What is a Relative Standard Deviation Calculator?
A relative standard deviation calculator expresses the spread of replicate measurements as a percentage of their mean: %RSD = 100 × s / |x̄|. Because it has no units, %RSD is the standard way to state the precision (repeatability) of an analytical method and to compare it with an acceptance limit. Enter replicates or a mean and SD, and optionally a limit, to get %RSD with steps, the standard error and a pass/fail verdict.
What the Relative Standard Deviation Calculator does
Paste the replicate results for one sample, such as six injection peak areas or five weighings, and the calculator returns the mean, the standard deviation, the %RSD and the standard error of the mean. If you already have a mean and SD from an instrument report, switch to Mean and SD. Enter an Acceptance limit (%RSD) and the result is checked against it, with the largest standard deviation that would still pass.
It works as an RSD calculator, a %RSD calculator or a percent relative standard deviation calculator: all three names describe the same number. You may also see it written as relative stdev, relative standard deviation percentage or percentage RSD. The result is shown both as a percentage and as a plain ratio (s / |x̄|), and the step-by-step solution shows every line of the arithmetic so you can copy it into a lab notebook.
RSD is the same quantity as the coefficient of variation, reported as a positive percentage. This page is built for the lab use of that number: judging repeatability against a criterion, seeing which replicate is furthest from the mean, and knowing how far an RSD from only a handful of replicates can be trusted. For comparing the relative spread of two different datasets in general, the coefficient of variation page is the better fit.
How to use the Relative Standard Deviation Calculator
- Choose Replicate measurements to paste raw results, or Mean and SD to enter summary values.
- For raw data, leave Standard deviation on Sample (n − 1) unless your procedure specifies the population formula.
- For summary values, enter the Mean, the Standard deviation and, if you know it, the Number of replicates, n.
- Optionally enter the Acceptance limit (%RSD) from your method, for example 2 for “not more than 2.0%”.
- Read the %RSD and the verdict, then check the replicate table for any single result that is driving the spread.
RSD formula
The relative standard deviation formula divides the standard deviation by the absolute value of the mean and multiplies by 100. Use the sample standard deviation (n − 1) for replicates, which is the usual case, or the population standard deviation (n) when the values are the whole population.
| Symbol | Meaning |
|---|---|
| xᵢ | individual replicate results |
| x̄ | mean of the replicates |
| s | sample standard deviation of the replicates |
| μ, σ | population mean and population standard deviation |
| n | number of replicates |
| χ²₀.₀₅, ₙ₋₁ | 5th percentile of the chi-square distribution with n − 1 degrees of freedom |
The absolute value of the mean keeps %RSD positive. That is the only difference from the coefficient of variation, which keeps the sign of the mean.
The upper bound treats the mean as known and assumes roughly normal measurement error. It is a good approximation when the RSD is small: with three or more replicates and an observed RSD of 10% or less it is within about 3% of the exact (noncentral t) bound, which covers nearly all method-precision work. For larger RSDs it runs low, by about 4% at an RSD of 20% with six replicates.
Worked example
Worked example: six replicate injections
Problem. A system suitability test requires the peak areas of six replicate injections to have an RSD of not more than 2.0%. The areas are 1254.3, 1261.8, 1249.6, 1258.2, 1266.1, 1252.7. Does the run pass?
- 01Mean: x̄ = 7542.7 / 6 = 1257.1167.
- 02Sum of squared deviations: Σ(x − x̄)² = 187.7483, so s = √(187.7483 / 5) = 6.1278.
- 03Ratio: s / x̄ = 6.1278 / 1257.1167 = 0.004874.
- 04%RSD = 100 × 0.004874 = 0.487%, which is below the 2.0% limit.
- 05Upper bound: with 5 degrees of freedom, χ²(0.05) = 1.1455, so the factor is √(5 / 1.1455) = 2.0893 and the 95% upper bound for the true RSD is 0.487% × 2.0893 = 1.02%.
Answer. The injections pass: %RSD = 0.49% against a 2.0% limit. Even allowing for only six replicates, the true RSD is unlikely to exceed about 1.0%, so the pass is comfortable.
