This empirical rule calculator applies the 68–95–99.7 rule to any normal distribution: give it a mean and standard deviation and it returns the ranges holding 68%, 95% and 99.7% of the data — plus the probability between any two values you choose.

–

Three Numbers That Sketch the Whole Bell Curve

For a normal distribution, knowing just the mean and standard deviation buys a surprisingly complete picture: about 68% of values fall within μ ± 1σ, 95% within μ ± 2σ, and 99.7% within μ ± 3σ. The defaults use the IQ scale’s canonical parameters (μ = 100, σ = 15): 68% of scores between 85 and 115, 95% between 70 and 130, 99.7% between 55 and 145 — the entire distribution’s geography in three sentences, which is why the rule is the first thing taught about the normal curve and the sanity check behind “six sigma” quality programs (a process whose spec limits sit at ±6σ has, in theory, two defects per billion on the pure normal model). The same three bands run backwards as a diagnostic: measure a real dataset’s fractions inside its own μ ± kσ bands, and departures from 68/95/99.7 are the fingerprint of non-normality — heavy tails, skew, or a mixture wearing a bell costume.

Precision note the rounded rule hides: the exact normal areas are 68.27 / 95.45 / 99.73%, and the boundary statisticians actually quote — 95% — sits at ±1.96σ, not ±2σ. This calculator’s custom-range line computes with the true cumulative normal (via z-scores), so it distinguishes the landmarks from the exacts: enter μ ± 1.96σ and watch it return 95.00%. One boundary of the rule matters more than any rounding, though: it describes NORMAL data only. Chebyshev’s theorem is the distribution-free cousin (weaker guarantees, any shape); the empirical rule’s tight numbers are purchased entirely with the bell-curve assumption, and spending them on skewed or bimodal data is the classic first-semester error this page’s diagnostic framing exists to prevent.

Worked Examples (Verified)

  • Defaults (IQ): μ 100, σ 15 → 68%: 85–115 · 95%: 70–130 · 99.7%: 55–145.
  • Exam scores: μ 72, σ 8 → 95% within 56–88; a 90 sits above the 95% band — top ~1.2% territory (exact CDF: P(X>90) ≈ 1.22%).
  • The 1.96 check: μ 100, σ 15, custom range 70.6–129.4 (= μ ± 1.96σ) → 95.00% exactly.
  • Heights: women μ 64.5 in, σ 2.5 → 99.7% between 57 and 72 in.

Empirical Rule Discipline

  • Normal data only — verify bell-shape (histogram, or the diagnostic of comparing your data’s own band fractions) before spending the 68/95/99.7 numbers.
  • Distinguish the landmarks (95.45% at 2σ) from the convention (95.00% at 1.96σ): confidence intervals are built on the second, mnemonics on the first.
  • σ here is the population/data standard deviation of the DISTRIBUTION — not the standard error of a mean; sampling distributions have their own (narrower) σ.
  • Tail reading: outside ±3σ lives 0.27% — in a dataset of 10,000, expect ~27 such points normally; hundreds means the model, not the data, is wrong.
  • Symmetry is assumed: for skewed quantities the two tails are NOT mirror images, and a single σ misdescribes both.

Related Tools

Frequently Asked Questions

What is the empirical rule?

The observation that for a normal (bell-shaped) distribution, about 68% of values lie within one standard deviation of the mean, 95% within two, and 99.7% within three. With μ = 100 and σ = 15, that’s 85–115, 70–130 and 55–145. This calculator returns all three ranges for any mean and standard deviation, plus the exact probability inside any custom range computed from the true normal CDF.

What percentage of data falls within 2 standard deviations?

Approximately 95% — exactly 95.45% for a perfect normal distribution. The famous “95%” used in confidence intervals corresponds to ±1.96σ rather than exactly ±2σ; the empirical rule rounds both the boundary and the area for memorability. Use this calculator’s custom range with μ ± 1.96σ to see the exact 95.00%, or μ ± 2σ for the landmark 95.45% — knowing which one a textbook means prevents a family of near-miss exam answers.

Does the empirical rule work for any dataset?

No — it is a property of the normal distribution’s shape. For arbitrary distributions, the guaranteed bounds are Chebyshev’s (at least 75% within 2σ, 88.9% within 3σ, for ANY distribution with a finite variance). Real datasets that are roughly bell-shaped (heights, measurement errors, test scores by design) obey the empirical rule closely; skewed data (incomes), bounded data (percentages near 0/100) and bimodal data (mixtures) break it — the mean and SD still exist, but the 68/95/99.7 fractions don’t follow. Plot first, then apply.

How do I find the probability between two values in a normal distribution?

Convert both values to z-scores (z = (x − μ)/σ) and subtract their cumulative probabilities: P(a < X < b) = Φ(z_b) − Φ(z_a). The custom-range fields on this calculator do exactly that computation with a high-accuracy normal CDF. Example: IQ between 85 and 115 with μ = 100, σ = 15 is z from −1 to +1 → Φ(1) − Φ(−1) = 0.8413 − 0.1587 = 68.27%, the unrounded heart of the empirical rule.

What does it mean if a value is 3 standard deviations from the mean?

On a normal model, it’s in the outer 0.27% — beyond the 99.7% band, rarer than 1 in 370. In quality control, that’s the trigger zone for special-cause investigation; in standardized testing it’s the “gifted program” or “intervention” fringe depending on direction; in finance, “3-sigma events” happen far more often than the model predicts because returns are heavy-tailed — the recurring lesson that the empirical rule prices the MODEL’s tails, and reality’s tails are frequently fatter.

What is the difference between the empirical rule and Chebyshev’s theorem?

Assumptions and strength. The empirical rule assumes normality and repays it with tight, exact-ish numbers (95.45% within 2σ). Chebyshev assumes nothing beyond finite variance and guarantees only ≥75% within 2σ and ≥88.9% within 3σ — for any distribution whatsoever. If your data is normal, Chebyshev is true but wasteful; if it isn’t, the empirical rule is precise but wrong. This site has a Chebyshev calculator alongside this one; the pair together mark the boundary between “model-based” and “guaranteed” statistics.

How is the empirical rule used in quality control?

Control charts are the empirical rule industrialized: a process mean with lines drawn at ±2σ (warning limits, ~95%) and ±3σ (action limits, ~99.7%) — a point outside the action lines, or runs and patterns inside them, signals the process has shifted rather than wobbled. Six Sigma methodology pushes the design target to ±6σ tolerances (theoretically ~2 defective parts per billion opportunities on a centered normal, famously adjusted to 3.4 per million for long-term drift). Every one of those charts is this calculator’s three ranges with a hard hat on.

Is this empirical rule calculator free?

Yes — free, instant, no sign-up. Enter a mean and standard deviation for the three landmark ranges plus an exact custom-range probability.