by N): your data is the complete group of interest, with nothing left unsampled.

The mean is identical either way; only the standard deviation in the numerator changes, and with it the CV.

Before You Rely on the CV

  • The CV assumes a meaningful zero (a ratio scale). Weights, prices, concentrations, and durations qualify; Celsius temperatures do not — details below.
  • All-positive data is the safe zone. Values straddling zero push the mean toward zero and can make the CV explode or flip sign.
  • Enter values separated by commas, spaces, or line breaks, all in the same unit.

Enter numbers separated by commas or spaces

The CV Formula

The coefficient of variation is a ratio of two quantities you may already know, and it inherits its two variants from the standard deviation in its numerator:

Sample: CV = s / x̄

Population: CV = σ / μ

As a percentage: CV% = CV × 100

Where:

  • s = sample standard deviation (n − 1 divisor), σ = population standard deviation (N divisor)
  • x̄ or μ = the mean of the same data

Because standard deviation and mean carry the same units, the units cancel: a CV of 0.15 means the typical deviation is 15% of the mean, whether the data was measured in dollars, grams, or milliseconds. Chemists usually call the same percentage the relative standard deviation (RSD). The numerator comes straight from the standard deviation calculator, and the n − 1 logic behind the sample version is unpacked on the variance calculator page.

Comparing Variability Across Different Scales

Standard deviations from different scales cannot be compared directly — a $1.41 daily swing is trivial for a $100 stock and violent for a $20 one, even though the SD is identical. Dividing by each mean converts both spreads to a common, relative footing. Typical uses:

  • Investment risk: the CV of returns approximates risk per unit of average return, letting a high-priced and a low-priced asset be ranked on one axis.
  • Laboratory precision: assay repeatability is quoted as CV. A common benchmark for immunoassays is an intra-assay CV under 10% and an inter-assay CV under 15%; higher values signal a method in need of troubleshooting.
  • Biology and manufacturing: comparing weight variability of mice against elephants, or fill-volume consistency across bottling lines running different container sizes, only works after scaling out the means.

CV vs Standard Deviation: Which to Report

The standard deviation answers "how much do values typically deviate, in original units?" — the right quantity when the units themselves matter, as in tolerance limits (±0.2 mm) or the spread of exam scores in points. The CV answers "how large is that deviation relative to the typical value?" — the right quantity when comparing groups with different means or different units entirely.

A useful habit: report the SD when comparisons stay within one scale, and add the CV whenever means differ substantially between the groups being compared. Note the direction of the trade: the CV gains comparability but loses the units, so it can no longer be turned back into a concrete interval around the mean without multiplying the mean back in.

When the CV Is Meaningless

The CV has sharp validity limits, all traceable to the mean in its denominator:

  • Mean near zero: the ratio explodes toward infinity even when absolute spread is tiny. Daily returns averaging 0.01% with an SD of 1% give a CV of 10,000% — a number that describes the denominator, not the variability.
  • Interval scales: Celsius and Fahrenheit have arbitrary zero points, so the CV changes with the unit conversion. Temperatures of 10 °C and 20 °C have CV ≈ 33%, but the same temperatures written as 50 °F and 68 °F give CV ≈ 15% — same physical reality, different answer, hence no answer at all. (Kelvin, with a true zero, is fine.)
  • Data straddling zero: positive and negative values pull the mean toward zero and can make the CV negative or arbitrarily huge; profit/loss figures and z-scored data are common offenders.

In these situations, report the standard deviation, the IQR, or a domain-specific measure instead.

Worked Example: Two Stocks, Same Swing, Different Risk

Four daily closing prices for two stocks. Stock A: 98, 100, 102, 100. Stock B: 18, 20, 22, 20. Treat each as a complete population of the four days:

  1. Means: Stock A averages 400 ÷ 4 = 100; Stock B averages 80 ÷ 4 = 20.
  2. Deviations are identical for both: −2, 0, +2, 0, so each has a squared-deviation sum of 8 and a population SD of √(8 ÷ 4) = √2 ≈ 1.4142.
  3. Stock A: CV = 1.4142 ÷ 100 = 0.0141, or 1.4142%.
  4. Stock B: CV = 1.4142 ÷ 20 = 0.0707, or 7.0711%.

Identical standard deviations, but Stock B is five times more variable relative to its price — a $2 move means far more at $20 than at $100. Choosing the sample toggle instead would use s = √(8 ÷ 3) ≈ 1.6330 for both stocks, giving CVs of 1.633% and 8.165%; the fivefold ratio between the stocks is unchanged, because the divisor affects both equally.

Frequently Asked Questions

What is a good coefficient of variation?

It depends entirely on the field. Laboratory assays often target an intra-assay CV under 10%, precision manufacturing may demand under 1%, while biological field measurements of 20-30% can be perfectly normal. The CV is most useful comparatively: an assay with CV 6% is more repeatable than one with 12%, whatever the absolute standard is.

Why can't I use the CV with temperatures in Celsius?

Because 0 degrees C is an arbitrary point, not a true absence of temperature. Converting the same temperatures to Fahrenheit shifts the zero and changes every CV, so the statistic depends on the unit rather than the data. The CV requires a ratio scale with a meaningful zero - Kelvin works, as do weights, prices, lengths, and durations.

Can the CV exceed 100%?

Yes, whenever the standard deviation is larger than the mean, which is common in strongly right-skewed data such as incomes, insurance claims, or rainfall. A CV above 100% is a signal to look at the distribution's shape: the mean may be a poor summary of such data, and median-based measures may describe it better.

Should I choose the population or sample version?

Choose sample (n - 1) when your values are a subset representing something larger - repeated measurements of an assay, a month of stock prices standing in for the asset's behavior. Choose population (N) only when the data is the entire group of interest. The sample CV is always slightly larger; with four values the difference is about 15%, so the choice genuinely matters on small data sets.

Is the coefficient of variation the same as relative standard deviation (RSD)?

Essentially yes. RSD is the term preferred in analytical chemistry and is conventionally reported as a positive percentage - the absolute value of the CV times 100. This calculator's percentage output equals the RSD whenever the mean is positive, which is the only situation where either quantity should be used anyway.