The formulas
Percent error says how far a measured value is from the true value, as a percentage of the true value. It is built up from two simpler quantities.
Absolute error = |measured − true|
Relative error = |measured − true| ÷ |true|
Percent error = relative error × 100
Here “measured” is your experimental, estimated or observed value, and “true” is the accepted, theoretical or reference value. The vertical bars mean the size of a number without its sign. The absolute error has the same unit as the measurement. The relative and percent errors have no unit, which is what allows errors in quite different quantities to be compared.
When there is no accepted value, for example when two students measure the same thing, use the percent difference, which divides by the average of the two so that neither is treated as correct.
Percent difference = |a − b| ÷ |(a + b) ÷ 2| × 100
Here a and b are the two measurements, in either order.
Worked example
A pendulum experiment gives the acceleration due to gravity as 9.6 m/s². The accepted value is 9.81 m/s².
- Subtract. 9.6 − 9.81 = −0.21. The minus sign says the measurement is too low.
- Absolute error. Drop the sign: 0.21 m/s².
- Relative error. 0.21 ÷ 9.81 = 0.0214.
- Percent error. 0.0214 × 100 = 2.14%.
If a second group measures 9.9 m/s² and you want to compare the two groups with each other, use the percent difference. The gap is 9.9 − 9.6 = 0.3 and the average is (9.6 + 9.9) ÷ 2 = 9.75, so the percent difference is 0.3 ÷ 9.75 × 100 = 3.08%.
Which measure to use
- Absolute error when the size of the miss in real units is what matters: a shelf cut 2 mm short, a clock 30 seconds fast.
- Relative or percent error when the miss has to be judged against the size of the thing measured. An error of 1 cm is 10% of a 10 cm length but only 0.01% of a 100 m length.
- Percent difference when comparing two measurements, two instruments or two estimates and neither is the reference.
- Percentage change, a different calculation again, when a single quantity moves from an old value to a new one over time. Use the percentage calculator for that.
The choice of denominator is not a detail. The same two numbers, 9.6 and 9.9, give 3.03% if 9.9 is taken as true, 3.125% if 9.6 is taken as true, and 3.08% as a percent difference. Always say which one you are reporting.
Mistakes and edge cases
Dividing by the measured value
The denominator is the true value, not the measured one. Swapping them gives a slightly different answer that looks plausible, which is why it often goes unnoticed.
Signed or unsigned
Most textbooks define percent error with absolute value bars, so it is never negative. Some courses and many laboratories keep the sign, because a consistent negative error across several trials points to a systematic cause such as friction or a miscalibrated instrument. The calculator gives the unsigned figure by default and the signed one if you tick the box.
A true value of zero, or close to it
Percent error cannot be worked out when the true value is zero, because that would mean dividing by zero. When the true value is merely small, the percentage balloons: a reading of 0.3 against a true value of 0.1 is a 200% error even though the absolute error is only 0.2. In those cases the absolute error is the more honest figure. A percent error above 100% is not a mistake in itself; it simply means the miss is larger than the true value.
Scales with an arbitrary zero
Percent error only makes sense on a scale where zero means “none”. Reading 21 °C when the true temperature is 20 °C looks like a 5% error, but the same two temperatures in kelvin are 294.15 and 293.15, an error of 0.34%. The absolute error, 1 degree, is the same either way. For temperatures in Celsius or Fahrenheit, quote the absolute error.
Percent error is not the same as tolerance
A resistor marked 100 ohms ± 5% may be anywhere from 95 to 105 ohms. If yours measures 103 ohms, its percent error from the nominal value is 3%, which is within tolerance. The tolerance is the limit the maker allows; the percent error is what you actually found.