The standard deviation measures the typical (root-mean-square) distance of values from the mean — small means clustered, large means spread out.
σ=n∑(x−μ)2
Quick Answer
The standard deviation measures the typical (root-mean-square) distance of values from the mean — small means clustered, large means spread out. Use it when you need one number for how consistent a set of measurements is, in the same units as the data: points on a test, minutes of waiting, millimeters of error. Find each value's distance from the mean, square those distances, average them, and take the square root. It is the yardstick behind z-scores and the 68-95-99.7 rule.
Why This Matters
Two classes can share a mean of 80 while one scores 78 to 82 and the other 50 to 110; the mean alone cannot tell them apart. The standard deviation can (about 1.4 points against 21.2), and it becomes the unit for how unusual a value is: in bell-shaped data, a score 2 SDs above the mean beats about 97.5% of the group, whatever the test.
Intuitive Explanation
Two classes both average 80. Class A scored 78, 79, 80, 81, 82; class B scored 50, 65, 80, 95, 110. Measure how far each score is from 80. In class A the distances are 2, 1, 0, 1, 2; in class B they are 30, 15, 0, 15, 30. The standard deviation boils each list of distances down to one typical distance: about 1.4 points for A and 21.2 points for B.
Why square and then take a root, instead of averaging the distances directly? The signed deviations always add to zero (for B, −30−15+0+15+30=0), so their plain average is useless. Squaring makes every deviation positive and gives large deviations extra weight, so one far-off value raises the SD more than several near ones. For B the squares are 900, 225, 0, 225, 900, averaging 52250=450, and 450≈21.2. The square root brings 'points squared' back to points.
The number before the square root, 450, is the variance. It measures the same spread, but in squared points, so it cannot be compared with the scores directly; if your spread has squared units, you stopped one step early. And when the five scores are a sample used to estimate a larger group, divide by n−1 instead of n: s=42250≈23.7.
Think of it like this.Mark the center of where a player's darts landed. The standard deviation is the typical distance of a dart from that center: a steady player's darts sit within a centimeter or two, a wild player's sprawl across the board. One difference from a plain average distance: squaring makes a single far miss count extra.
Core idea.Standard deviation is roughly how far a normal data value sits from the mean, on average.
When to Use
Use the standard deviation when a problem asks how consistent, spread out or variable values are around their mean and wants the answer in the data's units. It is also the step before z-scores and normal-curve questions ('within 2 standard deviations'). Divide by n for a whole population and by n−1 for a sample. If the data has strong outliers or is skewed, the IQR is a sturdier spread; if a formula needs squared units, keep the variance.
Tip.Ask: Am I measuring how far values typically fall from the mean, in the data's own units?
How to Recognize It
Before computing a standard deviation, check that the question is about how tightly values cluster around their mean and that the answer should come in the data's own units.
1.
Is the question about how consistent or spread out a whole group is, rather than where one value stands?
If yes, the standard deviation summarizes the group. If it asks how unusual one particular value is, you need a z-score, which uses the standard deviation as its unit of distance.
2.
Should the answer be in units you can picture, such as seconds or millimeters, rather than squared units?
If yes, finish with the square root to get the standard deviation. If a later formula needs squared units, for example when adding the spreads of independent quantities, stop before the root and keep the variance.
3.
Is the data a sample drawn from a larger group, such as 30 bolts tested from a day's production?
If yes, divide by n−1 to get the sample standard deviation s. If you have every member of the group, divide by n for the population standard deviation σ.
4.
Is the data roughly symmetric, without a few extreme values?
If yes, the standard deviation is a good summary. If the data is strongly skewed or has outliers, which squaring makes count extra, the interquartile range describes the spread more fairly.
Standard Deviation compared with Variance, Range, and Mean absolute deviation (MAD)
Standard deviation, variance, range and mean absolute deviation all put a number on spread. Ask what each one measures: the gap between the extremes (range), the average plain distance from the mean (MAD), the average squared distance (variance), or the square root of that (standard deviation).
