On this concept: Formula

Not sure which lesson to do first? Find your level and get a plan →

Statistics · Grade 9-12

Empirical Rule

The Empirical Rule (the 68-95-99.

P(μ−σ<X<μ+σ)≈0.68

Quick Answer

The Empirical Rule (the 68-95-99.7 rule) says that for a normal distribution about 68% of data lies within 1 standard deviation of the mean, 95% within 2, and 99.7% within 3. Use it for quick estimates when data are bell-shaped and a boundary sits a whole number of standard deviations from the mean: with mean 100 and SD 15, about 95% of values fall between 70 and 130. The leftover share splits evenly between the two tails, so about 2.5% lie above 130.

Why This Matters

It turns a mean and a standard deviation into a picture of the data without a table. You can say what range is ordinary, recognize that a reading 3 SDs out is very rare, and estimate tail shares in your head. Quality-control charts, children's growth charts and the common 'about 2 SDs' cut-off for unusual results all rest on these numbers. Applied to skewed data, though, the same percentages are simply false.

Intuitive Explanation

Take a bell curve with mean 50 and standard deviation 10. Mark 40 and 60, one standard deviation either side: about 68% of the data, roughly two-thirds, lies between those marks. Widen to 30 and 70 (two SDs) and you capture about 95%. Widen to 20 and 80 (three SDs) and nearly everything, 99.7%, is inside.

The tails come from subtraction and symmetry. Outside ±1 SD is 100%−68%=32%, split into 16% on each side, so about 16% of values exceed 60 and about 84% lie below it. Outside ±2 SD is 5%, or 2.5% per tail; outside ±3 SD is 0.3%, or 0.15% per tail. Halves work too: between 50 and 70 is half of 95%, about 47.5%.

The percentages belong to the normal shape, not to every data set that has a mean and an SD. Suppose waits at a clinic have mean 15 minutes and SD 10 minutes. The rule would put 2.5% of waits below −5 minutes, which is impossible, so the data must be skewed and the rule does not apply. A boundary that is not a whole number of SDs, such as 'below 63' in the first example, is also outside the rule's reach; it needs a z-score and a normal table.

Think of it like this. The rule works like a recipe that scales: double the batch and every amount doubles, but the proportions stay the same. Slide the mean or stretch the standard deviation and the bell moves or widens, yet the shares stay 68, 95 and 99.7 within 1, 2 and 3 SDs. Change the shape from bell to lopsided and it is a different recipe.

Core idea. For a roughly normal distribution, the empirical rule gives the percentage of data inside one, two, and three standard deviations of the mean: about 68%, 95%, and 99.7%.


When to Use

Use it when data are called bell-shaped or approximately normal, you know the mean and standard deviation, and the boundaries sit exactly 1, 2 or 3 SDs from the mean, or the question asks for a quick estimate. Tail questions such as 'what percent lie above μ+2σ' fit too: halve the leftover share. For a boundary like 1.4 SDs, compute a z-score and use a normal table; for skewed data, use the data's own percentiles.

Tip. Ask: Is the data roughly bell-shaped, with each boundary a whole number (1, 2 or 3) of standard deviations from the mean?

How to Recognize It

Before quoting 68, 95 or 99.7, check the shape of the data and where each boundary sits compared with the mean.

  1. Is the distribution described as normal, or does its histogram look like a single symmetric mound?

    Yes allows the rule. For skewed or two-peaked data the percentages fail, so describe the data through Skewness or Distribution Shape instead.

  2. If you subtract the mean from a boundary and divide by the SD, do you get exactly 1, 2 or 3?

    Yes means you can read the share straight from 68-95-99.7. A result like 1.4 needs a table, which is the full Normal Distribution.

  3. Does the question ask for the share in one tail only, such as above the mean plus 2 SDs?

    Then halve what lies outside: 100%−95%=5%, so 2.5% sits in each tail. The share beyond an awkward value needs a Z-Score (Standard Score) and a table.

  4. Is a single value being judged as unusual, such as a box 4 SDs below the mean?

    The rule says values beyond 3 SDs are very rare, about 0.3% in all. Formally deciding which values to flag is Outlier Detection.

Empirical Rule vs Normal Distribution vs Z-Score (Standard Score) vs Skewness

The empirical rule is a shortcut inside the normal model, so problems for it look like problems for the full curve. Ask whether every boundary is a whole number of SDs, and whether the data are bell-shaped at all.

Empirical Rule

Meaning
The empirical rule says that about 68%, 95% and 99.7% of bell-shaped data lie within 1, 2 and 3 SDs of the mean.
Key test
Whole-number SD boundaries on bell-shaped data; answer from memory.
Formula
P(μ−σ<X<μ+σ)≈0.68
Example
Exam scores with mean 64 and SD 9: about 68% lie between 55 and 73.

