Skip to main content
ponderloop

Percent Error Calculator

Percent error is |measured − accepted| ÷ |accepted| × 100. For a measured value of 9.52 against an accepted value of 9.81 it is 2.96%: the measurement is 0.29 below the accepted value, which is 2.96% of it. The signed version, (measured − accepted) ÷ |accepted| × 100, is −2.96%, the absolute error is 0.29 and the relative error is 0.0296. Percent difference is a different measure, 3.00% for the same two numbers: it divides by their mean and is for two values when neither is the reference. Switch to several measurements to get each error and the mean absolute percent error against one accepted value.

Updated Runs in your browser: nothing you enter is uploadedHow this tool works

Type the value you measured and the accepted (true, theoretical or reference) value. Negatives and E notation such as 6.5e-3 work.

|measured − accepted| ÷ |accepted| × 100 = |9.52 − 9.81| ÷ |9.81| × 100

Percent error 2.96%

The measured value 9.52 is 0.29 below the accepted value 9.81, which is 2.96% of the accepted value.

Percent error
2.96%
Signed percent error
−2.96%
Absolute error
0.29
Signed error (measured − accepted)
−0.29
Relative error
0.0296
Percent difference (a different measure)
3.00%

Rounded to 3 significant figures, the fewest in your values. Absolute and signed errors are exact differences and are not rounded.

The working

  1. Error = measured − accepted = 9.52 − 9.81 = −0.29.
  2. Absolute error = |−0.29| = 0.29.
  3. Relative error = absolute error ÷ |accepted| = 0.29 ÷ 9.81 ≈ 0.0296.
  4. Percent error = relative error × 100 = 0.29 ÷ 9.81 × 100 ≈ 2.96%.
  5. Percent difference (for two values with no reference) = |9.52 − 9.81| ÷ mean × 100 = 0.29 ÷ 9.665 × 100 ≈ 3.00%.

Error, relative error, percent error and percent difference compared

Each row is the calculator’s own answer for a measured 9.52 and an accepted 9.81.
MeasureFormulaDivides byUse it whenResult for 9.52 and 9.81
Signed errormeasured − acceptedNothing (keeps the unit)You need to know whether you read high or low−0.29
Absolute error|measured − accepted|Nothing (keeps the unit)You want the size of the miss in the quantity’s own unit0.29
Relative error|measured − accepted| ÷ |accepted|The accepted valueYou want the miss as a fraction of the true value0.0296
Percent errorrelative error × 100The accepted valueA measurement is compared with a known, accepted value2.96%
Signed percent error(measured − accepted) ÷ |accepted| × 100The accepted valueYou want percent error with its direction−2.96%
Percent difference|a − b| ÷ ((a + b) ÷ 2) × 100The mean of the two valuesTwo values are compared and neither is the reference3.00%

Percent error worked examples

Each row is the calculator’s own answer for the measured and accepted values in the first two columns.
MeasuredAcceptedAbsolute errorPercent errorSigned percent errorPercent difference
9.529.810.292.96%−2.96%3.00%
3.143.141590.001590.0506%−0.0506%0.0506%
98.6100.01.41.40%−1.40%1.41%
15001480201.4%1.4%1.3%
0.00520.00500.00024.0%4.0%3.9%
-4.9-5.00.12.0%2.0%2.0%
25.010.015150%150%85.7%
505Undefined (accepted value is 0)Undefined (accepted value is 0)200%

Percent error for measured values against an accepted 50.0

Measured values from 40.0 to 60.0 against an accepted value of 50.0, shown to 3 significant figures.
MeasuredSigned errorPercent errorSigned percent errorPercent difference
40.0−1020.0%−20.0%22.2%
45.0−510.0%−10.0%10.5%
48.0−24.00%−4.00%4.08%
49.0−12.00%−2.00%2.02%
49.9−0.10.200%−0.200%0.200%
50.50.51.00%1.00%0.995%
52.024.00%4.00%3.92%
55.0510.0%10.0%9.52%
60.01020.0%20.0%18.2%

The same 1-unit miss against accepted values of different size

A measured value 1 above the accepted value, for accepted values from 2 to 1000, shown to 3 significant figures.
AcceptedMeasuredAbsolute errorRelative errorPercent error
2.003.0010.50050.0%
5.006.0010.20020.0%
10.011.010.10010.0%
50.051.010.02002.00%
10010110.01001.00%
50050110.002000.200%
1000100110.001000.100%

Percent error of 9.52 against 9.81 at each rounding setting

The same two numbers under each display setting of the calculator; absolute and signed errors are never rounded.
SettingPercent errorSigned percent errorRelative errorPercent difference
Automatic: the fewest in your values2.96%−2.96%0.02963.00%
2 significant figures3.0%−3.0%0.0303.0%
3 significant figures2.96%−2.96%0.02963.00%
4 significant figures2.956%−2.956%0.029563.001%
5 significant figures2.9562%−2.9562%0.0295623.0005%
6 significant figures2.95617%−2.95617%0.02956173.00052%
Full precision (up to 12 digits)≈ 2.95616717635%≈ −2.95616717635%≈ 0.0295616717635≈ 3.00051733057%

