What Is Standard Deviation? A Plain-English Explanation with Examples

6 min read · Math Help

Standard Deviation Shows How Far Data Spreads From Average

  • Standard deviation tells you how spread out a set of numbers is around its average. A small value means the numbers huddle close to the average, and a large value means they scatter.

  • You can work it out yourself in five short steps. Find the average, measure each distance from it, square those distances, average the squares, then take the square root of the result.

  • Bell-shaped data follows a pattern that lets you judge any single result. Most values land within two standard deviations of the average, so a result far outside that zone stands out as unusual.

  • The number can mislead when the data are lopsided or full of extreme values. Checking the shape of the data first keeps you from trusting a figure that does not fit.

You are at the kitchen table with your child’s school portal open. The latest unit test shows a class average, and right beside it sits a line that reads standard deviation. Your child scored a little above average, but does that mean a strong result or an ordinary one? The average alone cannot say, and a note from the teacher will not settle it either.

This article answers the question “what is standard deviation?” in three parts: what the number means, how to build it by hand, and when to trust it.

What Does a Standard Deviation Number Tell You?

The Short Answer

Standard deviation tells you how far the values in a data set typically sit from their average. A small standard deviation means the values cluster tightly around the average. A large one means they are spread widely. It is measured in the same units as the data, such as test points or inches of height.

Why an Average Needs a Spread

An average hides a lot. Say two classes both average 70 on a quiz. In the first class every student scored 70. In the second, the scores were 50, 60, 70, 80 and 90. The averages match, yet the classes look nothing alike.

Standard deviation is the number that captures that difference. Treating each class as the whole group, it comes out to 0 for the first and about 14 for the second. Think of it as a typical distance from the average. Some values sit closer, others sit farther, and the number summarizes the whole group.

The National Institute of Standards and Technology, the U.S. government measurement agency, defines it in its NIST/SEMATECH e-Handbook of Statistical Methods as the square root of the variance. Variance is the average of the squared distances from the average, and it is stated in squared units, such as square points. Taking the square root returns the spread to the original units, which is why standard deviation is the number people quote.

Reading Results on a Bell Curve

Many real data sets, such as test scores and heights, pile up in a bell shape. JMP, the statistical software company, describes what happens when data follow that shape:

  • About 68% of values fall within one standard deviation of the average.

  • About 95% fall within two standard deviations.

  • About 99.7% fall within three standard deviations.

So a score within one standard deviation of the class average is an ordinary result. A score beyond two standard deviations, high or low, is rare. Statisticians call this the empirical rule, or the three sigma rule.

Treat it as a reading guide for bell-shaped data only. Wikipedia’s entry on the 68-95-99.7 rule gives the exact shares as about 68.27%, 95.45% and 99.73%. For data that are not bell-shaped, Chebyshev’s inequality, a general result in statistics, guarantees only that at least 88.8% of values sit within three standard deviations.

Reading Spread in Labs and Reports

Nursing and medical students meet the same idea in lab results. Whitley and Ball, writing in the medical journal Critical Care in 2002, explain that a 95% reference range runs from the average minus 1.96 standard deviations to the average plus 1.96 standard deviations. Their hemoglobin example has an average of 9.9 grams per deciliter and a standard deviation of 2.0, which gives a range of 5.98 to 13.82.

The method holds only when the data are roughly bell-shaped. The same authors show a lactate example where the lower limit came out negative, an impossible result, because the data were not bell-shaped.

Business reports work the same way. A monthly sales average with a small standard deviation points to steady months. A large one points to swings, so a manager treats that average with less confidence. Whether the numbers come from a gradebook, a chart or a budget meeting, read the average and the spread side by side.

How Do You Calculate Standard Deviation by Hand?

The Five Steps, Worked Through

Every calculation follows the same five steps, in this order:

  1. Add the values and divide by how many there are to get the average.

  2. Subtract the average from each value to get each distance.

  3. Square each distance, so negatives do not cancel positives.

  4. Average the squared distances to get the variance.

  5. Take the square root of the variance.

Math Is Fun, an education site, works this through with five dog heights in millimeters: 600, 470, 170, 430 and 300. The total is 1,970, so the average is 394. The distances from the average are 206, 76, negative 224, 36 and negative 94. Squared, they add up to 108,520. Dividing by 5 gives a variance of 21,704, and its square root is about 147 millimeters. Squaring gives extra weight to far-off values, which is why the gap of 224 dominates the total.

Rusty on squares and square roots? A prealgebra full course is a sensible place to rebuild them, and writing each step on its own line makes mistakes easy to spot.

Dividing by N or N Minus 1

Step four has a choice. If your numbers are the whole group, divide by the count, written N. If they are only a sample, a slice of a larger group, divide by N minus 1 instead. Math Is Fun notes that real-world data are usually a sample, so N minus 1 is the common choice. Treat the dog heights as a sample and the variance becomes 27,130, with a standard deviation of about 165 millimeters. The two answers differ most for small sets like this one, and the gap shrinks toward nothing as the number of values grows.

Wikipedia’s entry on Bessel’s correction, the name for this adjustment, explains the reason. Distances measured from the sample’s own average run a little small, so dividing by N would underestimate the spread of the larger group. The NIST handbook uses N minus 1 in its formula too.

A single standard deviation means little until you know whether the data are bell-shaped and whether you measured a sample or a whole group.

When Standard Deviation Misleads You

The number works best on roughly bell-shaped data. The NIST handbook warns that it is sensitive to long tails, meaning a few extreme values far from the rest. For data like that, it suggests the median absolute deviation, the average absolute deviation, or the interquartile range, which is the spread of the middle half of the values.

Do not mix it up with standard error, a different measure. Whitley and Ball give a simple rule: use standard deviation to describe how individual values vary, and standard error to judge how precisely a sample estimates the true average.

To see these steps worked on video, Cool Math Guy’s Statistics course has 36 lessons across 11 chapters and lists standard deviation among its topics, taught by Dana Mosely. The math study skills course covers how to practice and check your own work.

Back at the kitchen table, the line beside the class average now has a meaning: it says how tightly the scores hug that average. Together, the mean (the usual average) and the standard deviation describe a data set far better than either one alone.

Practice Standard Deviation Step by Step

Watch each step worked out on video, then try it on numbers from your own week

Explore the Statistics Course →

External References