How to Calculate Percentage Error
Percentage error measures how far an experimental or measured value is from an accepted, true, or expected value, as a percentage of that accepted value. It is a standard way to report the size of a discrepancy in a lab, a calibration check, or any setting where one number is treated as the reference.
It is not the same as percentage change, which compares a new value with an original value, and it is not the same as percentage difference, which uses the midpoint of two values and does not treat either one as the “correct” figure. Those distinctions matter if you pick a formula from a dropdown and get a number that looks plausible but answers the wrong question.
This site does not have a dedicated percentage-error calculator. The arithmetic is short enough to do by hand, and the sections below include fully worked numbers. If you need a different comparison, use the percentage difference calculator or the percentage change calculator instead, and use them only for their own formulas.
The percentage-error formula
- The *accepted* value is the reference: a known constant, a manufacturer’s figure, a textbook value, or an agreed target.
- The *experimental* value is the one you measured or observed.
- The absolute values make the error a magnitude. A result 2% high and a result 2% low have the same percentage error if you use this definition. If you need the direction (too high or too low), report a signed error as well, or state the two raw values.
The accepted value must not be zero. Dividing by zero is undefined, and a zero reference cannot be the base of a percentage error. In that case, report the absolute error (the difference itself) and say that a percentage error is not defined.
Step-by-step
- Identify the accepted value and the experimental value. Keep the units the same.
- Subtract, and take the absolute value if you want the usual unsigned percentage error.
- Divide by the absolute value of the accepted value.
- Multiply by 100.
Example: a measured 9.8 against an accepted 10
- Difference: |9.8 − 10| = 0.2.
- 0.2 ÷ 10 = 0.02.
- 0.02 × 100 = 2.
The percentage error is 2%. The measured value is a little low, which you can say in words; the 2% itself is the size of the gap relative to 10.
Example: 48 against an accepted 50
- |48 − 50| = 2.
- 2 ÷ 50 = 0.04.
- 0.04 × 100 = 4.
The percentage error is 4%. Notice that the raw difference here (2) is larger than in the previous example (0.2), but the accepted bases are different (50 versus 10). Percentage error scales the gap by the reference, so the two experiments can be compared even though the units and sizes differ.
Example: 102 against an accepted 100
- |102 − 100| = 2.
- 2 ÷ 100 = 0.02.
- 0.02 × 100 = 2.
The percentage error is 2%, the same size as the 9.8-versus-10 case, even though one measurement was below the reference and this one is above it. If direction matters, write “2% high” or report a signed error of +2%.
Signed error, when you need a direction
A signed (or relative) error keeps the sign of (experimental − accepted):
For 9.8 versus 10, that is −2%. For 102 versus 100, it is +2%. Some lab notebooks want this signed form; some mark schemes want the unsigned form. Use the definition you have been asked for, and label it.
How this differs from percentage change and percentage difference
- Percentage change uses an original value as the base and is directed from old to new: ((new − original) ÷ original) × 100. If a reading goes from 50 to 48, the change is −4%, which happens to match the 4% error in the second example only because the “accepted” and the “original” were the same number. If they are not, the results diverge. See How to calculate a percentage for the basic part-and-whole idea, and the percentage change calculator for the change formula.
- Percentage difference uses the midpoint of two values: (|A − B| ÷ ((|A| + |B|) ÷ 2)) × 100. Neither value is the official reference. For 48 and 50, the midpoint is 49, and the percentage difference is 2/49 × 100 ≈ 4.0816%, not 4%. That is a different question, and it is the one the percentage difference calculator answers.
If you have a published true value, use percentage error, not percentage difference. If you have two lab teams’ measurements and no official true value, percentage difference (or a plain difference with units) is the more honest tool.
Percentage error is not a verdict on quality by itself
A 2% error on one experiment and a 2% error on another are comparable only if the accepted values are the relevant references and the measurements are of the same kind of quantity. A 2% error might be excellent in one context and poor in another. This article does not invent typical lab tolerances or industry “acceptable” ranges, because those limits come from a method, a standard, or an instructor, not from the formula.
Report the accepted value, the measured value, and the percentage error together. Anyone can then check the working and apply whatever tolerance they were given.
Common mistakes
- Putting the experimental value in the denominator. The accepted value is the base.
- Dropping the absolute value when the mark scheme asked for unsigned error, or keeping it when a signed error was required.
- Using the midpoint (percentage difference) when a true value is known.
- Computing a percentage error when the accepted value is 0. The formula is undefined there.
- Rounding too early. If 58 and 70 were a measured and accepted pair, 58/70 is not involved; the error would be |58 − 70|/70 × 100 ≈ 17.142857%. Round only at the end, and say how you rounded.
Frequently asked questions
How do you calculate percentage error?
Take the absolute difference between the experimental value and the accepted value, divide by the absolute accepted value, and multiply by 100.
What is the percentage error if I measure 9.8 and the accepted value is 10?
2%. The working is |9.8 − 10| ÷ 10 × 100 = 2.
Is percentage error the same as percentage difference?
No. Percentage error uses a known accepted value as the base. Percentage difference uses the midpoint of two values and treats them symmetrically.
Can percentage error be greater than 100%?
Yes. If the measured value is 0 and the accepted value is 50, the percentage error is 100%. If the measured value is 0 and the accepted value is 10, it is also 100%. If the measured value is 30 and the accepted value is 10, the percentage error is 200%.
What if the accepted value is zero?
A percentage error is not defined, because the formula would divide by zero. Report the absolute difference instead.
