The empirical rule (also called the 68-95-99.7 rule) is a quick way to estimate how data is spread around the mean in a normal distribution, without doing a full probability calculation. It shows up constantly in statistics, quality control, standardized testing, and finance whenever data is roughly bell-shaped.

How the empirical rule works

For any normal distribution, roughly 68.27% of values fall within one standard deviation of the mean, 95.45% fall within two, and 99.73% fall within three. These figures come directly from the shape of the standard normal curve — they aren't approximations tied to a specific data set, they're a property of the normal distribution itself. That's why the rule only applies when your data is at least approximately normal (symmetric and bell-shaped); it does not hold for skewed or multi-modal distributions.

Inputs and what they mean

This calculator needs just two numbers: the mean (μ), the center of your data, and the standard deviation (σ), which must be greater than 0 and in the same units as the mean. From those two values it derives the ±1σ, ±2σ, and ±3σ ranges automatically. The Within a Range tab lets you check any custom interval — not just whole-sigma steps — using the full normal cumulative distribution function so the percentage stays exact. The Value Position tab converts a single data point into a z-score (how many standard deviations it sits from the mean) and reports which empirical-rule band it falls in.

Limits and edge cases

The empirical rule is an approximation that only holds for normal or near-normal distributions — applying it to skewed data (like income or reaction times) will give misleading ranges. It also breaks down for very small sample sizes, where the sample mean and standard deviation may not reflect the true population parameters. For exact probabilities on non-normal distributions, or for very precise tail calculations, use a full statistical distribution tool rather than the empirical rule alone. Values beyond ±3σ (about 0.27% of a normal distribution) are often flagged as outliers, but always confirm with domain knowledge before discarding a data point.