How to Read a Z-Score Without the Jargon

Fifteen points above a mean of 70

A test has a mean of 70 and a standard deviation of 10. A score of 85 sits 15 points above the mean. Divide that gap by 10 and the z-score is 1.5. The 1.5 means one and a half standard deviations above the center, which on this test is exactly those 15 points. It does not mean the student scored 1.5 points, and it does not mean 1.5 percent. A score of 55 sits 15 points below the mean, so its z-score is negative 1.5. The sign is the direction. The absolute value is the distance, measured in spreads rather than in raw points.

Change the spread and the same raw gap becomes a different z. If the standard deviation were 5 instead of 10, a score of 85 would be 15 divided by 5, a z of 3. The student did not gain points. The class got tighter, so the same 15-point lead is farther out in spread-units. Always write the mean and the standard deviation next to a z, or nobody can turn it back into a score. Here, 70 plus 1.5 times 10 returns 85. Drop the 10 and the return trip is impossible.

Zero, and the scores that land on it

A z of 0 is the mean itself. On this test, 70 plus 0 times 10 is 70. Students sometimes read a z of 0 as a failure, because zero looks like no credit. It is not a score of zero. It is an average performance on the scale you just standardized. If the mean is a respectable 70, a z of 0 is that respectable result. Say so when you hand z-scores back. Otherwise a student at the class mean will think the conversion erased their paper. Keep the unit, the count, and the name of the formula in the same sentence as any figure you copy into a report.

The Z-Score Calculator takes a raw score, a mean, and a standard deviation and returns the signed count of spreads. Feed it 62, 70, and 10. The gap is negative 8, and negative 8 divided by 10 is negative 0.8. That student is a bit under the mean, not a full spread under it. A cutoff written as z below negative 1 would not catch this paper. A cutoff written as 60 raw points would, because 62 is above 60. Convert the cutoff and the paper into the same language before you apply a rule.

Two classes on one scale

Class A has mean 70 and standard deviation 10. Class B has mean 62 and standard deviation 8. One student scores 78 in A and another scores 74 in B. The A z-score is 78 minus 70, divided by 10, which is 0.8. The B z-score is 74 minus 62, divided by 8, which is 1.5. The higher raw score is the smaller z. If the prize is for standing out inside a class, B's student leads. If the prize is raw points, A's student scored more. Compare z-scores only when each uses its own mean and spread.

Do not pool the papers and also keep the separate z-scores in one ranking without saying so. A z of 0.8 and a z of 1.5 are distances on different rulers. They can be compared as distances, and they are not raw points. A selection rule that adds a z to a raw score, 0.8 plus 78, mixes a count of spreads with a count of points and produces a number with no unit. Keep a column of raw scores and a column of z-scores. Apply the rule to one column. If both matter, write a weight, and show a worked row so the weight is visible.

What a large z asks you to check

A z of 4 on a mean of 70 and a standard deviation of 10 is a raw score of 110. If the test is out of 100, 110 is impossible, so the z is a data error: a wrong mean, a wrong spread, or a typed score. A z of 4 on a test that really does run past 100 is rare under a bell-shaped model, about a few per 100,000 in one tail of a normal curve, and it is not impossible in real marks. Check the arithmetic before you celebrate or investigate the student. Most extreme z-scores in a classroom spreadsheet are transposition errors.

A negative z of the same size is the same check in the other direction. Z equals negative 4 would be 70 minus 40, a raw score of 30. If the rest of the class sits between 60 and 85, a 30 is either a blank converted to a low number or a genuine crash. Look at the paper. The z-score did its job when it made that row visible. It is not a diagnosis of the student. Report the raw score with the z so a counselor sees 30 points, not only a minus 4 that could have come from any scale.

Z-scores do not create a bell

Standardizing a list gives it a mean of 0 and a standard deviation of 1. It does not make the shape normal. A skewed set of incomes, once converted to z-scores, is still skewed. The household at z equals 3 is still the long right tail. You may not quote a normal-curve percentage just because you now have a z. The percentage comes from a model you must be willing to defend, or from a count you make on the actual list. If 2 of 50 scores exceed z equals 2, that is 4 percent in this list, which may not match a textbook tail.

Use the z as a unit conversion first. On the test, every point is a tenth of a standard deviation, because the spread is 10. A three-point item is 0.3 of a z. That is a useful way to talk about whether a question is large relative to the class spread. It still does not say what fraction of students missed it. Count that fraction directly. Recommend the z-score when you are lining up different scales, and recommend a plain percent when you are describing this class alone and the class is the whole story.

A short report that keeps the return trip

Write raw score, mean, standard deviation, and z in one line. For the 85: raw 85, mean 70, standard deviation 10, z equals 1.5. A reader can divide 15 by 10 and can also multiply back. For the cross-class example, give both lines, and state that the z uses each class's own mean and spread. If you round the z to 2, say so. Rounding 1.5 to 2 would turn the score back into 90, a five-point fiction. Two decimals are enough for most grade conversations. More decimals pretend the mean was known to a precision the class did not have.

If you standardize a whole column, check two summaries before you circulate the file. The mean of the z-scores should be 0, apart from rounding, and the standard deviation of the z-scores should be 1, using the same divisor you used to build them. If the standard deviation of the new column is 10, you forgot to divide by the spread and you have only centered the scores. That mistake is common and invisible if you glance at signs alone. A column of centered scores is useful, and it is not a column of z-scores until the division is done.

Open the worksheet

The matching page is the Z-Score Calculator. It opens with a worked example, names the formula, and keeps the steps next to the result. Use it when you want the arithmetic on your own numbers rather than the argument on this page.