What Is Standard Deviation? A Plain-English Explanation

The average (mean) of a dataset tells you where its center is, but it tells you nothing about how tightly the individual values cluster around that center. Two classes can both average 75% on a test — one because every single student scored close to 75, the other because half the class scored near 100 and the other half near 50. Standard deviation is the number that captures exactly that difference: how spread out the data actually is around its mean. This guide builds the formula up from scratch, works through a complete numeric example by hand, and explains the population-vs-sample distinction that trips up most people the first time they meet it.

What standard deviation measures, intuitively

Standard deviation answers the question: "on average, how far does each data point sit from the mean?" A small standard deviation means most values sit close to the mean — the data is consistent and predictable. A large standard deviation means values are scattered widely above and below the mean — the data is variable and less predictable. A standard deviation of exactly zero means every single value in the dataset is identical to the mean, with no variation at all.

The formula, step by step

Standard deviation is built in five distinct steps, and it's worth understanding what each step is doing rather than memorizing the final formula as a black box:

1. Find the mean of the dataset.
2. Find each value's deviation from that mean (value minus mean).
3. Square each deviation (this makes every value positive and penalizes larger deviations more heavily).
4. Average the squared deviations — this average is called the variance.
5. Take the square root of the variance — this final number is the standard deviation, back in the original units of your data.

In formula form, for a population of size N with mean \( \mu \):

\[ \sigma^2 = \frac{\sum (x - \mu)^2}{N} \qquad \sigma = \sqrt{\sigma^2} \]

Here \( \sigma^2 \) (sigma squared) is the variance and \( \sigma \) is the standard deviation — the Greek letter \( \sigma \) is the conventional symbol for population standard deviation in statistics.

A fully worked example

Suppose a small café tracks how many coffees it sells during five separate hours in one morning: 4, 8, 6, 5, 3. Let's find the standard deviation of these five numbers from scratch.

Step 1 — the mean:

(4 + 8 + 6 + 5 + 3) ÷ 5 = 26 ÷ 5 = 5.2

Step 2 and 3 — each deviation from the mean, and each deviation squared:

Value (x)	Deviation (x − mean)	Squared deviation
4	4 − 5.2 = −1.2	1.44
8	8 − 5.2 = 2.8	7.84
6	6 − 5.2 = 0.8	0.64
5	5 − 5.2 = −0.2	0.04
3	3 − 5.2 = −2.2	4.84

Step 4 — average the squared deviations (variance):

Sum of squared deviations: 1.44 + 7.84 + 0.64 + 0.04 + 4.84 = 14.8

Treating this as the full population (all five hours we care about, not a sample of a larger set), variance = 14.8 ÷ 5 = 2.96

Step 5 — square root (standard deviation):

\( \sqrt{2.96} \approx \) 1.72

So this dataset has a mean of 5.2 coffees per hour and a population standard deviation of about 1.72 — meaning hourly sales typically sit within roughly plus-or-minus 1.72 coffees of the 5.2 average, rather than being identical every hour.

Population vs. sample standard deviation

The calculation above treated all five hours as the entire population we care about — every data point that exists. But very often, the data you have is only a sample drawn from some larger population you can't fully measure — for example, surveying 50 customers out of many thousands, or measuring 5 mornings as a stand-in for "typical" café performance in general. When your data is a sample rather than the full population, the standard deviation formula changes slightly: you divide by n − 1 instead of n.

\[ s^2 = \frac{\sum (x - \bar{x})^2}{n - 1} \qquad s = \sqrt{s^2} \]

Using our same five numbers, but now treating them as a sample:

Sample variance = 14.8 ÷ (5 − 1) = 14.8 ÷ 4 = 3.7

Sample standard deviation = \( \sqrt{3.7} \approx \) 1.92

Notice the sample version (1.92) is slightly larger than the population version (1.72). This isn't a coincidence — dividing by a smaller number (n − 1 instead of n) always produces a slightly larger result, and that's precisely the point.

Why n − 1? (Bessel's correction)

When you calculate a sample's mean and then measure each point's deviation from that same sample's mean, the sample mean is, by construction, the value that minimizes the sum of squared deviations for that specific sample. In other words, the sample mean fits the sample slightly "too well" compared to how well the true (unknown) population mean would fit it — so a sample's deviations tend to slightly understate the real spread of the whole population. Dividing by n − 1 instead of n inflates the result just enough to correct for that built-in understatement, giving an unbiased estimate of the population's true variance. This adjustment is called Bessel's correction, and the practical rule is simple: use n when you have data for an entire population you're describing; use n − 1 whenever your data is a sample being used to estimate a larger population you haven't fully measured. Sample standard deviation is, by far, the more common case in real-world statistics — polls, experiments, and studies almost always work from a sample.

