Significant Figures and Rounding: How Precise Should a Conversion Be?
A converter will give you 3.6576 metres from a tape measure reading. Most of those digits are fiction. How to decide what precision an answer deserves.
Introduction
Measure a room with a tape measure, call it 12 feet, and ask any converter for the metric equivalent. It will tell you 3.6576 metres. Four decimal places, implying you know the width of that room to a tenth of a millimetre, from a reading you took by eye against a tape marked in eighths of an inch.
The number is not wrong. The precision is a fiction. Conversion multiplies your measurement by a constant, and multiplication cannot create information that was never in the measurement to begin with. This guide covers how much of a converted answer is real, the rules for deciding where to cut it off, and the rounding behaviour that quietly changes results in spreadsheets and code.
The Governing Idea: Precision Cannot Be Created
Every measurement carries an implied uncertainty. "12 feet" from a tape measure means something closer to "12 feet, give or take half an inch". Converting it does not sharpen it. The uncertainty travels through the arithmetic and comes out the other side, proportionally the same size as it went in.
So the honest version of that conversion is about 3.66 metres, or even 3.7. The extra digits are noise dressed up as knowledge, and reporting them tells a reader your measurement was more careful than it was.
Counting Significant Figures
Significant figures are the digits in a number that carry real information. The rules are short:
- Non-zero digits always count. 47.3 has three.
- Zeros between non-zero digits count. 4007 has four.
- Leading zeros never count. 0.0045 has two - they only mark the decimal position.
- Trailing zeros after a decimal point count. 0.00450 has three, and that final zero is a deliberate claim about precision.
- Trailing zeros in a whole number are ambiguous. 1200 could be two, three or four.
That last case is the reason scientific notation exists. Writing 1.2 × 10³ means two significant figures and 1.200 × 10³ means four, with no ambiguity left for the reader to guess at.
The Two Arithmetic Rules
Different operations propagate precision differently, which catches people who learn one rule and apply it everywhere.
For multiplication and division - which covers nearly every unit conversion - the result carries the fewest significant figures of the inputs. A board measured as 4.5 inches has two significant figures. Multiply by 25.4 and the calculator says 114.3 mm, but the honest answer is 110 mm.
For addition and subtraction, the result carries the fewest decimal places, not the fewest significant figures. 12.11 plus 0.3 gives 12.41 on a calculator, but the second value is only known to one decimal place, so the answer is 12.4.
Exact Conversion Factors Do Not Count
This is the part that trips up people who have learned the rules properly, because it looks like an exception and is really a consequence.
Some conversion factors are definitions, not measurements. One inch is defined as exactly 25.4 millimetres. One pound is exactly 0.45359237 kilograms. These carry infinite significant figures, because there is no uncertainty in them at all - they are agreements, not observations.
So an exact factor never limits the precision of your answer. Only your measurement does. Converting 165 pounds gives 74.842741 kilograms on the calculator, and the question of how to round it depends entirely on how the 165 was obtained. From a bathroom scale reading to the nearest pound, 74.8 kg is the honest answer. Our guides to inches and millimeters and kilograms and pounds cover both of those defined constants in detail.
Counts behave the same way. If you have exactly 12 items, that 12 is not a measurement and does not limit anything.
Never Round Halfway Through
Rounding intermediate results is one of the most reliable ways to produce a wrong answer while following every other rule correctly.
Convert 100 kilometres to feet in one pass and you get 328,084 feet. Do it in two steps, rounding to 62 miles along the way, and you get 327,360 feet. The two answers differ by 724 feet - more than an eighth of a mile, produced entirely by tidying up a number mid-calculation.
The rule is simple: carry full precision through every step and round once, at the very end, when you report the result.
Two Ways to Round a Half
Everyone learns to round 0.5 upward. Most software does not.
Round half up is the schoolroom rule: 2.5 becomes 3, 3.5 becomes 4. It is simple, and it has a bias - across many values it pushes totals slightly upward, because halves always travel in the same direction.
Round half to even, often called banker's rounding, sends a half to whichever neighbour is even: 2.5 becomes 2, 3.5 becomes 4. Half the time it rounds down and half the time up, so the bias cancels across a large set of numbers. This is the default in the IEEE 754 floating point standard and in many statistical and financial tools.
Neither is more correct. What matters is knowing which one your tool uses, because a spreadsheet and a programming language can disagree on the same column of data, and the gap only shows up once totals stop matching.
Why Computers Produce Strange Decimals
There is a further wrinkle underneath all of this. Computers store decimals in binary, and most decimal fractions have no exact binary representation - in the same way that a third has no exact decimal one.
The consequences are visible in any browser console. Adding 0.1 and 0.2 gives 0.30000000000000004. Converting 12 feet to metres by multiplying by 0.3048 gives 3.6576000000000004. Those trailing digits are not a bug in the conversion; they are the gap between binary storage and decimal notation.
It also means rounding functions can surprise you. Asking many languages to round 1.005 to two decimal places returns 1.00 rather than 1.01, because the stored value is fractionally below 1.005. Rounding to a sensible number of significant figures makes the whole issue disappear, since it discards exactly the digits where the noise lives.
When Precision Genuinely Matters
The right number of digits is a judgement about consequences, not a universal rule.
Where it matters
- Medicine. Doses calculated per kilogram of body weight need real precision, and every rounding step compounds into the final quantity.
- Engineering and machining. Tolerances are specified precisely because parts must fit, and treating an inch as 25 mm rather than 25.4 introduces a 1.6 percent error that accumulates across an assembly.
- Aviation and marine fuel. Load calculations mix weight and volume units, and errors here have famously ended in emergency landings.
- Finance. Fractions of a currency unit, repeated across thousands of transactions, become real money.
Where it does not
- Cooking. An oven that cycles ten degrees either side of its setpoint does not care about 356 versus 350.
- Weather and travel. Nobody needs their forecast to a hundredth of a degree.
- Casual estimates. Rounding for a mental sanity check is exactly what mental arithmetic is for.
A Practical Rule of Thumb
If you do not want to count significant figures every time, this approximation covers most everyday work:
- Match the precision of your original measurement, and no more.
- Carry full precision through intermediate steps.
- Round once, at the end.
- If the answer will feed into another calculation, hand over the unrounded value.
- When in doubt, one more digit than you need is safer than one fewer - but four more is just noise.
Common Mistakes Worth Avoiding
Copying every digit the calculator offers
A converter shows what the arithmetic produced, not what your measurement justifies. The trimming is your job.
Letting an exact factor limit the answer
25.4 is a definition with infinite significant figures. It never constrains your result.
Applying the multiplication rule to addition
Multiplication counts significant figures; addition counts decimal places. They frequently give different answers.
Rounding for display, then reusing the displayed value
A rounded number is for a human reader. Feeding it back into a calculation restarts the compounding problem.
Conclusion
A conversion is only as good as the measurement behind it. Count the significant figures in what you actually measured, treat defined constants as exact, carry full precision through every intermediate step, and round once when you report the result. Do that and your answers stop claiming a confidence your tape measure never had.
Our millimeters to inches converter and pounds to kilograms converter work at full precision so you can round at the point that suits your purpose, and the scientific calculator handles the intermediate arithmetic. For the metric side of the same question, our guide to metric prefixes explains why the ladder itself never introduces rounding error.