Errors of the 5 example measurements against 9.81

The list mode’s own table for the measurements 9.78, 9.81, 9.85, 9.74, 9.52, shown to automatic significant figures.
#MeasuredErrorPercent errorAbsolute percent error
19.78−0.03−0.306%0.306%
29.8100%0%
39.850.040.408%0.408%
49.74−0.07−0.714%0.714%
59.52−0.29−2.96%2.96%

How the list results differ for five kinds of data against an accepted 10

Each row is a set of four measurements compared with an accepted value of 10, shown to 3 significant figures.
MeasurementsValuesMean absolute percent errorBias (mean signed percent error)Percent error of the meanLargest percent error
Scattered around 10, no bias9.5, 10.5, 9.8, 10.23.50%0%0%5.00% (measurement 1: 9.5)
Mostly high10.4, 10.6, 10.5, 10.34.50%4.50%4.50%6.00% (measurement 2: 10.6)
Mostly low9.4, 9.6, 9.5, 9.74.50%−4.50%4.50%6.00% (measurement 1: 9.4)
One wild reading10.1, 9.9, 10.0, 12.05.50%5.00%5.00%20.0% (measurement 4: 12.0)
Tight and on target10.01, 9.99, 10.02, 9.980.150%0%0%0.200% (measurement 3: 10.02)

How to calculate percent error

  1. Find the error. Subtract the accepted value from the measured value: 9.52 − 9.81 = −0.29. A negative result means the measurement is below the accepted value.
  2. Take the absolute value. Drop the sign to get the absolute error: |−0.29| = 0.29.
  3. Divide by the accepted value. Divide the absolute error by the absolute value of the accepted value to get the relative error: 0.29 ÷ 9.81 ≈ 0.0296.
  4. Multiply by 100. Multiply by 100 to express it as a percentage: the percent error is 2.96%. Keep the sign from step 1 if you want the signed percent error, −2.96%.
  5. Round sensibly. Report about as many significant figures as the values you measured had. The calculator rounds to the fewest in your two values (0.0506% for 3.14 against 3.14159) unless you choose another setting.