Normal Distribution

Meaning
The normal distribution is the full bell-shaped model, which gives the share between any two values.
Key test
Awkward boundaries, such as 140 g, need the full curve and a table.
Example
For exam scores that are N(64,92), the share above 80 is about 3.8%.

Z-Score (Standard Score)

Meaning
A z-score converts a boundary into standard deviations from the mean.
Key test
The step that shows whether the rule applies: is z a whole number?
Formula
z=x−μσ
Example
A score of 82 with mean 64 and SD 9 has z=2, so the rule applies.

Skewness

Meaning
Skewness measures lopsidedness, and strongly skewed data break the 68-95-99.7 shares.
Key test
A long tail on one side means the rule's percentages cannot be trusted.
Example
Hospital stays with mean 5 days and SD 4 days must be skewed, since a bell would put 2.5% of stays below −3 days.

Formula & Notation

P(μ−σ<X<μ+σ)≈0.68
For X∼N(μ,σ2): P(μ−σ<X<μ+σ)≈0.6827, P(μ−2σ<X<μ+2σ)≈0.9545, P(μ−3σ<X<μ+3σ)≈0.9973.

Worked Examples

Example 1 — Recognize when the empirical rule applies

Easy

A class records 200 student heights. A histogram of the data is roughly bell-shaped, with mean 165 cm and standard deviation 7 cm. About what percent of students are between 158 cm and 172 cm?
Show the steps and the answer
  1. Check that the data are approximately normal.

    The empirical rule applies only to roughly bell-shaped distributions; otherwise the 68/95/99.7 numbers are unreliable.

  2. Translate the interval into standard deviations from the mean.

    The rule is stated in standard-deviation units, not raw values. 158 cm and 172 cm are exactly one standard deviation (7 cm) below and above the mean (165 cm).

  3. Read off the corresponding percentage from the rule.

    About 68% of values lie within one standard deviation of the mean for a normal distribution.

Answer: About 68% of students — roughly 136 of the 200 — have heights between 158 cm and 172 cm.

Example 2 — Avoid the nearby trap

Standard

A student sees the word “chance” in a problem about a spinner and writes "By the empirical rule, P(red) ≈ 68%." Explain why the empirical rule is the wrong tool here.
Show the steps and the answer
  1. Decide whether the situation is a normal distribution.

    The empirical rule describes how values spread around a mean for continuous, bell-shaped data. A spinner produces a discrete chance outcome, not a normal distribution.

  2. Match the tool to the structure.

    Probabilities for a spinner come from counting equally-likely outcomes or from a probability model, not from the 68-95-99.7 percentages.

Answer: The empirical rule does not apply because the spinner is not a normal distribution. The student should use a basic probability model (favourable outcomes over total outcomes), not the 68/95/99.7 percentages.

Example 3 — Find the mistake: 95% is the middle, not the tail

Application

Heights of 2,000 adult women in a survey are approximately normal with mean 163 cm and standard deviation 7 cm. Priya writes: '177 cm is 2 standard deviations above the mean, so by the empirical rule about 95% of the women are taller than 177 cm.' Find the mistake and correct it.
Show the steps and the answer
  1. Check the boundary: 163+2×7=177, so 177 cm is exactly 2 SD above the mean.

    Priya got this part right, and the heights are bell-shaped, so the rule applies.

  2. Read what the 95% covers: the heights between 163−14=149 cm and 163+14=177 cm.

    95% is the share between −2 and +2 SD, the middle of the curve. Priya used it for the tail above +2 SD.

  3. Take what is left outside, 100% minus 95%, which is 5%, and halve it: about 2.5% are taller than 177 cm.

    The normal curve is symmetric, so the tall tail and the short tail hold the same share.

  4. Count them: 2.5% of 2,000 is 0.025×2000=50 women.

    A count is easy to sense-check against the whole survey.

  5. Sense-check: 177 cm is above the mean, so fewer than half the women can be taller than it, never 95%.

    Half of a symmetric distribution lies above its mean, and any boundary above the mean has less than half beyond it.

Answer: About 2.5% of the women, roughly 50 of the 2,000, are taller than 177 cm. The 95% is the share between 149 cm and 177 cm.

Takeaway: 95% sits between −2 and +2 SD; for the share beyond 2 SD on one side, take the leftover 5% and halve it.

Where You'll See It

  • IQ scores

    IQ is built to be normal with mean 100, SD 15. The empirical rule says about 68% of people score between 85 and 115, and about 95% between 70 and 130.