The empirical rule (68-95-99.7)

For data that follows a normal distribution (the familiar symmetric, bell-shaped curve), standard deviation has a remarkably consistent practical meaning known as the empirical rule:

• About 68% of values fall within 1 standard deviation of the mean.
• About 95% of values fall within 2 standard deviations of the mean.
• About 99.7% of values fall within 3 standard deviations of the mean.

Worked example: IQ scores are constructed to be normally distributed with a mean of 100 and a standard deviation of 15. Applying the empirical rule:

• 68% of people score between 100 − 15 = 85 and 100 + 15 = 115
• 95% of people score between 100 − 30 = 70 and 100 + 30 = 130
• 99.7% of people score between 100 − 45 = 55 and 100 + 45 = 145

This is exactly why standard deviation is so useful beyond just describing a single dataset: once you know a distribution is roughly normal, the standard deviation alone tells you how likely any given value is, without needing to look at every individual data point.

Why square the deviations instead of just using distance?

A natural question is why step 3 squares each deviation rather than simply taking its absolute value (ignoring the negative sign). Two reasons: first, squaring is easier to work with mathematically — it's differentiable everywhere, which matters for the calculus used to build more advanced statistical methods on top of variance. Second, squaring deliberately penalizes larger deviations more than proportionally — a deviation twice as large contributes four times as much to the variance, not just twice as much — which is often exactly the behavior you want when a few extreme outliers should count more heavily than many small ones. (There is an alternative measure, mean absolute deviation, that uses absolute values instead of squares and doesn't have this amplifying property — it's used occasionally, but standard deviation remains the far more common default.)

What a large vs. small standard deviation tells you in practice

Consider two classes that both score an average of 75% on the same exam. Class A has a standard deviation of 5; Class B has a standard deviation of 20. Even though the average is identical, these describe very different classrooms: in Class A, almost every student scored somewhere between 65 and 85 — a consistent group clustered tightly around the average. In Class B, scores are scattered far more widely, with plenty of students well below 55 and plenty well above 95 — a group with wildly different levels of understanding, even though the class average looks the same on paper. This is the core reason standard deviation matters alongside the mean rather than instead of it: the mean alone can hide enormous differences in consistency.

Common practical uses

• Finance — standard deviation of returns is the standard measure of an investment's volatility; a stock with a higher standard deviation of historical returns is considered riskier, even if its average return is identical to a calmer stock.
• Quality control — manufacturers track the standard deviation of a measured part dimension to catch a process that's becoming inconsistent, even before any individual part actually falls outside spec.
• Grading and standardized testing — test scores are often reported alongside the population mean and standard deviation so an individual score can be interpreted relative to everyone else, not just as a raw number.
• Weather and climate data — the standard deviation of daily temperatures distinguishes a climate with reliably mild weather from one with wide day-to-day swings, even if both have the same average temperature.

Frequently asked questions

Can standard deviation be negative?

No — because it's a square root of a sum of squared values, standard deviation is always zero or positive. A standard deviation of exactly 0 means every value in the dataset is identical.

What's the difference between variance and standard deviation?

Variance is the average of the squared deviations; standard deviation is its square root. Standard deviation is generally more useful for interpretation because it's expressed in the same units as the original data (dollars, coffees per hour, points), while variance is in squared units (squared dollars, for instance) that don't have an intuitive real-world meaning.

Should I use the population or sample formula?

Use the population formula (divide by n) only when your dataset genuinely contains every member of the group you're describing. Use the sample formula (divide by n − 1) whenever your data is a subset being used to estimate something about a larger population — which describes the overwhelming majority of real-world statistics, surveys, and experiments.

Does a bigger dataset always have a bigger standard deviation?

No — dataset size and spread are independent. A dataset of 10,000 tightly clustered values can have a smaller standard deviation than a dataset of just 5 widely scattered ones. Standard deviation measures spread, not sample size.

Calculate it yourself

Working through the mean, deviations, squares, and square root by hand is a great way to understand the formula once, but it's tedious for real datasets. DocNectar's Standard Deviation Calculator accepts any list of numbers and instantly returns the mean, variance, and both population and sample standard deviation, with the full step-by-step working shown.