Knowing the average isn't enough — two classes can have the exact same mean test score and still look completely different. Standard deviation tells you how tightly (or loosely) your data clusters around that center.
Standard deviation measures the typical distance between each value and the mean. A small standard deviation means values huddle close to the average; a large one means they're scattered wide.
σ = √( Σ(x − mean)² ÷ n )
For each value, measure its distance from the mean, square it (so distances in both directions count positively), average those squared distances, then take the square root to bring the units back to normal.
68% within 1σ · 95% within 2σ · 99.7% within 3σ
For data that follows a roughly bell-shaped (normal) distribution, these percentages hold no matter what the actual mean or spread is — which is what makes the bell curve so useful for spotting what's "normal" versus unusual.
Both classes scored an average of 70 on the same test. Toggle between them below — the 3D surface height shows how densely scores cluster at each point, and the width shows how far they spread.
Class A: 68, 69, 71, 70, 72 — tightly clustered around 70. Class B: 40, 55, 70, 85, 100 — same average, much wider spread.
Scores: 68, 69, 70, 71, 72. Mean = 70. Standard deviation ≈ 1.4 — almost everyone scored within 2 points of average.
Scores: 40, 55, 70, 85, 100. Mean = 70. Standard deviation ≈ 21.2 — the same average, but scores range across the entire scale.
If you only reported "class average: 70" for both, you'd miss that Class B has students who may need serious extra support and others who are already excelling — Class A doesn't.
Low standard deviation. Most values sit close to the mean — predictable, consistent process or population.
Example: a well-calibrated manufacturing lineHigh standard deviation. Values are scattered across a broad range — more variability, less predictability.
Example: response times under inconsistent loadNot all data is symmetric. Income and delivery-time data often pile up on one side with a long tail — the 68-95-99.7 rule no longer applies cleanly.
Example: household income distribution