RSD versus coefficient of variation
They are the same statistic: the standard deviation divided by the mean. Chemists and quality-control laboratories call it relative standard deviation and almost always quote it as a percentage (%RSD). Statisticians, biologists and finance writers call it the coefficient of variation (CV) and may quote it as a ratio or a percentage.
The one technical difference is the sign. RSD divides by the absolute value of the mean, so it is never negative; the CV keeps the sign of the mean. For concentrations, masses, areas and other positive measurements the two are identical. The Coefficient of Variation Calculator covers the comparison-across-scales use of the same number.
| Relative standard deviation | Coefficient of variation | |
|---|---|---|
| Formula | 100 × s / |x̄| | s / x̄ (often × 100) |
| Usual form | Percentage, always positive | Ratio or percentage, sign of the mean |
| Where the name is used | Analytical chemistry, pharma QC, method validation | Statistics, biology, clinical labs, finance |
| Typical question | Is the method precise enough to pass? | Which dataset is relatively more variable? |
Relative standard deviation vs standard deviation
The standard deviation is an absolute measure of spread, in the same units as the data: milligrams, peak-area counts, seconds. The relative standard deviation divides that spread by the mean, so the units cancel and the result is a percentage of the typical value.
That makes RSD the better choice for comparing precision across different concentrations, instruments or units. A standard deviation of 2 mg means a lot for a 10 mg tablet (RSD 20%) and very little for a 1,000 mg tablet (RSD 0.2%). Use the standard deviation when you need the spread in real units, for example to set a tolerance, and RSD when you need to judge it against a percentage limit.
| Standard deviation | Relative standard deviation | |
|---|---|---|
| Formula | s = √[ Σ(x − x̄)² / (n − 1) ] | %RSD = 100 × s / |x̄| |
| Units | Same as the data | None (percentage) |
| Compares datasets with different means? | No | Yes |
| Behaviour near a zero mean | Stable | Unstable, can blow up |
Sample vs population %RSD: the same data both ways
The six injection areas from the worked example give two slightly different answers depending on which standard deviation you use. The sample formula divides the sum of squared deviations (187.7483) by n − 1 = 5; the population formula divides it by n = 6. The calculator shows the other version in its statistics list so you can check which one a report used.
For replicate measurements, use the sample version. The population version is only right when the values are every member of the group you care about, not a sample of results the method could produce.
| Version | Standard deviation | %RSD | Excel |
|---|---|---|---|
| Sample (n − 1) | s = √(187.7483 / 5) = 6.1278 | 100 × 6.1278 / 1257.1167 = 0.487% | STDEV.S |
| Population (n) | σ = √(187.7483 / 6) = 5.5939 | 100 × 5.5939 / 1257.1167 = 0.445% | STDEV.P |
What %RSD is acceptable?
There is no universal number. The acceptable RSD depends on the concentration being measured, the technique and the purpose of the result, and the binding limit is always the one in your own method or SOP. As a guide, instrument repeatability for a main-component chromatographic assay is commonly held to about 1–2%, while bioanalytical methods measuring drugs in plasma typically allow up to 15%, and 20% at the lower limit of quantitation.
Typical ranges by field: pharmaceutical assay and system suitability about 2% or less; dissolution and content uniformity a few percent; environmental and food trace analysis 5–20% depending on level; bioanalysis (drug in plasma) up to 15%; immunoassays and cell-based assays often 10–20%. Outside the lab, an RSD under about 10% is usually read as low relative variability, but that rule of thumb has no formal basis.
Sources for these ranges: ICH Q2(R2), Validation of Analytical Procedures for repeatability and intermediate precision, the FDA M10 bioanalytical method validation guidance for the 15% and 20% limits, and Horwitz and Albert, J AOAC Int. 2006;89(4):1095–1109, for the Horwitz curve below.