Standard Deviation compared with Variance, Range, and Mean absolute deviation (MAD)
Type
Meaning
Key test
Formula
Example
Standard Deviation
Use this when you want a typical spread around the mean that uses every value and stays in the data's units.
Am I measuring how far values typically fall from the mean, in the data's own units?
σ=n∑(x−μ)2
Find the population SD of 2,4,4,4,6 (mean =4).
Variance
The same spread before taking the square root — in squared units.
Use when the math (adding spreads, ANOVA) is easier in squared units.
σ2=n∑(x−μ)2
Variance25 means SD 5
Range
Total span from max to min, using only the two extremes.
Use for a quick spread when you do not need all values or the mean.
max−min
Span of daily temperatures
Mean absolute deviation (MAD)
Average of the absolute (not squared) distances from the mean.
Use when you want average distance without squaring, often in middle school.
n∑∣x−μ∣
Average miss of dart throws from center
Standard Deviation
Meaning
Use this when you want a typical spread around the mean that uses every value and stays in the data's units.
Key test
Am I measuring how far values typically fall from the mean, in the data's own units?
Formula
σ=n∑(x−μ)2
Example
Find the population SD of 2,4,4,4,6 (mean =4).
Variance
Meaning
The same spread before taking the square root — in squared units.
Key test
Use when the math (adding spreads, ANOVA) is easier in squared units.
Formula
σ2=n∑(x−μ)2
Example
Variance25 means SD 5
Range
Meaning
Total span from max to min, using only the two extremes.
Key test
Use for a quick spread when you do not need all values or the mean.
Formula
max−min
Example
Span of daily temperatures
Mean absolute deviation (MAD)
Meaning
Average of the absolute (not squared) distances from the mean.
Key test
Use when you want average distance without squaring, often in middle school.
How to read it:σ for population SD, s for sample SD (which divides by n−1)
Worked Examples
Example 1 — SD of five values
Easy
Find the population SD of 2,4,4,4,6 (mean =4).
Show the steps and the answer
Subtract the mean from each value: −2,0,0,0,2.
Each deviation says how far, and in which direction, a value sits from 4; they add to 0, as deviations from a mean always do.
Square the deviations: 4,0,0,0,4.
Squaring removes the signs, so distances below the mean no longer cancel those above.
Average the squares over n=5: 54+0+0+0+4=58=1.6.
The problem asks for the population SD, so divide by n; this 1.6 is the variance.
Take the square root: 1.6≈1.26.
The root returns to the data's units; 1.26 sits between the three distances of 0 and the two distances of 2, as a typical distance should.
Answer:≈1.26
Takeaway:Deviations, squares, average, square root; stopping before the root leaves the variance, in squared units.
Example 2 — They handed you variance
Standard
A data set's variance is given as 49 and you are asked for the spread in the data's units.
Show the steps and the answer
Note the units: a variance averages squared deviations, so it is in the data's units squared.
If the data were in centimeters, 49 would be 49 square centimeters, which is not a distance on the data's scale.
Take the square root: 49=7.
The standard deviation is the square root of the variance, which converts squared units back to the data's units.
Interpret it: a typical value lies about 7 units from the mean.
Reporting 49 as the spread would overstate it seven-fold.
Answer:49=7
Takeaway:Given the variance, one square root gives the standard deviation; there is no need to recompute the deviations.
Example 3 — Find the mistake: sample or population?
Application
An inspector weighs a sample of 4 bags of chips from a day's production: 28, 30, 31 and 31 g. Dev writes: "Mean =30, squared deviations 4,0,1,1, so s=6/4≈1.22 g." Find the mistake and correct it.
Show the steps and the answer
Check the work so far: mean 4120=30; deviations −2,0,1,1; squares 4,0,1,1, total 6.
Dev's arithmetic is right up to the division, so the error is in the divisor.
Notice the data is a sample of 4 bags standing in for the whole day's production, and the answer is labeled s.
A sample's own mean sits closer to its values than the true mean does, so dividing by n makes the spread look too small; n−1 corrects for that.