Frequently asked questions

How do you calculate percent error?
Subtract the accepted value from your measured value, take the absolute value, divide by the accepted value, and multiply by 100: |measured − accepted| ÷ |accepted| × 100. For 9.52 measured against 9.81 accepted that is |−0.29| ÷ 9.81 × 100 = 2.96%. The calculator writes this working out under the result.
What is the difference between percent error and percent difference?
Percent error compares a measurement with an accepted value and divides by that accepted value, because it is the reference. Percent difference compares two values that have equal standing, such as two trials or two instruments, and divides by their mean: |a − b| ÷ ((a + b) ÷ 2) × 100. For 9.52 and 9.81 the percent error is 2.96% and the percent difference is 3.00%. Use percent error when one of the values is known to be right, and percent difference when neither is. Percent difference is not defined when the two values add to 0, because their mean is 0. For X% of a number or a percentage change, use the percentage calculator.
Can percent error be negative?
The usual definition takes the absolute value, so percent error is never negative: 2.96% for 9.52 against 9.81. The signed version keeps the direction, −2.96% here, where a minus sign means the measurement is below the accepted value. This calculator divides by the size of the accepted value, so a plus sign always means above it, even for negative values: 2.0% for −4.9 against −5.0, because −4.9 is above −5.0.
What if the accepted value is 0?
Percent error is undefined, because it divides by the accepted value and nothing can be divided by 0. Entering a measured 0.4 against an accepted 0 gives the message “Undefined (accepted value is 0)” and still shows the absolute error, 0.4. With an accepted value of 0, report the absolute error with its unit or compare it with the size of the quantity you were measuring.
Can percent error be more than 100%?
Yes. Percent error passes 100% whenever the miss is bigger than the accepted value itself: a measured 25.0 against an accepted 10.0 is 150%. A measurement of 0 gives exactly 100%. A large percent error usually points to a mistake in the method, a wrong unit or a slipped decimal point.
What is absolute error and what is relative error?
Absolute error is the size of the miss in the quantity’s own unit: |measured − accepted|, 0.29 here. Relative error divides it by the accepted value, so it has no unit: 0.0296 here. Percent error is relative error written as a percentage: 0.0296 × 100 = 2.96%.
What is an acceptable percent error?
It depends on the instrument and the experiment. In typical teaching labs a percent error under about 5% is often considered good and under 10% acceptable, but a precision measurement may need a fraction of a percent. Compare your percent error with the uncertainty of your instruments rather than with a fixed cut-off.
How do I find the percent error of several measurements?
Use the several measurements mode with one accepted value. Each measurement gets its own error and percent error, and the mean absolute percent error averages their sizes: for 9.78, 9.81, 9.85, 9.74, 9.52 against 9.81 it is 0.877%. The percent error of the mean measurement is different: it averages the measurements first, so errors above and below cancel. For 9.6, 10.4, 9.8, 10.2 against 10 the mean absolute percent error is 3% but the percent error of the mean is 0%. The standard deviation of the measurements describes their spread, which the percent error does not.
How many significant figures should percent error have?
A common rule of thumb is as many as the least precise value in the calculation. For 3.14 against 3.14159 that is 3, giving 0.0506%, and 48 against 50 gives 4% because the trailing zero of 50 is not counted as significant, while 48 against 50. (with a point) gives 4.0%. Automatic rounding follows that rule; you can choose 2 to 6 figures or full precision, ≈ 2.95616717635% for 9.52 against 9.81. Absolute and signed errors are shown as the exact difference of what you typed.
Why does this calculator not show floating-point noise?
Most programs add and subtract decimals as binary floating-point numbers, where 9.81 − 9.52 comes out as 0.29000000000000092. This calculator reads every number as an exact decimal and does the subtraction, the multiplication and the division with whole-number arithmetic, so the difference is exactly 0.29 and only the final display is rounded.
Does it matter which value I call the accepted one?
Yes, because the accepted value is the divisor. With 9.52 measured against 9.81 accepted the percent error is 2.96%; swap the roles, so that 9.81 is measured against 9.52 accepted, and it is 3.05%. The absolute error, 0.29, is the same both ways. Percent difference, 3.00%, does not depend on the order because it divides by the mean, which is why it is the measure for two values with no reference.
Is percent error the same as percent change?
The arithmetic is the same: both divide a difference by a reference value. Percent change takes the starting value as the reference and reports a rise or fall; percent error takes the accepted value as the reference and reports how far a measurement is from it. Going from 9.81 to 9.52 is a change of −2.96%, the signed percent error above, and the percent error drops the sign: 2.96%. Use the percent change calculator for a before-and-after comparison.
Why is the same miss a bigger percent error on a small accepted value?
Percent error divides the miss by the accepted value, so the same absolute error weighs more against a small value. A miss of 1 unit is 50.0% against an accepted 2.00 but only 0.100% against an accepted 1000; the table above lists the steps in between. When a quantity is small, quote the absolute error as well, so the reader can judge the miss in the quantity’s own unit.
Should I round before calculating percent error?
No: round only the final answer. The relative error for 9.52 against 9.81 is 0.0296; rounded to 2 decimal places first it would be 0.03, and multiplying by 100 would give 3% instead of 2.96%. Keep every digit through the subtraction, the division and the multiplication, then round the percentage to the number of significant figures your measurements justify. The calculator follows that order.
How do I calculate percent error in a spreadsheet?
With the measured value in A2 and the accepted value in B2, the formula =ABS(A2-B2)/ABS(B2)*100 gives the percent error, and =(A2-B2)/ABS(B2)*100 gives the signed version. Format the cell as a number, not as a percentage, because the formula already multiplies by 100. For 9.52 and 9.81 both formulas agree with the calculator, 2.96% and −2.96%, apart from display rounding: a spreadsheet works in binary floating point and shows more digits than the measurements justify, so round the result yourself.
Is anything I type sent anywhere?
No. The numbers are used only in your browser to work out the result and are not sent anywhere.

How it works

Every number you type is read as an exact decimal (its digits and the position of the point), never as a binary float, using the same parser as the scientific notation converter, and combined with the exact BigInt arithmetic of the significant figures calculator. With m the measured value and a the accepted value: error = m − a; absolute error = |m − a|; relative error = |m − a| ÷ |a|; percent error = relative error × 100; signed percent error = (m − a) ÷ |a| × 100. The signed version divides by |a| rather than a, so a plus sign always means the measurement is above the accepted value, whatever the sign of a; for a positive accepted value this is the textbook formula. Percent difference = |m − a| ÷ ((m + a) ÷ 2) × 100 divides by the mean of the two values instead, and is shown for comparison because the two measures are often confused; it is undefined when m + a = 0. When the accepted value is 0 the percent error and relative error are undefined, and the calculator says so instead of printing infinity. In the several measurements mode, the mean absolute percent error is (Σ|mᵢ − a| ÷ n) ÷ |a| × 100, which equals the mean of the individual absolute percent errors because they share one accepted value; the mean signed percent error (bias) is the same with signed errors; and the percent error of the mean measurement is |mean − a| ÷ |a| × 100. Display: quotients are rounded half up to the chosen number of significant figures, by default the fewest among the non-zero values you typed (capped at 15); the zeros are kept (2.0% says two figures), and the choice of full precision shows up to 12 significant digits with ≈ marking a cut. Absolute and signed errors are exact differences and are never rounded. Separators in the list: spaces, new lines, semicolons and a comma followed by a space; a comma inside a number is a thousands separator (1,234) or a decimal comma (12,5). Up to 1,000 measurements are used and the table lists the first 100. Everything runs in your browser and nothing you type is sent anywhere.

Sources

More: Percent, fractions and arithmeticAll math & numbersAll toolsHome & DIY calculators