• Understand the basic idea of the coefficient of variation and how to compute it
• Explain why the CV is dimensionless and what advantage that gives
• Compute and interpret the CV for simple data sets
Deeper study
• Use the CV to compare the variability of data with different units or magnitudes
• Apply the common rule-of-thumb thresholds for interpreting the CV
• Apply the CV in practical settings such as investing and quality control
• Recognize the caveats and limits of using the CV
Real-world practice
• Understand and apply advanced variants such as the robust CV
• Make CV-informed decisions in complex settings such as A/B testing and quality control
• Explain the theoretical background of the CV and its relationship to other statistical measures
• Evaluate appropriate uses of the CV in research and practice
At a glance
The coefficient of variation (CV) is a relative measure of spread equal to the standard deviation divided by the mean; because the units cancel, it is a dimensionless ratio (CV = σ/μ, usually reported as a percentage). The term is defined in the international standard ISO 3534-1, and analytical chemistry also calls it the relative standard deviation (RSD).
Use the coefficient of variation to compare the spread of data measured in different units. A standard deviation is locked to its unit — kg versus g, or won versus dollars, cannot be compared directly — but the CV cancels the units and puts everything on one scale: "how many percent of the mean does the data wobble?"
A measured example from the NumLM CV simulator: for elephants (mean 4,000 kg, σ = 400 kg) and mice (mean 20 g, σ = 5 g), the elephants' standard deviation is 80,000 times larger, yet the CV verdict flips — elephants 10% < mice 25%, so mice are the "more variable" group. Drawing 300 actual samples reproduced the same reversal: standard deviations of 444 kg vs 4.9 g, CVs of 11.2% vs 24.8%.
The coefficient of variation is only meaningful for positive data on a ratio scale, where zero is an absolute reference point. If the mean is near zero the denominator shrinks and the value explodes, and on an interval scale such as Celsius temperature, where the position of zero is arbitrary, the phrase "relative to the mean" has no meaning at all.
The mean is the denominator of the CV, so you need to understand what the mean is
Standard Deviation
The CV is the standard deviation divided by the mean, so you need to know what the standard deviation is and how to compute it
Basic: What Is the Coefficient of Variation?
Difficulty 2/5
About 25 min
The coefficient of variation is the standard deviation divided by the mean. It is a dimensionless measure of how variable a data set is relative to its own average, which lets you compare the variability of data sets that have different units or very different magnitudes on fair terms.
Think of comparing heights with weights. If the students in Class A have a mean height of 170 cm with a standard deviation of 5 cm, their CV is 5/170 = 2.9%. If the students in Class B have a mean weight of 60 kg with a standard deviation of 4 kg, their CV is 4/60 = 6.7%. Even though centimetres and kilograms cannot be compared directly, the CVs tell you that Class B's weights vary relatively more than Class A's heights.
Key points
•
The ratio of the standard deviation to the mean
•
The numerator (σ) and denominator (μ) share the same unit, so the units cancel — the CV is dimensionless and unaffected by the unit of measurement
•
Lets you compare the variability of different data sets
•
Usually expressed as a percentage (%)
A simple example
Compare the math scores of two classes. Class 1: mean 80 points, standard deviation 8 points → CV = 8/80 = 10%. Class 2: mean 90 points, standard deviation 9 points → CV = 9/90 = 10%. The two classes have the same relative variability.
Check your understanding
Answer:
A larger CV means the standard deviation is large relative to the mean, so the data has high relative variability and low consistency.
Answer:
Because the numerator (the standard deviation) and the denominator (the mean) have the same unit, the unit cancels the moment you divide. That is why the CV of a data set is identical whether you measure it in kilograms or grams, and why data sets with different units can be compared fairly.
Answer:
When the mean is near zero the CV can become extremely large or unstable, so interpret it with caution. In that case consider a different measure of variability.
The coefficient of variation (CV), also called the relative standard deviation, is computed as CV = (σ/μ) × 100%. It expresses the standard deviation as a percentage of the mean. As a common rule of thumb, 10–20% is moderate variability and anything above 20% is high variability.
Mathematical formula
CV is the coefficient of variation, σ the standard deviation and μ the mean. For a sample, use s (the sample standard deviation) and x̄ (the sample mean) instead.
Learn through examples
Example 1: Comparing the Risk of Two Investments
Using the CV to compare the risk of two investment products
Stock A: mean return 12%, standard deviation 3% → CV = 3/12 = 25%. Stock B: mean return 8%, standard deviation 1.5% → CV = 1.5/8 = 18.75%
Answer:
Stock B carries less relative risk than Stock A.