  • How unusual is a tall classmate?

    If 6th-grade heights average 150 cm with SD 6 cm, then about 95% of kids are between 138 cm and 162 cm, so a kid taller than 162 cm is in the rare top 2.5%.

  • Quality control on cereal boxes

    A machine fills boxes to a mean of 500 g with SD 5 g. About 99.7% of boxes weigh between 485 g and 515 g, so a 480 g box is a red flag that the machine drifted.

Common Mistakes

  1. The slip: The 68-95-99.7 rule works for any data set with a mean and standard deviation.

    Why it's tempting: Every data set has a mean and SD, so the percentages seem universal. But the rule is derived from the normal curve and only holds when data is roughly bell-shaped.

    Instead: Only apply 68-95-99.7 when the distribution is approximately normal. For skewed or non-normal data, use the actual data, a histogram, or Chebyshev's inequality instead.

  2. The slip: 95% of the data falls within ONE standard deviation of the mean.

    Why it's tempting: 95 is the most memorable number, so students attach it to the simplest case (one SD). But 95% goes with TWO SDs; one SD is 68%.

    Instead: Match the level to the count: 1 SD → ~68%, 2 SD → ~95%, 3 SD → ~99.7%. Within one SD is 68%, not 95%.

  3. The slip: To find the percent ABOVE mean + 2 SD, use 95%.

    Why it's tempting: The 95% figure is tied to '2 SD,' so it feels like the answer to any 2-SD question. But 95% is the area BETWEEN −2 and +2 SD, not above +2 SD.

    Instead: Within ±2 SD is 95%, so 5% is outside, split evenly into two tails: about 2.5% above +2 SD and 2.5% below −2 SD.


Mini Practice

Try these on your own. Tap Reveal when you want to check.

  1. Adult resting heart rates are normal with mean 72 bpm and standard deviation 8 bpm. About what percent of adults have a resting heart rate between 64 and 80 bpm?

    Hint: How many standard deviations from the mean are 64 and 80?

  2. Exam scores are bell-shaped with mean 500 and standard deviation 100. About what percent of students score above 700?

    Hint: 700 is how many standard deviations above the mean? Then split the leftover percent.

  3. Cereal boxes are filled with a normal amount, mean 500 g and standard deviation 4 g. About what percent of boxes hold between 488 g and 512 g?

    Hint: Divide the distance from the mean by the standard deviation.

  4. Battery life is normal with mean 40 hours and standard deviation 5 hours. About what percent of batteries last between 35 and 50 hours?

    Hint: Split the interval at the mean and handle each side separately.

  5. Household incomes have a long right tail: most are moderate, but a few are enormous. Can you use the 68-95-99.7 rule to estimate the middle 95% of incomes?

    Hint: What shape of distribution does the rule need?

  6. Commute times are normal with mean 12 min and standard deviation 2 min. A student says: "About 95% of commutes take between 10 and 14 minutes." Correct this.

    Hint: How many standard deviations from the mean are 10 and 14?

Want the full set?

50 practice questions for this concept — free to try, every one with a complete worked solution showing the why, not just the answer.

Frequently Asked Questions

What can I say about data that are not bell-shaped?

Chebyshev's rule works for any shape at all, but it gives only guaranteed minimums. At least 75% of the values lie within 2 SDs of the mean, and at least about 89% within 3 SDs. Compare that with 95% and 99.7% for bell-shaped data. The guarantee is weaker because it has to hold even for strange, lopsided data sets.

Where do 68, 95 and 99.7 come from, and are they exact?

They are areas under the normal curve, rounded. More precisely, 68.27% lies within 1 SD, 95.45% within 2 SDs and 99.73% within 3 SDs. The exact distance that holds 95% is 1.96 SDs, not 2, which is why confidence intervals and margins of error use 1.96. For quick estimates, the rounded 68-95-99.7 values are good enough.

If 99.7% is within 3 SDs, why worry about the other 0.3%?

Because in large numbers it adds up. A factory making a million bolts a day would expect about 2,700 of them beyond 3 SDs, even when the machine works perfectly. For medicine doses or aircraft parts that is far too many, so engineers set tighter limits or aim for much smaller spreads. The tails never quite reach zero, so rare values still turn up when you look at enough data.

Learning Path

Empirical Rule

You are here

Next →

Nothing after this yet
This builds directly on the Normal Distribution, the bell curve fixed by a mean and a standard deviation. The empirical rule adds three remembered areas that replace a table whenever a boundary is a whole number of SDs. With it you can flag unusual measurements in quality control and estimate tail shares quickly, and later in statistics the '95% within about 2 SDs' fact underlies 95% confidence intervals.

See Also