Sample Standard Deviation Calculator
Measure the spread of your data around the mean. Paste any list of numbers and get the sample standard deviation, variance, mean, and a full step-by-step breakdown — plus a live schematic, deviation chart, and companion spread metrics.
Standard deviation is in the same units as your data
This is the key advantage over variance. If your data is in centimetres, the standard deviation is in centimetres, so you can compare it directly to the mean. A standard deviation of 6.09 on a mean of 86.8 means a typical data point sits about 6 points away from the average. Variance, being squared, is in cm² and has no intuitive physical meaning — which is exactly why we take the square root at the end.
n − 1 corrects for using the sample mean
When you compute deviations from the sample mean rather than the true population mean, the deviations are systematically a little too small — the sample mean sits at the exact centre of your data by construction, so it minimises the sum of squared deviations. Dividing by n − 1 instead of n inflates the result just enough to make it an unbiased estimate of the population variance. This is Bessel's correction, and it matters most for small samples.
It is the natural measure of uncertainty
The standard deviation is the square root of the average squared distance from the mean. Because it squares before averaging, it punishes large deviations far more than small ones — a single far-off value can dominate the result. That is why it is the right tool when large errors matter more than small ones, and the wrong tool when you want every deviation to count equally (in which case use the mean absolute deviation instead).
It is sensitive to outliers
A single extreme value can inflate the standard deviation dramatically, which is why the interquartile range is often reported alongside it. Compare the two: if the IQR is small but the standard deviation is large, you have outliers. If they agree, the data is roughly symmetric. The Advanced panel runs a quick outlier sensitivity check, showing what happens to s when the largest or smallest value is removed.
How to Use This Calculator
The standard deviation formula turns a list of numbers into a single measure of spread. Paste your data, choose your denominator, and read the result.
1. Paste your data
Type or paste numbers into the box, separated by commas, spaces, tabs or new lines. Decimals, negative values and scientific notation all work. The chip bar below the box tells you how many values were detected and whether anything was skipped.
2. Choose sample or population
Use the toggle at the top. Sample (n − 1) is what you want when your data is a sample drawn from a larger population — the usual case. Population (N) is correct only when you genuinely have every single member of the group you care about.
3. Pick what to solve for
Standard deviation and variance are both reported for the dataset. The third option, missing value, is an inverse solver: give the calculator the rest of your data and a target standard deviation, and it finds the unknown number that produces exactly that spread.
4. Read the result and verdict
The result card shows the standard deviation in your data's own units, plus the variance, mean and sample size. A colour-coded badge interprets the coefficient of variation — a scale-free measure of how large the spread is relative to the average.
5. Check the schematic and charts
The schematic traces your data through every step of the formula, from the raw points to the final square root. The deviation chart shows each point's distance from the mean, and the distribution chart places the ±1s and ±2s bands over your data.
6. Switch to Advanced for depth
Advanced mode adds the full working table, an outlier sensitivity check showing how s changes when extreme values are removed, and companion statistics including median, quartiles, IQR, skewness, kurtosis and the standard error of the mean.
What This Tool Offers
Built around the sample standard deviation formula s = √( Σ(xᵢ − x̄)² / (n − 1) ), the standard measure of how far data spreads around its mean.
Sample and population
One toggle switches the denominator between n − 1 and N, with the formula strip and every step updating to match.
Missing value solver
Supply the rest of your data and a target standard deviation, and the quadratic solver finds the unknown value — including both symmetric roots when two answers exist.
Live formula schematic
A data strip with the mean line and ±1s band feeds into a six-stage pipeline showing every value as it flows through the formula.
Deviation chart
Every point's distance from the mean drawn as a bar, coloured by direction, so you can see exactly which values drive the spread.
Full working table
Advanced mode lays out every xᵢ, its deviation, and its squared deviation in a scrollable table with a live sum row.
Companion statistics
Mean, median, quartiles, IQR, range, mode, MAD, skewness, kurtosis, standard error, and the empirical-rule coverage check.
Reference Tables & Interpretations
How to read a standard deviation, when to use which formula, and the empirical rule.