Stock A delivers the higher return, but judged by the CV, Stock B is the more stable investment. The CV is useful when you want to evaluate relative stability rather than the absolute level of returns.
Example 2: Product Quality Control
Comparing how consistently two production lines meet their specifications
Line 1: mean weight 500 g, standard deviation 15 g → CV = 15/500 = 3%. Line 2: mean weight 200 g, standard deviation 8 g → CV = 8/200 = 4%
Answer:
Line 1 shows relatively more consistent quality than Line 2.
Looking only at the absolute standard deviations (15 g vs 8 g), Line 2 seems more consistent. Once you account for variability relative to the mean, Line 1 is the more stable line.
Example 3: Regional Income Inequality
Comparing the degree of inequality in regions with different income levels
Region A: mean income 40 million won, standard deviation 8 million won → CV = 8/40 = 20%. Region B: mean income 30 million won, standard deviation 4.5 million won → CV = 4.5/30 = 15%
Answer:
Income inequality is more severe in Region A than in Region B.
Region A has the higher mean income, but its larger CV shows greater income inequality. When shaping economic policy, look at how fairly income is distributed, not just at the average.
Common mistakes
Mistake:
Using the CV when the mean is negative or close to zero
Why is it wrong?
If the mean is negative or near zero, the CV becomes meaningless or blows up toward infinity
Correct approach:
Use the CV only for data whose mean is positive and comfortably far from zero
Mistake:
Looking only at absolute spread and ignoring relative variability
Why is it wrong?
The same standard deviation implies different relative variability when the means differ
Correct approach:
Use the coefficient of variation to compare relative variability
Mistake:
Comparing data in different units by standard deviation alone
Why is it wrong?
When the units differ, as with height (cm) and weight (kg), comparing standard deviations is meaningless
Correct approach:
Compare data with different units using the coefficient of variation
Practice questions
Hint:
Compare the ratio to the mean (CV = σ/μ), not the absolute standard deviation.
Answer:
Stock B (CV 18.75% < Stock A's CV of 25%)
Explanation:
A: CV = 3/12 = 25%; B: CV = 1.5/8 = 18.75%. Stock A has the larger standard deviation, but relative to its mean, Stock B is the more stable one. Keep in mind that CV comparisons work best when the two assets are similar in nature; a real investment decision should also weigh the absolute level of returns and the shape of their distribution.
Hint:
Think about what property the zero of a scale must have for the CV to make sense.
Answer:
No — Celsius temperature is an interval scale, and the CV cannot be used on it.
Explanation:
Celsius is an interval scale: 0 °C is an arbitrarily chosen reference point, not "no temperature". Convert the same readings to Fahrenheit or Kelvin and both the mean and the CV change completely, so the interpretation "percent of the mean" simply does not hold. The CV is only valid on ratio scales with an absolute zero, such as weight, length or price.
Hint:
A smaller standard deviation does not automatically mean more uniform — the means are different.
Answer:
Line 1 (CV 3% < Line 2's CV of 4%)
Explanation:
Line 1: CV = 15/500 = 3%; Line 2: CV = 8/200 = 4%. Judged by standard deviation alone, Line 2 (8 g) looks more uniform than Line 1 (15 g), but relative to the mean, Line 1 is more consistent. The CV is the fair yardstick when comparing the quality of products with different weight levels.
Case study: Using the CV in an A/B Test
Background
An online store ran an A/B test comparing the conversion rates of two checkout page designs, A and B.
The problem
Average conversion rate alone says little about how consistent each design is. The team needed to choose the design that performs reliably across many different user groups.
The data
Design A: mean conversion rate 5.2%, standard deviation 0.8% (CV = 15.4%). Design B: mean conversion rate 4.8%, standard deviation 1.2% (CV = 25.0%)
Method
The team computed the CV of each design to compare relative variability, then evaluated average performance and consistency together.
Solution and results
Design A not only has the higher mean conversion rate, its lower CV also shows more consistent performance. In other words, Design A delivers reliably good results across diverse user groups.
Design A achieved both the higher mean conversion rate (5.2% vs 4.8%) and the lower coefficient of variation (15.4% vs 25.0%), securing superior performance and consistency at the same time.
Conclusion:
CV analysis supports decisions that weigh consistency, not just average performance. Because predictable, stable results matter in marketing, Design A is the sensible choice.
References
ISO 3534-1:2006, Statistics — Vocabulary and symbols — Part 1: General statistical terms and terms used in probability — international standard definition of the coefficient of variation