Precision worsens as concentration falls. The Horwitz equation, an empirical summary of thousands of inter-laboratory studies, predicts the between-laboratory RSD as 2^(1 − 0.5·log₁₀C)%, where C is the concentration as a mass fraction. Repeatability within one laboratory is usually one half to two thirds of these values.
| Analyte concentration | Mass fraction, C | Horwitz predicted between-lab %RSD |
|---|---|---|
| 100% (pure substance) | 1 | 2.0% |
| 10% | 0.1 | 2.8% |
| 1% | 0.01 | 4.0% |
| 0.1% | 0.001 | 5.7% |
| 100 ppm | 0.0001 | 8.0% |
| 1 ppm | 0.000001 | 16% |
| 1 ppb | 0.000000001 | 45% |
How far to trust an RSD from a few replicates
A standard deviation estimated from a handful of results is itself imprecise, and so is the RSD built on it. The table shows how much larger the true RSD could plausibly be than the one you observed (95% one-sided bound, normal errors). With three replicates an observed 1% is consistent with a true value above 4%; with ten it is pinned below about 1.6%.
| Replicates, n | Degrees of freedom | Upper-bound factor | Observed 1.0% RSD could be as high as |
|---|---|---|---|
| 3 | 2 | 4.42 | 4.4% |
| 5 | 4 | 2.37 | 2.4% |
| 6 | 5 | 2.09 | 2.1% |
| 10 | 9 | 1.65 | 1.6% |
| 20 | 19 | 1.37 | 1.4% |
| 30 | 29 | 1.28 | 1.3% |
How to calculate RSD in Excel, Google Sheets, R and Python
No spreadsheet has a built-in RSD function, so the relative standard deviation formula in Excel is built from STDEV.S and AVERAGE. With the replicates in cells A2:A7, the formulas below return the same 0.487% as this calculator for the worked example. Format the Excel cell as a number (the formula already multiplies by 100), or drop the × 100 and format it as a percentage.
For the population version use STDEV.P in Excel and Google Sheets, sd(x) * sqrt((n − 1) / n) in R, ddof=0 in NumPy and σx on the TI-84.
| Tool | Formula for %RSD (sample SD) |
|---|---|
| Excel | =STDEV.S(A2:A7)/ABS(AVERAGE(A2:A7))*100 |
| Google Sheets | =STDEV(A2:A7)/ABS(AVERAGE(A2:A7))*100 |
| R | 100 * sd(x) / abs(mean(x)) |
| Python (NumPy) | 100 * np.std(x, ddof=1) / abs(np.mean(x)) |
| TI-84 | STAT → CALC → 1-Var Stats, then 100 × Sx / x̄ |
Common use cases
- System suitability: check that replicate injections of a standard meet the method’s %RSD limit before running samples.
- Method validation: report repeatability and intermediate precision as %RSD at each concentration level.
- Balance, pipette and dispenser checks: express the scatter of repeated weighings or deliveries relative to the nominal amount.
- Content uniformity and batch consistency: summarise unit-to-unit variation in a batch as a single percentage.
- Comparing instruments or analysts: a lower %RSD on the same sample means better precision, whatever the units.
Assumptions and limitations
- RSD measures precision, not accuracy. Replicates can agree closely with each other and still all be wrong; check trueness with a reference standard or recovery experiment.
- It is unstable when the mean is near zero, for example close to a detection limit or for blank-corrected signals, and it is meaningless for scales without a true zero such as °C or pH. Report the standard deviation instead.
- One outlying replicate inflates the SD and therefore the RSD. Investigate it before deleting it; do not drop results only to pass a limit.
- With few replicates the RSD is a rough estimate. Use the upper confidence bound, or the confidence interval for a standard deviation, before drawing conclusions about long-run precision.
- RSD describes scatter of individual results. The uncertainty of their mean is the standard error, which is smaller by a factor of √n.
Frequently asked questions
How do I calculate relative standard deviation?
What is RSD in statistics?
How can I calculate RSD online?
Is RSD the same as the coefficient of variation?
What is a good RSD%?
Should %RSD use the sample or the population standard deviation?
Can relative standard deviation be negative?
How do I calculate RSD in Excel?
How many replicates do I need for a reliable RSD?
Cite this page
Statculator. (2026, November 14). Relative Standard Deviation Calculator. Retrieved November 14, 2026, from https://statculator.com/descriptive-statistics/relative-standard-deviation-calculator/