Interpreting the Coefficient of Variation
The coefficient of variation CV = s / |x̄| × 100% expresses the spread as a percentage of the average. Unlike the standard deviation itself, it is scale-free, so it can be compared across datasets with different units.
| CV range | Interpretation | Typical example |
|---|---|---|
| CV < 10% | Low relative variability — the data is tightly clustered around the mean | Manufacturing tolerances, calibrated instrument readings |
| 10% – 25% | Moderate variability — typical for most natural measurements | Adult heights, exam scores, daily temperatures |
| 25% – 50% | High variability — the mean is only a loose summary | Household incomes, reaction times, rainfall totals |
| CV > 50% | Very high variability — the distribution is likely skewed or has heavy tails | Stock returns, insurance claims, city populations |
| x̄ ≈ 0 | CV is undefined or meaningless — use the standard deviation directly | Data centred on zero, e.g. measurement errors |
Sample vs. Population Formulas
Both formulas measure spread. The only difference is the denominator, and the choice depends on whether your data is the whole group or just a sample of it.
| Quantity | Sample | Population | Notes |
|---|---|---|---|
| Mean | x̄ = Σxᵢ / n | μ = Σxᵢ / N | Same calculation, different notation |
| Sum of squares | SS = Σ(xᵢ − x̄)² | SS = Σ(xᵢ − μ)² | Identical in form |
| Variance | s² = SS / (n − 1) | σ² = SS / N | n − 1 is Bessel's correction |
| Standard deviation | s = √(SS / (n − 1)) | σ = √(SS / N) | Always the positive square root |
| When to use | Data is a sample from a larger population | Data is the entire population | Sample is correct in almost all real work |
| Bias | Unbiased estimator of σ² | Biased low for samples | The gap shrinks as n grows |
The Empirical Rule (68 – 95 – 99.7)
For roughly bell-shaped data, the standard deviation tells you exactly where the values fall. This rule is the basis of most statistical intuition about spread.
| Range | Expected share of data | Outside the range | Meaning |
|---|---|---|---|
| x̄ ± 1s | ≈ 68% | ≈ 32% | About two-thirds of values sit within one standard deviation |
| x̄ ± 2s | ≈ 95% | ≈ 5% | About one value in twenty is further out |
| x̄ ± 3s | ≈ 99.7% | ≈ 0.3% | About one value in 370 is further out |
| x̄ ± 4s | ≈ 99.994% | ≈ 0.006% | About one value in 15,787 |
| x̄ ± 6s | ≈ 99.999999% | ≈ 2 × 10⁻⁹ | The basis of Six Sigma quality control |
Variable Reference
| Symbol | Meaning | Typical range | Units |
|---|---|---|---|
| s | Sample standard deviation | ≥ 0 | same as data |
| σ | Population standard deviation | ≥ 0 | same as data |
| s² | Sample variance | ≥ 0 | data units squared |
| x̄ | Sample mean | any | same as data |
| n | Number of data points | ≥ 2 for sample SD | count |
| SS | Sum of squared deviations | ≥ 0 | data units squared |
| CV | Coefficient of variation | any | percent |
| SEM | Standard error of the mean | ≥ 0 | same as data |
What Does s = √( Σ(xᵢ − x̄)² / (n − 1) ) Measure?
The standard deviation answers a simple question: on average, how far is a typical data point from the mean? The formula builds that answer in four steps. First, subtract the mean from every value to get its deviation. Second, square each deviation so that positive and negative distances do not cancel out. Third, add them all up and divide by n − 1 to get the average squared deviation — the variance. Fourth, take the square root to return to the original units.
The result is the single most widely used measure of spread in all of statistics. It appears in confidence intervals, hypothesis tests, control charts, risk models, and quality standards. Whenever you need to say "the data is typically within about this much of the average," you are quoting a standard deviation.
The Standard Deviation Formula
The sample standard deviation is defined as:
Expanded step by step, the computation is:
2. dᵢ = xᵢ − x̄ for each i
3. SS = Σdᵢ²
4. s² = SS / (n − 1)
5. s = √s²
An algebraically equivalent form, useful when you already know Σx and Σx², avoids computing the mean first:
For a population rather than a sample, replace n − 1 with N:
Why Divide by n − 1?