Divide by n−1=3: s2=36=2.
This is the sample variance.
Take the root: s=2≈1.41 g.
With only 4 values, the divisor changes the answer by about 15%.
Answer:s=2≈1.41 g
Takeaway:Sample data uses n−1; population data uses n. With small samples the difference is large, 1.22 g against 1.41 g here.
Where You'll See It
Choosing batteries for a camping trip
Brand A batteries last a mean of 10 hours with SD 0.5 hours; brand B also averages 10 hours, with SD 3 hours. If you need 8 hours, A is the safer pick: 8 hours is 4 SDs below A's mean, but under 1 SD below B's.
Filling 500 mL drink bottles
A bottling line tracks the SD of fill volumes. With a mean of 500 mL and SD 2 mL, roughly normal fills put 99.7% of bottles between 494 and 506 mL. If the SD drifts up to 5 mL, bottles as low as 485 mL appear and the machine is recalibrated.
Risk of an investment fund
In finance the SD of yearly returns is called volatility. A fund averaging 8% a year with SD 15% has bad years one SD below the mean at 8−15=−7%, a loss, while a savings account with SD near 0 almost never surprises you.
Common Mistakes
The slip:Forgetting the square root
Instead:that gives variance, not standard deviation.
The slip:Dropping the squaring of deviations
Instead:positive and negative distances would cancel; square them first.
The slip:Dividing a sample by n when you should divide by n−1
Instead:sample SD (s) uses n−1, population SD (σ) uses n.
Mini Practice
Try these on your own. Tap Reveal when you want to check.
1.
Find the population standard deviation of 10,10,16,16. The mean is 13.
Hint:Find each distance from the mean, square it, average the squares, then take the root.
2.
Find the population standard deviation of 5,7,8,9,11.
Hint:Start by finding the mean of the five values.
3.
A sample of 4 plant heights is 12,15,18,15 cm. Find the sample standard deviation s.
Hint:For a sample, divide the sum of squared deviations by n−1.
4.
A set of test scores has a standard deviation of 4.2 points. The teacher adds 5 bonus points to every score. What is the new standard deviation?
Hint:Does adding the same amount to every score change how far apart the scores are?
5.
A coach wants one number for how far a typical sprint time is from the team's mean time, using every runner's time. Standard deviation or range?
Hint:Think about which measure uses all the values and which uses only two.
6.
For 2,4,6 (mean 4), a student finds the squared deviations add to 8, divides by 3, and reports "SD ≈2.67." Correct the answer.
Hint:Check whether the result is still in squared units.
Want the full set?
50 practice questions for this concept — free to try, every one with a complete worked
solution showing the why, not just the answer.
It can be zero, but only when every value is the same: 7, 7, 7, 7 has no spread at all, so every deviation is 0. It can never be negative, because it is the square root of an average of squares, and squares are never negative. A standard deviation of 0 on real measurements is a warning sign: either the values really are identical, or the instrument rounds too coarsely to show the differences.
Why square the distances instead of just using absolute values?
Absolute values work too, and they give the mean absolute deviation. For 4, 6, 8, 10 and 12 the MAD is 2.4, while the standard deviation is about 2.83; squaring gives large deviations extra weight, so the standard deviation is never smaller than the MAD. Squares are preferred because they behave well in algebra: variances of independent quantities add, and the normal curve's formula is built on σ.
How do investors use standard deviation?
They use it to measure risk. A fund's volatility is the standard deviation of its yearly returns. Two funds might both average a 7% return, one with a standard deviation of 5 percentage points and the other with 20. A year one standard deviation below average would mean a 2% gain for the first fund and a 13% loss for the second. Same average, very different ride, and the standard deviation is the number that shows it.
You need the Mean, since every deviation is measured from it, and Square Roots, to undo the squaring at the end. Standard deviation adds one typical-distance number for spread. Next, Variance is the same calculation without the final root and is the version used inside formulas, and the Normal Distribution uses the SD as its ruler: about 68% of values within one SD of the mean.