Paste your replicate results to get the mean, standard deviation and %RSD instantly, with a pass/fail check against your acceptance limit. The calculator uses the sample standard deviation (n − 1), the same as USP, ICH and Excel’s STDEV.S. It shows every step, so you can check it against your worksheet.
- n
- –
- Mean
- –
- Standard deviation
- –
- Minimum
- –
- Maximum
- –
- Range
- –
Show the calculation step by step
What is relative standard deviation (RSD)?
Relative standard deviation (RSD) expresses the standard deviation as a percentage of the mean. Because it is relative, you can compare the precision of results with different units or magnitudes. A %RSD of 0.5% means the same thing whether the results are peak areas in the millions or assay values near 100%. RSD is also called the coefficient of variation (CV). When it is multiplied by 100 and reported as a percentage, it is often written as %RSD.
In pharmaceutical QC, %RSD is the standard way to report precision: replicate injections in system suitability, repeatability and intermediate precision in method validation, content uniformity, and dissolution profiles.
RSD formula
%RSD = (s ÷ x̄) × 100
where the sample standard deviation is
s = √[ Σ(xi − x̄)² ÷ (n − 1) ]
xi = each individual result · x̄ = mean of the results · n = number of results
How to calculate RSD, step by step
Six assay results (% of label claim) from a repeatability study: 99.6, 100.4, 99.1, 100.8, 99.9 and 100.2.
- Calculate the mean. Add the results and divide by n: 600.0 ÷ 6 = 100.0.
- Find each deviation from the mean (x − x̄) and square it.
- Add the squared deviations. The sum is 1.82.
- Divide by n − 1 to get the variance: 1.82 ÷ 5 = 0.364.
- Take the square root to get the standard deviation: √0.364 = 0.603.
- Divide by the mean and multiply by 100: 0.603 ÷ 100.0 × 100 = 0.60%.
| Result | x | x − x̄ | (x − x̄)² |
|---|---|---|---|
| 1 | 99.6 | −0.4 | 0.16 |
| 2 | 100.4 | 0.4 | 0.16 |
| 3 | 99.1 | −0.9 | 0.81 |
| 4 | 100.8 | 0.8 | 0.64 |
| 5 | 99.9 | −0.1 | 0.01 |
| 6 | 100.2 | 0.2 | 0.04 |
| Sum of squares | 1.82 | ||
The result of 0.60% meets the typical repeatability criterion for an assay of not more than 2.0%.
Example 2: system suitability (six replicate injections)
Peak areas of a standard: 1,254,310; 1,251,872; 1,256,904; 1,249,566; 1,253,118; 1,255,430. The mean is 1,253,533, the SD is 2,617 and %RSD = 0.21%. Load “Example: peak areas” in the calculator above to see every step.
How to calculate RSD in Excel
If your results are in cells B2 to B7, use:
=STDEV.S(B2:B7)/AVERAGE(B2:B7)*100
Use STDEV.S (sample), not STDEV.P (population). Google Sheets and LibreOffice use the same function names. Format the cell to two decimal places and add the “%” label yourself, because the formula already multiplies by 100.
Sample vs population standard deviation: why n − 1 matters
Laboratory replicates are a sample of all possible measurements, so the sample formula (dividing by n − 1) is used. Dividing by n underestimates the spread, especially with the small n typical in QC. For the six assay results above, the sample SD gives 0.60%, but the population SD gives only 0.55%. With three replicates the difference is even larger. Pharmacopoeial and ICH calculations use n − 1.
RSD acceptance limits in pharmaceutical analysis
ICH Q2(R2) asks you to report precision as SD, %RSD and confidence interval, but it does not set numerical limits. Limits come from the pharmacopoeial monograph, your validated method or your SOP. The values below are typical criteria used in the industry. Always follow the requirement that applies to your method.
| Application | Typical %RSD criterion | Notes |
|---|---|---|
| HPLC/GC system suitability, replicate standard injections (assay) | NMT 2.0% | USP <621> uses 5 replicate injections when the requirement is 2.0% or less, and 6 when it is more than 2.0%, unless the monograph says otherwise. Ph. Eur. 2.2.46 and harmonized USP <621> also allow a maximum RSD calculated from the assay limits and the number of injections. |
| Method validation: repeatability, assay | NMT 2.0% | ICH Q2: at least 6 determinations at 100%, or 9 across the range (3 levels × 3 replicates). |
| Method validation: intermediate precision, assay | NMT 2.0% (some labs allow up to 3.0% for pooled data) | Different day, analyst and/or instrument. |
| Related substances at the specification level | About NMT 10% | Wider criteria are often justified near the quantitation limit. |
| Dissolution profile comparison (f2) | NMT 20% at early time points; NMT 10% at later points | FDA dissolution guidance for immediate-release dosage forms. |
| Bioanalytical methods (ICH M10) | NMT 15% (NMT 20% at the LLOQ) | Expressed as %CV for QC samples. |
When a system suitability %RSD fails, do not report sample results. Investigate the system first. Common causes are injector or autosampler problems, air in the pump, leaks, poor integration and unstable standard solutions. See the HPLC troubleshooting guide.
Common mistakes
- Using the population SD (STDEV.P or dividing by n). This gives a lower, non-compliant %RSD.
- Rounding too early. Keep full precision until the final %RSD, then round as your SOP requires.
- Removing “outliers” without justification. Excluding a result needs an assignable cause or a predefined statistical test (see USP <1010>). Unexplained exclusion is a data integrity issue.
- Mixing results from different conditions. Repeatability uses one analyst, one instrument and a short time interval. Keep intermediate precision data separate.
- Using %RSD for values near zero. When the mean is close to zero (for example blank responses or very low impurity levels), %RSD becomes huge and meaningless. Report the SD instead.
Frequently asked questions
What is a good %RSD value?
It depends on the test. For an HPLC assay, a %RSD below 1% for replicate injections is good and below 2.0% usually meets requirements. Trace-level impurity and microbiological methods naturally show higher %RSD.
What is the difference between SD and RSD?
The standard deviation has the same units as the results, for example mg or peak area. RSD divides the SD by the mean, giving a unitless percentage that can be compared across methods and concentration levels.
Is RSD the same as the coefficient of variation (CV)?
Yes. RSD and CV are the same quantity. %RSD and %CV are the same value expressed as a percentage.
How many replicates do I need to calculate RSD?
Mathematically, at least two. In practice, use what the method requires: commonly 5 or 6 replicate injections for system suitability and at least 6 determinations for repeatability.
Can %RSD be negative?
No. The standard deviation is always positive, and %RSD is reported using the absolute value of the mean.
Related tools and guides
- Method validation guide (ICH Q2): all parameters explained
- HPLC troubleshooting guide
- OOS procedure in pharma
- Pharmaceutical Lab Toolkit: all calculators
References
- ICH Q2(R2) Validation of Analytical Procedures.
- USP General Chapter <621> Chromatography.
- USP General Chapter <1010> Analytical Data: Interpretation and Treatment.
- Ph. Eur. 2.2.46 Chromatographic Separation Techniques.
- FDA Guidance for Industry: Dissolution Testing of Immediate Release Solid Oral Dosage Forms (1997).
- ICH M10 Bioanalytical Method Validation and Study Sample Analysis.