Concepts · 4 min read
Z-Scores and the Normal Distribution: A Plain-English Guide

Why z-scores exist
Raw numbers are hard to compare across contexts. Is 86 on a test good? It depends on how the rest of the class did. Is a height of 190 cm tall? It depends on the population. A z-score removes the units by measuring distance from the average in standard-deviation steps, so results from different scales can be compared directly.
To follow this you need two ideas. The mean is the average value. The standard deviation measures the typical spread around that mean. A z-score simply combines them.
The z-score formula
z = (x − μ) ÷ σ, where x is the value, μ is the mean and σ is the standard deviation. A positive z means above average, a negative z below average, and zero means exactly average.
z-score for a result of 86 when the mean is 70 and the standard deviation is 8
Order-of-operations steps 2 steps
Start: (86 − 70) ÷ 8
- Subtract
86 − 70 = 16→ 16 ÷ 8 - Divide
16 ÷ 8 = 2→ 2
So 86 is two standard deviations above the mean. A result of 62 would give z = (62 − 70) ÷ 8 = −1, one standard deviation below.
You can find the mean and standard deviation of your own data with the mean, median and mode calculator or the standard deviation calculator, then plug them into the formula.
The normal distribution and the 68-95-99.7 rule
Many measurements, such as heights, measurement errors and averages of many small effects, form a bell-shaped curve called the normal distribution. It is symmetric around the mean, and its shape is fully set by the mean and the standard deviation.
For a normal distribution, the share of values within a given number of standard deviations of the mean is fixed. This is the empirical rule, often written 68-95-99.7.
| Within | Share of values | Outside (both tails) |
|---|---|---|
| ±1 standard deviation | 68.27% | 31.73% |
| ±2 standard deviations | 95.45% | 4.55% |
| ±3 standard deviations | 99.73% | 0.27% |
Roughly two thirds of values sit within one standard deviation, about 95% within two, and almost everything within three. That last fact is why a value beyond z = 3 is considered unusual: only about 0.27% of values fall that far from the mean in either direction.
From a z-score to a percentile
A z-score becomes a percentile through the cumulative distribution function, usually written Φ(z). It gives the proportion of a normal distribution that lies below z. Statistical tables list it, and software computes it directly; you do not need to do the integral by hand.
| z-score | Share below (percentile) | Plain reading |
|---|---|---|
| -2 | 2.28% | Far below average |
| -1 | 15.87% | Below average |
| 0 | 50% | Exactly average |
| 0.5 | 69.15% | Slightly above average |
| 1 | 84.13% | Above average |
| 1.645 | 95% | Top 5% boundary |
| 2 | 97.72% | High |
| 3 | 99.87% | Very high |
For the exam example, z = 2 corresponds to the 97.7th percentile. The same z-score from a different exam would mean the same relative standing, even if the raw scores looked nothing alike.
Comparing two different scales
Suppose a student scores 86 on a maths test (mean 70, SD 8) and 71 on a science test (mean 60, SD 5). Which result is stronger relative to the class?
| Test | Score | Mean | SD | z-score |
|---|---|---|---|---|
| Maths | 86 | 70 | 8 | 2 |
| Science | 71 | 60 | 5 | 2.2 |
The maths result has z = 2.00 and the science result z = 2.20, so the science score is slightly stronger relative to its group even though the raw number is lower. This kind of comparison is the main practical use of z-scores.
When the normal model does not fit
- Skewed data. Incomes and waiting times have long right tails, so the 68-95-99.7 shares do not hold. A median can be more honest, as in mean versus median.
- Small samples. With only a handful of values, the sample mean and SD are unreliable estimates.
- Outliers. One extreme value can inflate the standard deviation and make everything else look closer to average.
- Bounded scales. Percentage scores cannot go beyond 100%, so the far tail may be cut off.
Check the shape of your data with a quick histogram before reading percentiles from a z-score. If it is clearly lopsided, quote the raw percentile of the data instead.
Frequently asked questions
Can a z-score be negative?
Yes. A negative z-score means the value is below the mean.
What z-score is unusual?
Many people treat |z| above 2 as unusual and above 3 as very unusual, but this is a convention, not a law.
Do z-scores only work for normal data?
The z-score itself can be calculated for any data. The percentile interpretation and the 68-95-99.7 rule assume a roughly normal distribution.
What is the difference between a z-score and a standard deviation?
A standard deviation measures spread in the original units. A z-score uses the standard deviation as a ruler to say how far one value is from the mean.
Keep reading
Standard Deviation Explained With a Worked Example
What standard deviation measures, how to calculate it step by step, the sample vs population difference and the 68-95-99.7 rule.
ConceptsMean vs Median: When Averages Lie
How a single extreme value changes the mean but barely moves the median, with a worked data set and a simple rule for choosing which average to report.
ConceptsFibonacci Sequence and the Golden Ratio, Computed
The first 20 Fibonacci numbers, how their ratios converge to the golden ratio φ ≈ 1.618034, and Binet’s formula — with every value computed.
ConceptsLogarithms Explained Simply: What They Are and Why They Matter
Logarithms as “the power you need”, with tables for base 10 and base 2, the product rule, and real uses like pH and the Richter scale.