All lessons

Module 3: Distributions · Lesson 3.4 · about 25 min

The normal distribution and its areas

The bell curve: its two numbers μ and σ, the 68-95-99.7 rule, turning any normal value into a z-score and reading probabilities as areas, and why it turns up everywhere averages are taken.

Key points, meanings and formulas on one page: choose "Save as PDF" in the print box.

Module 3 · about 25 minutes · school maths. You’ll be able to use the 68-95-99.7 rule, convert values to z-scores and read normal probabilities, and find a value for a given percentile.

The bell curve is described by its centre μ and spread σ, and the share of values in any range is the area under it.

In everyday terms: pour sand slowly onto one spot: it piles into a smooth hump, high in the middle and thin at the edges. Anything made by adding up many small independent nudges (fill weights, heights, sums of sales) piles up the same way.

1. The bell curve

The normal distribution is the symmetric bell-shaped curve set by two numbers: its mean μ (the centre) and standard deviation σ (the width). Written X∼N(μ,σ2). Its density is

f(x)=1σ2πe−(x−μ)22σ2

You’ll never need to compute that by hand. What matters is that every normal curve is the same shape, just shifted and stretched, so one table (or one spreadsheet function) of areas serves them all.

2. The 68-95-99.7 rule

For any normal distribution:

  • about 68 % of values lie within 1σ of the mean,
  • about 95 % within 2σ (exactly 95 % within 1.96σ),
  • about 99.7 % within 3σ.
Try it: area under the bell curve
area between (probability)
left tail
right tail
The standard normal has mean 0 and standard deviation 1; any normal value turns into this scale with z = (x − μ) ÷ σ. The shaded area is the probability of landing between the two limits. ±1.96 holds exactly 95 %, which is where the 1.96 in confidence intervals (Module 5) comes from.

So with daily orders N(500, 40²): two days in three see 460 to 540 orders; one day in 20 falls outside 420 to 580; a day below 380 should happen about once in 2 years.

3. Any normal → the standard normal

Convert a value to a z-score (Lesson 1.5):

z=x−μσ

Z follows the standard normal, N(0, 1). Now any question becomes “area under the standard normal curve”:

  • P(X ≤ x): area to the left of z. Excel NORM.S.DIST(z, TRUE); Python scipy.stats.norm.cdf(z).
  • The value for a percentile: find the z with that area to its left (NORM.S.INV(p), norm.ppf(p)), then x=μ+zσ.

Key z-values worth knowing: 1.282 (90th percentile), 1.645 (95th), 1.96 (97.5th), 2.326 (99th).

4. Where the normal fits, and where it doesn’t

It fits measurement errors, heights, many quality measurements and, most importantly, averages and totals of many independent values (Lesson 4.3). Daily totals of a busy shop are often close to normal even though single baskets are skewed.

It fits badly for: money per customer (long right tail, Lesson 3.5), small counts (use the binomial or Poisson), anything with a hard limit nearby, and data with fat tails (big moves far more often than “3σ is rare” suggests). Lesson 3.6 shows how to check.

μ
The centre of the bell: its mean, median and mode.
σ
The spread: the distance from the centre to where the curve bends.
68-95-99.7
Shares within 1, 2 and 3 σ of μ.
z = (x − μ) ÷ σ
Turns any normal value into the standard normal, μ = 0 and σ = 1, so one table of areas works for all.
Physical meaning
Area is probability. "10.6 % of bags are underweight" is the area of the bell to the left of 500 g. Moving μ slides the whole bell; shrinking σ makes it taller and thinner, so less of it spills past a limit.

“Most data is normal.” Many business numbers are not: order values, waiting times and incomes are skewed (Lesson 3.5), and counts can be lumpy. The normal fits averages of many things (Lesson 4.3) much better than single raw values.

Watch it happen, one step at a time:

The line. Bags labelled 500 g; fills are normal with μ = 505 g, σ = 4 g.
Underweight z. (500 − 505) ÷ 4 = −1.25.
Read the area. Area to the left of z = −1.25 ≈ 0.106: one bag in ten is underweight.
Target 1 %. The 1st percentile of the standard normal is z = −2.326, so 500 must sit 2.326 σ below μ.
New centre. μ = 500 + 2.326 × 4 ≈ 509.3 g. Either give away 4.3 g a bag, or buy a machine with smaller σ.
Maths and principle

Principle: every normal curve is the standard normal shifted by μ and stretched by σ, so probabilities depend only on the z-score.

Equations: z=(x−μ)/σ; P(a≤X≤b)=Φ(b−μσ)−Φ(a−μσ), where Φ is the standard normal CDF. Inverse: xp=μ+zpσ.

Analogy: like converting every currency to dollars before comparing prices: convert every normal to z, then use one table.

5. Worked example

Made-up numbers. A packing line fills bags labelled 500 g. Fill weights are normal with μ = 505 g and σ = 4 g.

  • Underweight bags (below 500 g): z = (500 − 505) ÷ 4 = −1.25. Area left of −1.25 ≈ 10.6 %. One bag in ten is underweight.
  • To get underweight below 1 %: need 500 g to sit at z = −2.326. So μ = 500 + 2.326 × 4 ≈ 509.3 g. Either raise the target by about 4.3 g (giving away product) or reduce σ (better machines).
  • Middle 95 % of bags: 505 ± 1.96 × 4 = 497.2 g to 512.8 g.

This trade-off, shift the mean or cut the spread, is the heart of quality control (Lesson 8.6).

6. Practice

  1. Test scores are N(70, 10²). What share score above 85?
  2. Same scores: what score marks the top 10 %?
  3. Call lengths are N(6, 1.5²) minutes. What share are between 4.5 and 7.5 minutes?
Answers
  1. z = 1.5; area right of 1.5 ≈ 6.7 %.
  2. z = 1.282; 70 + 12.82 ≈ 82.8.
  3. That’s ±1σ: about 68 %.

Try it on your own data: the normal curve lab and normal lab shade any normal probability; the probability distribution explorer compares the normal with other curves.

New words: normal distribution, standard normal, 68-95-99.7 rule, Φ (normal CDF), inverse normal.

  • A normal distribution is set by μ (centre) and σ (spread).
  • About 68 %, 95 % and 99.7 % of values lie within 1, 2 and 3 σ of μ.
  • z = (x − μ) ÷ σ; probabilities are areas under the standard normal curve.
  • To hit a target share, solve x = μ + z·σ for the z of that percentile.
  • The normal fits averages and sums well; raw money and time data are often skewed.

Check yourself

1Daily orders are normal with mean 500 and SD 40. About what share of days have more than 580 orders?
2What z-values hold exactly the middle 95 % of a normal distribution?
3Why is the normal distribution so common in statistics?

Sources

Your ticks stay in this browser. Save them as a file on the course home.

Updated 2026-11-22. Numbers in the examples are made up to show the idea unless a source is given. Simulations run in your browser; nothing you type is sent anywhere.