Standard Deviation Calculator
Spread, variance and the sample-versus-population choice made explicit.
Paste a list of numbers and get mean, median, mode, range, count and sum together, so a single outlier cannot hide behind the mean and go unnoticed.
About precision. Results are computed in double-precision floating point, which is exact for whole numbers up to about sixteen digits and very slightly approximate beyond that. Figures are rounded for display, so a long chain of calculations can differ from a hand-worked answer in the last decimal place. Where an exact fraction or radical exists, this page shows it alongside the decimal.
“Average” means at least three different things, and which one someone chose usually tells you more than the number itself. This calculator gives all of them at once, so a mean pulled sideways by one enormous value cannot pass unnoticed.
Negative numbers and decimals are both fine. The only limit is 5,000 values, which is well past the point where a summary is more useful than the list.
Four measures of centre and two of spread.
| Symbol | Meaning | Unit | Typical range |
|---|---|---|---|
n | Count of values | count | 1 – 5,000 |
Mean | Arithmetic average | same as data | — |
Median | Middle value in order | same as data | — |
Mode | Most frequent value | same as data | — |
IQR | Spread of the middle half | same as data | — |
The mean is the balance point: if the values were weights on a beam, it is where the beam would balance. That mental picture explains its weakness immediately — one very heavy weight a long way out moves the balance point a long way with it, no matter how many small weights sit at the other end.
Add a single value of 210 to the same list. The mean jumps from 53.25 to 65.31 — a 23% increase caused by one number. The median moves from 52.00 to 53.00, a change of one point. That contrast is the entire case for reading both: the mean noticed the outlier and the median barely did, and which behaviour you want depends on whether the outlier is real signal or a typing error.
On the original data the mean is 53.25, the median is 52.00 and the mode is 58.00. All three are correct and all three are “the average”. Anyone quoting a single figure without saying which one they used has left out the most informative part of the sentence — and in a dataset with a long tail, the choice can move the headline number by a great deal more than it does here.
Every summary figure for the worked dataset.
| Measure | Value | How it is found | Note |
|---|---|---|---|
| Count | 12 | How many values | |
| Sum | 639 | All values added | |
| Mean | 53.25 | Sum ÷ count | Moved by every outlier |
| Median | 52.00 | Middle value in order | Barely moved by outliers |
| Mode | 58 | Most frequent value | Appears three times |
| Range | 30 | 72 − 42 | Spread from lowest to highest |
The mean sits above the median here, which is the signature of a distribution with a longer tail on the right — in this case the 72 pulling upward with nothing equally far below to balance it. When the mean exceeds the median, look at the top of the sorted list; when it falls below, look at the bottom.
Which average to use is a question about what you are trying to say. Mean is right for anything that will be multiplied or totalled: mean order value times orders gives revenue, and the median would not. Median is right for anything describing a typical case: house prices, salaries and response times are all reported as medians precisely because a handful of extreme values would otherwise dominate.
Mode is the least used and occasionally the only one that works. It is the only average defined for categories rather than numbers — the most common colour, the most frequent error code — and on numeric data it flags clustering that the other two cannot see. For the spread around whichever centre you choose, the standard deviation calculator is the natural next step.
One habit worth adopting: never report a mean without the count beside it. A mean of 53.25 from twelve values and the same mean from twelve thousand are very different claims, and the count is the cheapest possible signal of how much weight the figure deserves. Adding the median costs two more characters and doubles the information.
Five ways an average misleads even when the arithmetic is perfect.
Averaging averages. The mean of two means is only the true mean if both groups are the same size. Two branches averaging 40 and 60 do not combine to 50 unless they served the same number of customers — you have to go back to the totals and the counts.
Averaging rates. Driving out at 30 miles per hour and back at 60 does not average 45. The correct figure is the harmonic mean, 40, because you spend twice as long at the slower speed. Any average of speeds, prices per unit or percentages deserves this check.
Ignoring the distribution. A mean of 53 describes a dataset clustered around 53 and one split evenly between 20 and 86 identically. Two datasets with the same mean can tell entirely opposite stories, which is why the range and interquartile figures sit beside it here.
Averaging over the wrong period. A monthly average that spans a seasonal peak and trough describes a year that never happens. Match the window to the cycle in the underlying process, or compare the same month across years.
The average of a bimodal set describes nobody. If half your customers spend $20 and half spend $200, the mean of $110 is a number no customer has ever spent. That is the clearest case for looking at the shape of the data rather than summarising it, and the mode is usually the figure that gives the game away.
Add every value and divide by how many there are. On the worked example, 639 ÷ 12 = 53.25.
Mean is the arithmetic average, median is the middle value in order, and mode is the most frequent. On this dataset they are 53.25, 52.00 and 58 — all correct, all different.
Whenever extreme values would distort the picture. Adding one value of 210 here moves the mean by 23% and the median by one point, which is why salaries and house prices are reported as medians.
Average the middle two. With twelve values sorted, that is the sixth and seventh: (51 + 53) ÷ 2 = 52.00.
Yes. If several values tie for most frequent, all of them are modes. If no value repeats at all, the dataset has no mode — which is itself informative.
The spread of the middle half of the data, from the 25th to the 75th percentile. It ignores the extremes entirely, so a wide range with a narrow IQR means a few values sit a long way out.
Only if the groups are the same size. Two branches averaging 40 and 60 combine to 50 only if they served equal numbers; otherwise go back to the totals and the counts.
Not with the arithmetic mean. Out at 30 and back at 60 averages 40, not 45, because you spend twice as long at the slower speed. Rates need the harmonic mean.
No. Every calculation runs in your browser, so a column of salaries or measurements never leaves the device.
The data is skewed. If the mean is higher, look at the top of the sorted list for large values pulling it up; if lower, look at the bottom. Either way the median is usually the fairer summary.
Six tools that pick up where this one leaves off.
Spread, variance and the sample-versus-population choice made explicit.
MathThree percentage questions in one tool, with the working shown.
MathSingle, combined and at-least-once probabilities without the algebra.
MathSimplify, solve for a missing term, or scale a mix.
MathWeighted assignment scores turned into a current course grade.
EverydayEngagement by followers and by reach — two different stories.
MarketingCookies. We use cookies for site functionality and to serve ads via Google AdSense. Nothing you type into a calculator is ever stored or sent anywhere. Read our Cookie Policy.