If you divide by n, the resulting variance is systematically too small when computed from a sample. The reason is subtle but important: the sample mean is not the true population mean. It is the value that sits closest to your particular sample — by definition, it minimises the sum of squared deviations. So measuring spread around the sample mean always underestimates the spread around the true mean.
Dividing by n − 1 rather than n compensates exactly for this. The result is an unbiased estimator: if you took many independent samples, averaged their variances, and compared to the true population variance, the two would match. With n = 10 the correction inflates the variance by about 11%; with n = 100 it is only about 1%. For very large samples the distinction stops mattering, but for small ones it is essential.
How Each Input Affects the Result
The standard deviation responds to two things: how far the values sit from the mean, and how many values there are. Adding a value that is close to the mean pulls s down; adding one far from the mean pulls it up. Removing an extreme value almost always reduces s noticeably, which is why the outlier sensitivity check in the Advanced panel is worth running before you trust a single number.
Note that the standard deviation does not change if you add a constant to every value — shifting all your data by 10 units shifts the mean by 10 and leaves the spread untouched. It does scale, though: multiply every value by k and the standard deviation multiplies by |k|. These two properties (invariance to translation, scaling with multiplication) are what make the standard deviation a natural measure of spread rather than an arbitrary one.
Worked Examples
x̄ = 40 / 8 = 5
Deviations: −3, −1, −1, −1, 0, 0, 2, 4
Squares: 9, 1, 1, 1, 0, 0, 4, 16 → SS = 32
s² = 32 / 7 = 4.5714
s = √4.5714 = 2.1381
x̄ = 868 / 10 = 86.8
SS = 333.6
s² = 333.6 / 9 = 37.0667
s = √37.0667 = 6.0882
σ² = 333.6 / 10 = 33.36
σ = √33.36 = 5.7758
(Note how the population value is always slightly smaller.)
Solving the quadratic gives x = 91 or x = 74.28
Both produce exactly s = 6.0000
Frequently Asked Questions
Use the sample formula (divide by n − 1) whenever your data is a subset drawn from a larger group you want to draw conclusions about. This covers essentially all real-world statistics: a clinical trial, a customer survey, a batch of measurements from a production line. Use the population formula (divide by N) only when your dataset genuinely contains every member of the group you care about — every student in a particular class, every country in a fixed list, every card in a deck. In practice, the population formula is rare. When in doubt, use the sample formula; it is the safer and more conservative choice, and the difference becomes negligible for large n anyway.
Squaring does three useful things. First, it makes all deviations positive so they cannot cancel out. Second, it makes the function smooth and differentiable everywhere, which matters enormously for calculus-based statistics — the mean is the value that minimises the sum of squared deviations, and that property would not hold with absolute values. Third, it weights large deviations more heavily, which is often exactly what you want: a measurement that is 10 units off is more than twice as bad as one that is 5 units off. The cost is sensitivity to outliers, and the fact that the variance ends up in squared units — which is why we take the square root at the end. If you want a measure that treats all deviations proportionally, use the mean absolute deviation instead, which the Advanced panel reports.
It can be zero but never negative. A standard deviation of zero means every value in the dataset is identical — there is no spread at all. This is the minimum possible value, since it is the square root of a sum of squares, and sums of squares cannot be negative. If you see a negative standard deviation, something has gone wrong in the calculation. In this calculator, entering identical values gives s = 0, and the deviation chart will show a flat line of zeros.
Yes, heavily. Because each deviation is squared before being averaged, a value that is twice as far from the mean contributes four times as much to the sum of squares. A single extreme outlier can inflate the standard deviation substantially, sometimes by a factor of two or more. This is why the interquartile range is often reported alongside it — the IQR only looks at the middle half of the data and is essentially immune to extremes. A useful diagnostic: if the IQR is small but the standard deviation is large, the dataset contains outliers and the standard deviation is not a good summary of typical spread. The Advanced panel runs this comparison automatically, showing what s becomes when the largest or smallest value is removed.
