Coefficient of Variation Calculator
Calculate the coefficient of variation (CV) from mean and standard deviation, and classify the relative variability as low, moderate, or high.
Inputs
Results
What is the coefficient of variation?
The coefficient of variation (CV) is a dimensionless measure of relative spread: the ratio of a dataset's standard deviation to its mean. Because it normalizes spread by the mean, it expresses variability independent of the unit of measurement, which makes it useful for comparing the variability of datasets on different scales.
Definition
The coefficient of variation, also called the relative standard deviation (RSD), expresses how large the standard deviation is relative to the mean:
CV=μσ,CV%=μσ×100%- σ — standard deviation (sample s or population σ)
- μ — mean (arithmetic average)
Because the mean is in the denominator, CV cancels out the unit of measurement. A dataset of masses in grams and one in kilograms produce the same CV if their relative spread is identical. This unit-independence is the key advantage of CV over plain standard deviation.
CV versus standard deviation
Standard deviation is the right tool for absolute spread in known units — for example, tolerances on a machined part specified in millimeters. CV applies to relative spread, where comparisons must survive a change of scale:
- Comparing two assays — a blood glucose test (mean 5 mmol/L, SD 0.3) and a cholesterol test (mean 5.2 mmol/L, SD 0.8) have very different SDs, so CV is needed to compare their analytical precision fairly.
- Finance — comparing the risk-adjusted return of two assets with different price ranges: a $10 stock with SD $2 has the same CV as a $1,000 stock with SD $200, even though their SDs differ by a factor of 100.
- Quality control — process capability assessments require knowing whether variability is small relative to the target value, not just in absolute terms.
When the mean is near zero or negative, standard deviation or variance is the safer choice, because CV loses its intuitive meaning in those cases.
The non-zero mean requirement
CV is undefined when μ = 0, because the mean sits in the denominator and division by zero has no result. More fundamentally, a dataset centered at zero has no meaningful relative center, so the question "how large is the spread relative to the mean?" has no sensible answer. For data that can legitimately average to zero (e.g., daily return differences, temperature anomalies), standard deviation or variance applies instead.
Interpreting the result
No universal threshold exists, but the following rough guideline is widely used across science and industry:
| CV (%) | Interpretation | Typical context |
|---|---|---|
| < 10% | Low variability | Precision analytical methods, tight manufacturing |
| 10–30% | Moderate variability | Clinical labs, social-science surveys |
| > 30% | High variability | Ecological counts, financial returns, heterogeneous populations |
CV should always be interpreted within the context of its field. A CV of 5% is excellent for a clinical assay but may be unacceptably high for a reference standard. A CV of 80% is alarming in pharmaceutical batch testing but entirely normal for wildlife population counts.
Worked example
A quality-control lab measures the moisture content of 30 biscuit samples:
- Mean moisture content: μ = 3.8%
- Standard deviation: σ = 0.57%
A CV of 15% falls in the moderate variability range. In food manufacturing, typical in-process moisture CVs are 5–20%, so this batch is within the expected range but not exceptionally tight. The lab might investigate whether the high end of the spread corresponds to samples from a specific production time, suggesting an equipment drift.
Negative coefficients of variation
Mathematically, CV = σ / μ is negative whenever the mean is negative. Standard deviation is always non-negative, so a negative CV signals a negative mean, not a physically impossible variance. In practice, most fields that use CV restrict it to positive-mean datasets. When a negative mean is unavoidable, the common convention is to report |CV| and state the sign of the mean explicitly.
Applications across fields
Analytical chemistry and clinical labs — Inter-laboratory comparisons, method validation, and proficiency testing all express reproducibility as CV (or RSD). Regulatory guidelines from organizations such as ISO 17511 and the Clinical and Laboratory Standards Institute (CLSI) specify acceptable CV limits by analyte concentration level.
Finance and investment — CV equals the ratio of an asset's standard deviation to its expected return, making it a measure of risk per unit of return. An asset with a lower CV offers more return for each unit of risk, all else equal. However, negative expected returns make CV meaningless in this context.
Manufacturing and process control — In Six Sigma and statistical process control, CV (as a ratio) is related to process capability indices. A CV below roughly 17% (σ < μ/6) is a rough proxy for a Six Sigma capable process.
Biology and ecology — Body measurements within a species, species abundance across sampling sites, and gene expression levels all carry CVs. In morphometric studies, CVs below 10% typically indicate a homogeneous sample; CVs above 50% suggest high intra-group diversity.
Frequently Asked Questions (FAQ)
When is CV more useful than standard deviation?
Standard deviation reports spread in the original units (e.g., grams, dollars), which is only meaningful when comparing datasets measured in the same unit at similar magnitudes. CV normalizes spread by the mean, so it is unit-free and scale-free.
CV is the better choice when comparing variability across different measurements (e.g., protein content in mg vs. caloric content in kcal), when one dataset has a much larger mean than another (e.g., salaries vs. age), or when assessing relative precision in repeated experiments. When the mean is near zero or negative, standard deviation is safer, because CV loses its intuitive meaning in those cases.
Why must the mean be non-zero?
CV is defined as σ / μ. When μ = 0, the denominator is zero and the ratio is undefined.
Conceptually, a dataset centered at zero has no meaningful "relative" variation: half the values are positive and half are negative, so asking "how large is the spread relative to the center?" has no sensible answer. For data that can legitimately have a zero mean (e.g., returns on a zero-drift portfolio), standard deviation or variance applies instead of CV.
Can the coefficient of variation be negative?
Yes, technically. If the mean is negative and the standard deviation is positive, CV = σ / μ will be negative. However, a negative CV has no standard interpretation because standard deviation is always non-negative. This situation arises when the data are measurements of a naturally negative quantity (e.g., temperature in °C below zero, net losses).
In practice, most fields that use CV — analytical chemistry, clinical labs, finance — restrict it to datasets with a positive mean. For a negative mean, the common convention is to report the absolute value |CV| and note the sign of the mean explicitly.
What counts as a 'high' coefficient of variation?
There is no universal threshold, but a common guideline is: CV < 10% is low (tight, reproducible data), 10–30% is moderate (acceptable in many fields), and CV > 30% is high (heterogeneous or imprecise data).
In analytical chemistry, method validation typically requires CV < 5–15% depending on concentration. In clinical labs, acceptable inter-assay CVs range from 5% to 20%. In finance, a CV > 100% on an investment return indicates extreme relative risk. CV should always be interpreted in the context of its field — what is "high" in precision manufacturing (>2%) is "low" in ecology (>50%).
Recommended Next
Variance and Standard Deviation Calculator
Calculate variance and standard deviation from a comma-separated list of numbers. Switch between sample and population formulas.