Relative Error Calculator

Relative error measures a measurement against the truth: how far off you were, scaled by the size of the true value. This calculator reports the signed and absolute error, the relative and percent error, and the accuracy, with a four-step working, an interpretation of the result, an optional tolerance PASS/FAIL verdict, and a symmetric percent-difference mode for comparisons with no true value.

Quick Answer

How to use the relative error calculator

Choose whether you are comparing a measurement to a true value or tracking change between two readings, review the pre-filled example, then enter your own numbers. Results stay hidden until you click Calculate.

1. Percent error mode: enter the measured value (what you observed) and the true value (the accepted or theoretical answer). You get the signed error, absolute error, relative error, percent error and accuracy, with a four-step working and an interpretation of the result.
2. Set a tolerance in the optional field and the result carries a PASS or FAIL verdict against your specification, exactly what quality checks and lab reports need. Leave it blank to skip the verdict.
3. Percent-difference mode: enter two values when neither is the true one. The tool compares them symmetrically as |A − B| ÷ average × 100, with its own three-step working.
4. Copy or share from the results box: the copied text carries the full working, so a shared error figure never arrives without its method.

What relative error measures

Relative error is the size of your measurement error compared to the true value: |measured − true| ÷ |true|. Percent error is the same thing times 100. Because it is relative, a 1-unit error on a true value of 10 (10%) looks very different from a 1-unit error on a true value of 1,000 (0.1%), which is exactly the context absolute error alone cannot give you.

The true value goes on the bottom because error is measured relative to what the result should have been, not relative to your own reading. Using the measured value as the denominator answers a different question. The absolute value calculator explains the |x| notation used above, and the accepted values themselves come from reference data, such as the fundamental physical constants maintained by NIST.

Worked example: measuring gravity, checked against this exact tool

You measure gravity as 9.8 m/s²; the accepted value is 9.81. Every number below is the calculator's own output for the pre-filled inputs.

• Signed error = 9.8 − 9.81 = −0.01 (an underestimate)
• Absolute error = 0.01
• Relative error = 0.01 ÷ 9.81 = 0.001019
• Percent error = 0.1019%
• Accuracy = 100% − 0.1019% = 99.8981%
• Interpretation: Excellent: within precision-measurement territory.

Note the sign convention: percent error itself is always zero or positive because it uses absolute values, while the signed error keeps the direction separately. The classic aluminum-density example from LibreTexts chemistry works the same way: measuring 2.45 g/cm³ against an accepted 2.70 g/cm³ gives |2.45 − 2.70| ÷ 2.70 × 100 = 9.26%, the same arithmetic this tool runs. Density and yield belong together in lab work, which is why the percent yield calculator is the natural next stop after an error check.

The tolerance check: turning a number into a decision

Standard relative-error tools compute a percent and stop. They cannot tell you whether the error passes your specification. Enter a tolerance and the result carries an explicit verdict, which is what quality checks, calibration records and lab reports actually need.

The verdict is strict about inputs: a negative tolerance is refused outright, because a negative specification is meaningless. The same PASS/FAIL idea applies wherever a limit matters, from manufacturing tolerances to the bonus tolerance calculator on the finance side.

Percent-difference mode: when there is no true value

Sometimes neither number is the reference: two estimates, two methods, a before and after. Percent difference compares them symmetrically as |A − B| ÷ average × 100, so swapping the inputs changes nothing. Values 110 and 100 give an average of 105 and a difference of 10 ÷ 105 × 100 = 9.5238%, a number the plain error formula cannot produce honestly, because it would depend on which value you called true (9.0909% one way, 10% the other).

The tolerance field works in this mode too: at a 10% tolerance, 9.5238% reads "PASS: 9.5238% is within your 10% tolerance." The analytical-chemistry teaching at LibreTexts draws the same line between percent error (accuracy against a true value) and percent difference (precision-style comparison of two measurements).

Reading the interpretation: what counts as acceptable

What counts as a good percent error depends entirely on the field, so the tool interprets the number against rough bands rather than declaring it good or bad itself.

These bands line up with the usual field guides: school labs typically accept under 5 to 10%, general engineering under 2 to 5%, precision work under 1%, and analytical chemistry often under 0.5%. One honest edge case: accuracy is computed as 100% minus the percent error, so an error above 100% gives a negative accuracy. The tool shows it as-is rather than clamping it, because a measurement that missed by more than the whole true value is genuinely worse than zero accuracy describes.

Accuracy is not precision

Percent error measures accuracy: closeness to the true value. It does not measure precision, which is how tightly repeated measurements cluster with each other. A tight cluster of readings can be precise yet inaccurate, all landing far from the true value, or accurate on average but scattered. Use percent error for accuracy and statistics like standard deviation for precision; the relative standard deviation calculator covers the spread side of the story.

What competing relative-error calculators miss

The top results for this keyword each do one slice of the job well and stop there:

• 1dollars.org percent error calculator: signed, absolute, relative and percent error with steps and decimal-place control, plus an honest note that input order matters. No tolerance verdict, no percent-difference mode.
• gigacalculator.com: many worked examples from different starting information, including treehouse and scale scenarios. One-way tool only; nothing on whether the error passes a specification.
• Omni Calculator percent error: clean formula teaching with a speed-of-sound example, and useful disambiguation of standard error and margin of error. No tolerance check, no symmetric comparison mode.
• calculatorlib.com: the closest rival, with an interpretation guide and acceptability bands by field plus an accuracy-vs-precision section. Still no PASS/FAIL tolerance verdict and no percent-difference mode.
• calculatorspot.online: a quick-reference table of lab scenarios with precomputed errors. A lookup table, not a decision tool.
• truegeometry error analysis: goes further into multi-measurement statistics (mean absolute error, RMSE) but as a formula reference, not a guided tool.
• The gap this page fills: signed plus absolute plus relative plus percent error with a four-step working, an explicit PASS/FAIL tolerance verdict, a symmetric percent-difference mode with its own working, and an interpretation guide, on one page with one set of inputs.

Frequently asked questions

Add This Calculator to Your Website

References