Z-score vs percentile: the difference

By Jude Wallis · Published

A z-score says how many standard deviations a value is above or below the mean. A percentile says what percent of the values are at or below it. Converting between them requires knowing the shape of the distribution: under a normal model, z = 1.5 is about the 93rd percentile.

In the Fall 2026 AP Statistics course, standardizing a value with a z-score is part of Unit 1 topic 1.9, Comparisons of the Distributions for One Quantitative Variable, and percentiles sit with the summary statistics of topic 1.7. Converting between a z-score and a percentile with a normal model is Unit 2 topic 2.11, The Normal Distribution, so this page spans Units 1 and 2.

Z-score vs percentile: the short answer

Both locate one value within a distribution, and they measure location differently. A z-score measures distance: how far the value is from the mean, in units of standard deviations. A percentile measures rank: what percent of the distribution is at or below the value.

Distance and rank are connected only through the shape of the distribution. When a normal model applies, each zz-score corresponds to one percentile and a table or calculator converts between them. When the distribution is not normal, the zz-score and the percentile are both still valid descriptions, but the normal conversion gives wrong answers. The second worked example has a value with z≈0.27z \approx 0.27 that sits at the 90th percentile of its data set, where a normal model would put it near the 61st.

What a z-score tells you

The zz-score of a value xx is

z=x−μσz = \frac{x - \mu}{\sigma}

where μ\mu ("mu") is the mean and σ\sigma ("sigma") is the standard deviation of the distribution. With sample data, use xˉ\bar{x} and ss in their place. The sign gives the direction: positive for a value above the mean, negative for one below, and zero for a value equal to the mean. The size gives the distance in standard deviations.

A zz-score has no units, which makes it useful for comparing values from different scales, such as a score on one exam with a score on another. It can be computed for any quantitative distribution. Nothing in the formula requires the distribution to be normal. The details are in how to find a z-score, and the z-score calculator does the arithmetic.

What a percentile tells you

A percentile is a value with a stated percent of the distribution at or below it. Saying that a score of 650 is at the 93rd percentile means about 93% of scores are 650 or lower. In a data set, the percentile rank of a value is found by counting:

percentile rank of x=number of values at or below xn×100\text{percentile rank of } x = \frac{\text{number of values at or below } x}{n} \times 100

On a density curve it is the area to the left of the value, written as a percent.

A percentile uses only the order of the values. It does not depend on how far apart they are, so it is unaffected by how extreme the largest or smallest values are. It also needs no assumption about shape: counting works for any data set. A percentile rank runs from just above 0 to 100 and is never negative, unlike a zz-score. More on reading percentiles is in what does percentile mean.

The differences side by side

FeatureZ-scorePercentile
MeasuresDistance from the mean, in standard deviationsPercent of values at or below
Computed fromThe mean and the standard deviationThe order of the values, or area under a curve
Possible valuesAny real number, negative or positiveA percent from just above 0 to 100
Value at the mean of a normal distribution050th
Affected by extreme valuesYes, through the mean and standard deviationNo, only order matters
Needs a normal modelNoNo
Converting to the otherNeeds the shape of the distributionNeeds the shape of the distribution

Each one can be found without any assumption about shape. The assumption enters only when you convert between them.

Converting between them with a normal model

When a distribution is approximately normal, the percentile of a value is the area under the standard normal curve to the left of its zz-score. The z-table or the normal distribution calculator gives that area.

zz-scoreArea to the leftPercentile
−2-20.0228about 2nd
−1-10.1587about 16th
000.500050th
110.8413about 84th
220.9772about 98th

The relationship is not linear. Moving from z=0z = 0 to z=1z = 1 gains about 34 percentile points, and moving from z=1z = 1 to z=2z = 2 gains only about 14. Equal steps in zz are equal steps in the raw value, but they pass very different numbers of observations, because a normal distribution is crowded near its mean and sparse in its tails.

The conversion also runs backward. For the 80th percentile, find the zz-score with area 0.80 to its left, which is about 0.84, then convert to a raw value with x=μ+zσx = \mu + z\sigma. With mean 500 and standard deviation 100, the 80th percentile is about 500+0.84(100)=584500 + 0.84(100) = 584.

The classic mix-up and how to avoid it

The first mistake is using the normal table on a distribution that is not normal. The table turns a zz-score into a percentile only for a normal curve. For skewed data, compute the percentile by counting, and report the zz-score separately if it is asked for.