The standard deviation describes the spread of individual data points around the mean — it tells you how variable your measurements are. The standard error of the mean (SEM = s / √n) describes the uncertainty in the mean itself — it tells you how precisely you have pinned down the average. The two answer different questions and shrink at different rates: the standard deviation converges to the true population spread as n grows and stops changing, while the standard error keeps shrinking, roughly by a factor of two every time you quadruple your sample size. When you report a confidence interval for a mean, you use the standard error. When you describe how variable the population is, you use the standard deviation.
Mathematically, the sample formula needs at least n = 2, because with a single value there is nothing to compare it to and the denominator n − 1 becomes zero. In practice, though, a standard deviation computed from two or three values is extremely unstable and tells you almost nothing. As a rough guide, n = 5 is the minimum for a meaningful spread estimate, n = 20 gives a reasonably stable result, and n = 30 or more is where the standard deviation becomes a reliable summary of the underlying distribution. The companion statistics that depend on higher moments — skewness and kurtosis — need even more data, requiring n ≥ 3 and n ≥ 4 respectively, and are only trustworthy with n in the hundreds.
Adding one unknown value x to a known dataset turns the standard deviation formula into a quadratic equation in x. Expanding the sum of squared deviations gives (n − 1)x² − 2Sx + (nA − S² − n(n − 1)s²) = 0, where S is the sum of the known values, A is their sum of squares, and n is the total count including the unknown. Solving with the quadratic formula yields up to two answers, which makes intuitive sense: a value far above the mean and a value equally far below it produce the same spread. If the discriminant is negative, no real value can produce the target — usually because the requested standard deviation is smaller than what the known values already force.
Variables
| Symbol | Quantity | Unit |
|---|---|---|
| s | Sample standard deviation | |
| xi | Each individual data value | |
| x_bar | Sample mean | |
| n | Number of data points in the sample |
What it means
The sample standard deviation (s) is a measure of the dispersion or spread of a set of data points relative to the sample mean. It is calculated by taking the square root of the variance. The denominator is n−1 (Bessel’s correction) to provide an unbiased estimate of the population standard deviation. This metric is widely used in statistical analysis to quantify uncertainty, in quality control (six sigma), in finance (volatility of returns), and in many scientific fields. A small s indicates data points are close to the mean; large s indicates wide spread. Understanding standard deviation is essential for interpreting confidence intervals, hypothesis testing, and regression analysis.
Worked example
Sample Standard Deviation – Two Examples
Real‑World| Parameter | Value |
|---|---|
| n | 5 |
| Sum sq dev | 10 |
| Parameter | Value |
|---|---|
| n | 5 |
| Sum sq dev | 250 |
Common mistakes
- Sample standard deviation: s = √[Σ(xᵢ − x̄)²/(n−1)] – denominator is n−1 (degrees of freedom).
- Population: For population σ, use N in denominator (not n−1).
- Mean: x̄ is the sample mean – compute it first.
- Units: The standard deviation has the same units as the data.
- Zero: If all values are equal, s = 0.
Applications
The sample standard deviation, s = √(Σ(xi − x̄)²/(n−1)), measures the spread of a dataset around its mean. It is essential in quality control, engineering tolerances, and statistical inference. Engineers use it to assess process capability, to set control limits, and to evaluate measurement uncertainty. In finance, it measures risk (volatility). In research, it quantifies variability. By understanding standard deviation, professionals can judge the consistency of processes, identify outliers, and make data‑driven decisions. This formula is a cornerstone of descriptive statistics and is used in virtually every field that handles numerical data.
- Quality control and process capability (Six Sigma)
- Measurement system analysis and uncertainty estimation
- Risk assessment in finance and investment
- Experimental data analysis and hypothesis testing
- Performance variation analysis in manufacturing
Related calculators · Descriptive Statistics
Recently updated formulas
Suggest an improvement
Found a mistake? Have a better example? Let us know.
You may send up to 3 suggestions per 24 hours.