How this significant figures calculator works
Type a number exactly as it is written, trailing zeros included, and it tells you how many figures are significant, why, and what it looks like rounded. The input is a text box rather than a number box on purpose: 1.5 and 1.500 are the same value but not the same measurement, and a number field would quietly throw away the difference that this whole page is about.
The second mode does arithmetic, and it exists because addition and multiplication follow different rules. Getting those two the wrong way round is the most common error in the whole topic, and it is easy to make because both rules are about precision and neither is about the answer being wrong.
The rules
Zeros between non-zero digits always count
Leading zeros never count
Trailing zeros count only if there is a decimal point
Add or subtract: keep the fewest DECIMAL PLACES
Multiply or divide: keep the fewest SIGNIFICANT FIGURES
Exact numbers have unlimited significant figures
The first four rules are about reading a number; the last three are about what to do with it. The one that surprises people is the fourth, because it means 0.004500 has four significant figures (the two leading zeros place the decimal point, the trailing zeros were measured) while 4500 has somewhere between two and four and nothing in plain decimal notation can tell you which.
Worked example
0.004500 has 4 significant figures. The leading zeros do not count, because moving the decimal point is all they do; the 4 and the 5 obviously count; and the two trailing zeros count because there is a decimal point, which means somebody deliberately wrote them down to claim the measurement went that far. Written unambiguously, that is 4.500 × 10-3.
Now the arithmetic trap. Add 12.11 and 0.3 and your calculator says 12.41. But 0.3 is written to one decimal place, so the answer keeps one: 12.4. Multiply the same two numbers and the rule changes: 12.11 has 4 significant figures and 0.3 has 1, so 3.633 becomes 4, a single figure. Same two numbers, two operations, and the answers do not even have the same shape.
And the rounding tie. Round 1.005 to 3 significant figures. Almost every calculator that goes through floating point answers 1.00, because 1.005 is not exactly 1.005 in binary. The honest answer is 1.01, and this page gets there by working on the digits rather than on a float, which is the same approach our rounding calculator uses.
Why the rule changes between adding and multiplying
This is worth understanding once rather than memorising twice, because the reason makes both rules obvious. When you add, the uncertainties line up by place value. If one number is only known to the nearest tenth, then nobody knows anything about the hundredths of the total, however precisely the other number was measured. Precision in addition is about where the digits sit.
When you multiply, the uncertainties combine as fractions of the whole. A number good to one part in a thousand multiplied by one good to one part in ten gives a product good to about one part in ten. Where the decimal point sits is irrelevant; what matters is how many figures each number carries. Precision in multiplication is about how many digits, not which.
One consequence catches people out: an addition can change the number of significant figures dramatically. Subtract 10.11 from 10.12 and both inputs have four significant figures while the answer, 0.01, has one. That is not a mistake in the rule, it is the rule reporting something real. Subtracting two nearly equal measured numbers destroys precision, which is why numerical analysts have a name for it (catastrophic cancellation) and why a well-designed experiment avoids measuring a small difference between two large quantities.
The ambiguity nobody can rule their way out of
Write 4500 and you have written something genuinely undecidable. It might be a measurement good to four figures, or a rough count good to two with the zeros holding the decimal point in place. No convention fixes this, and any calculator that confidently reports a single answer is picking one and not telling you.
There are three honest ways out, and they are worth having. Scientific notation settles it completely: 4.5 × 103 is two figures, 4.500 × 103 is four, and neither can be misread. A trailing decimal point, written as 4500., declares that all four count, though it is easy to miss on a page and some styles discourage it. Best of all, state the uncertainty: 4500 ± 50 tells the reader exactly what you know, and it is what a lab report should carry anyway. Significant figures are a shorthand for uncertainty, and where the shorthand fails, the long form always works.
Two rules that get skipped, and one that is not what you were taught
Exact numbers have unlimited significant figures. Counted objects, defined conversions and the integers inside a formula are not measurements and carry no uncertainty, so they never limit an answer. Divide a measured 10.0 g between exactly 3 people and the answer keeps three figures, 3.33 g, not one. There are exactly 2.54 centimetres in an inch and exactly 1,000 metres in a kilometre, both by definition, so neither ever costs you precision. Missing this is how a careful measurement gets thrown away by its own arithmetic.
Round only at the end. Carry extra digits through every intermediate step and round once, when you report. Rounding at each stage lets small errors accumulate into a wrong final digit, which is exactly what the rules are meant to prevent.
And the one that is probably not what you were taught: when a digit is exactly 5 with nothing after it, science rounds to the nearest even number rather than always rounding up. So 2.5 becomes 2 and 3.5 becomes 4. It looks arbitrary until you add up a long column: always rounding ties upward biases the total upward, while sending half of them down cancels out. Most school teaching and most spreadsheets use round-half-up, which is why the two disagree on exactly these cases and only these. The selector above lets you use either, and the result tells you when your particular number was one where the choice actually mattered.