The second mistake is reading a zz-score as a percent. A zz-score of 1.5 does not mean 1.5% or 15% of anything. It means 1.5 standard deviations above the mean, which under a normal model is about the 93rd percentile.

The third mistake is treating percentile differences as equal distances. Under a normal model with mean 500 and standard deviation 100, the 50th percentile is 500 and the 84th is about 600, a gap of 100 points for 34 percentile points. From the 84th to the 98th percentile is also about 100 points, for only 14 percentile points. To compare how far apart two values are, use the raw values or their zz-scores.

From a score to a z-score to a percentile

Scores on an exam are approximately normal with mean 500 and standard deviation 100. A student scores 650. Find the z-score and the percentile.

  1. Compute the zz-score: z=x−μσ=650−500100=1.5z = \frac{x - \mu}{\sigma} = \frac{650 - 500}{100} = 1.5.

  2. Interpret it: the score is 1.5 standard deviations above the mean.

  3. Because the distribution is approximately normal, the percentile is the area to the left of z=1.5z = 1.5 under the standard normal curve. From the table, that area is 0.9332.

  4. Convert to a percent: 0.9332×100=93.320.9332 \times 100 = 93.32, so the score is at about the 93rd percentile.

  5. Check for sense: a positive zz-score should give a percentile above 50, and 1.5 is between 1 (about the 84th percentile) and 2 (about the 98th). The 93rd fits.

The zz-score is 1.5 and the score is at about the 93rd percentile. About 93% of scores are 650 or lower. The zz-score came from the mean and standard deviation alone; turning it into a percentile used the normal model.

A skewed data set where the normal conversion fails

Ten annual incomes, in thousands of dollars, are 30, 32, 35, 38, 40, 42, 45, 50, 58, 130. For the income of 58, find the z-score and the percentile rank within this data set. Then find the percentile a normal model would assign to that z-score.

  1. Mean: the sum is 30+32+35+38+40+42+45+50+58+130=50030 + 32 + 35 + 38 + 40 + 42 + 45 + 50 + 58 + 130 = 500, so xˉ=50010=50\bar{x} = \frac{500}{10} = 50.

  2. Deviations from 50: −20,−18,−15,−12,−10,−8,−5,0,8,80-20, -18, -15, -12, -10, -8, -5, 0, 8, 80. Squares: 400,324,225,144,100,64,25,0,64,6400400, 324, 225, 144, 100, 64, 25, 0, 64, 6400, which sum to 7746.

  3. Standard deviation: s=77469=860.67≈29.34s = \sqrt{\frac{7746}{9}} = \sqrt{860.67} \approx 29.34.

  4. The zz-score of 58: z=58−5029.34≈0.27z = \frac{58 - 50}{29.34} \approx 0.27.

  5. Percentile rank by counting: nine of the ten incomes are at or below 58, so the rank is 910×100=90\frac{9}{10} \times 100 = 90.

  6. Normal conversion: the area to the left of z=0.27z = 0.27 under the standard normal curve is 0.6064, which would be about the 61st percentile.

  7. Compare: the count says 90th percentile and the normal model says 61st. The data are strongly skewed right by the income of 130, which inflates both the mean and the standard deviation, so the normal model does not describe them.

The income of 58 has z≈0.27z \approx 0.27 and is at the 90th percentile of this data set. A normal model would place z=0.27z = 0.27 at about the 61st percentile, which is wrong here by 29 percentile points. Both the zz-score and the percentile are correct descriptions; the error is in converting between them with a normal curve when the data are skewed.

Frequently asked questions

What percentile is a z-score of 0?

Under a normal model, the 50th. A zz-score of 0 is a value equal to the mean, and a normal distribution is symmetric, so half the area lies to its left. In a skewed distribution the mean and the median usually differ, so a zz-score of 0 can be above or below the 50th percentile.

Can I convert a z-score to a percentile without a normal distribution?

Not with a z-table. The table gives areas under a normal curve only. For a data set that is not approximately normal, find the percentile rank directly by counting the values at or below the one you care about.

Can a percentile be negative like a z-score?

No. A percentile rank is a percent of the data, so it lies between 0 and 100. A negative zz-score corresponds to a percentile below the 50th under a normal model. For example, z=−1z = -1 is about the 16th percentile.