What it means
Averages hide as much as they reveal. Two sales territories can both average $50,000 of monthly profit while one delivers between $45,000 and $55,000 every month and the other alternates between $10,000 and $90,000.
Standard deviation is the number that tells you which is which, and it is calculated by measuring how far each observation sits from the mean, squaring those distances so positives and negatives do not cancel out, averaging them, and taking the square root to get back to the original units. The squaring step matters more than it looks.
It means large deviations count disproportionately, so a single catastrophic month affects the standard deviation far more than several mildly bad ones. That is deliberate, because in most business settings one very bad outcome is genuinely worse than several slightly disappointing ones.
If the data follow a roughly bell shaped distribution, standard deviation gives a quick sense of likely range: about 68% of observations fall within one standard deviation of the mean and about 95% within two. Investment returns are not perfectly bell shaped, and real markets produce extreme moves more often than the model suggests, so this is a rule of thumb rather than a guarantee.
There is an important distinction between the population and the sample version. If your numbers are the complete set of observations you care about, divide the sum of squared deviations by the number of observations; if they are a sample being used to estimate a wider pattern, divide by that number minus one, which produces a slightly larger figure.
Spreadsheet functions offer both, and mixing them up on small datasets produces noticeably different answers. Standard deviation is only comparable across things measured on the same scale.
To compare the volatility of a $2,000,000 revenue line with a $200,000,000 one, divide the standard deviation by the mean to get the coefficient of variation, a percentage that strips out the size difference. The same logic underlies the Sharpe ratio, which measures how much return an investment delivers per unit of standard deviation.
In practice
Real-world examples
Example 1
A pension trustee compares two funds that both returned an average of 7% a year over a decade, one with a standard deviation of 5 points and the other with 18 points. The trustee chooses the first for members close to retirement, because a bad year in the second fund could take a decade to recover from.
Example 2
A factory manager tracks the weight of filled containers and finds a mean of 500 grams with a standard deviation of 2 grams. When the standard deviation climbs to 9 grams after a machine service, the process is stopped even though the average is still correct, because too many containers now fall outside the legal tolerance.
Example 3
A finance team measures monthly cash collections and finds a standard deviation of $180,000 around a mean of $900,000. That spread justifies holding a larger cash buffer than the average alone would suggest, because a month two deviations below the mean would bring in only $540,000.
Think of it
Imagine a classroom where students' heights are measured. If most students are about the same height, the standard deviation is low, like a row of books neatly lined up. But if some students are much taller or shorter, the standard deviation is high, like books of different sizes on a shelf.
Formula
Calculation
Standard deviation = square root of ( sum of (each value - mean) squared / number of values )
A fund reports annual returns of 4%, 8%, 10%, 12% and 16% over five years. The mean is (4 + 8 + 10 + 12 + 16) / 5 = 50 / 5 = 10%.
The deviations from the mean are -6, -2, 0, 2 and 6 percentage points. Squaring each gives 36, 4, 0, 4 and 36, and those add to 36 + 4 + 0 + 4 + 36 = 80. Dividing by the five observations gives a variance of 80 / 5 = 16, and the square root of 16 is 4, so the standard deviation is 4 percentage points.
Read that back into plain language: returns average 10% and typically land within 4 points either side, so roughly two years in three should fall between 6% and 14%. If the five years were treated as a sample rather than the whole population, the divisor would be 5 - 1 = 4, giving a variance of 80 / 4 = 20 and a standard deviation of about 4.47 percentage points.Case study
Seen in the real world
This is an illustrative and entirely fictional example. Norbrook Seasonal Foods, an invented supplier of chilled ready meals, ran two divisions that each generated average monthly gross profit of $50,000, and the board treated them as equally valuable when allocating capital.
An analyst then calculated the standard deviation of each division's monthly profit over three years. The retail own label division came out at $5,000, meaning a typical month landed between $45,000 and $55,000. The festive and events division came out at $22,000, which implied that a month two standard deviations below the mean would deliver only $50,000 - 2 x $22,000 = $6,000, and several months had in fact been close to breakeven.
That single number changed the fictional company's decisions. The events division still earned its keep on average, but it could not be relied on to fund fixed costs, so the board set its working capital facility against the retail division's steadier cash flows and required the events division to hold a $150,000 reserve. Nothing about the average profit had changed; the board simply started managing the spread as well as the mean.
Watch out
Common mistakes
- Treating standard deviation as a complete measure of risk, when it says nothing about whether the surprises are on the upside or the downside.
- Using the population formula on a small sample, which understates the true spread because the divisor is too large.
- Comparing standard deviations across items of very different size instead of converting them to a coefficient of variation first.
Questions
People also ask
What is the difference between variance and standard deviation?
Variance is the average of the squared deviations, and standard deviation is its square root, which returns the figure to the same units as the original data and makes it interpretable.
Does a high standard deviation always mean a bad investment?
No, it means an uncertain one, and investors are often compensated for that uncertainty with higher expected returns, which is why the two are usually assessed together.
How many observations do you need for it to be meaningful?
There is no hard rule, but a figure based on fewer than about 20 observations should be treated as indicative, because a single unusual value can dominate the result.