The short answer: Rounding replaces a number with the nearest value at a chosen place. Find the rounding digit, look at the digit immediately to its right, and leave the digit alone if that digit is 0–4 or increase it by one if it is 5–9. Half-up sends ties away from zero; significant figures round to a count of meaningful digits rather than a fixed decimal place.

Key takeaways

  • Look one digit to the right. Only the next digit decides; the rest are ignored.
  • Half-up is not the only rule. Half-even sends ties to the nearest even digit to reduce bias.
  • Significant figures count meaning, not decimals. 0.004560 has four significant figures.
  • Round once, at the end. Rounding intermediate values can change the final answer.
  • Zeros can be significant or not. Scientific notation removes the ambiguity.

What rounding does and what it costs

Rounding replaces a number with a nearby, simpler value at a chosen place. It is an approximation, so it discards information: the rounded value differs from the original by at most half a unit in the last place kept. Rounding to the nearest whole number gives an error of at most 0.5, rounding to the nearest ten gives at most 5, and rounding to the nearest tenth gives at most 0.05. That bound is why rounding is safe for presentation but risky in the middle of a calculation.

The purpose is communication and estimation, not changing the underlying quantity. A measurement of 3,847.265 metres is still that measurement after it is reported as 3,847 metres; the report simply claims less precision. Rounding does not make a rough measurement precise.

Place value and the rounding digit

Place value names the position being kept: thousands, hundreds, tens, ones, tenths, hundredths, thousandths. To round, identify the digit in the target place, then look at the single digit immediately to its right. If that digit is 0, 1, 2, 3, or 4, keep the target digit unchanged. If it is 5, 6, 7, 8, or 9, increase the target digit by one and replace everything to its right with zeros or drop it if it is after the decimal point.

For 3,847.265, the digit in the tens place is 4 and the digit to its right is 7, so rounding to the nearest ten gives 3,850. The digit in the hundredths place is 6 and the digit to its right is 5, so rounding to the nearest hundredth gives 3,847.27. The digit in the tenths place is 2 and the digit to its right is 6, so rounding to the nearest tenth gives 3,847.3.

Half-up, half-even, and negative numbers

Half-up is the rule most people learn: a tie rounds away from zero, so 2.5 becomes 3 and −2.5 becomes −3. Half-even, also called banker’s rounding, sends a tie to the nearest even digit: 2.5 becomes 2, 3.5 becomes 4, and −2.5 becomes −2. Neither rule is universally correct. Half-up is common in school arithmetic and consumer pricing; half-even reduces cumulative bias in long sums and is common in statistics and accounting.

Direction matters for negatives because “up” is ambiguous. Rounding −2.5 to −3 moves away from zero, while rounding it to −2 moves toward zero. Floor rounds toward negative infinity, ceiling rounds toward positive infinity, and truncation drops the fractional part toward zero.

Worked example at every place

Take 3,847.265 and round it to each common place. Nearest thousand: the hundreds digit is 8, so 4,000. Nearest hundred: the tens digit is 4, so 3,800. Nearest ten: the ones digit is 7, so 3,850. Nearest whole: the tenths digit is 2, so 3,847. Nearest tenth: the hundredths digit is 6, so 3,847.3. Nearest hundredth: the thousandths digit is 5, so 3,847.27 under half-up.

Carries need care. Rounding 9.97 to the nearest tenth looks at the hundredths digit 7, increases the tenths digit 9, and carries: 9.97 becomes 10.0, not 9.10. Rounding 99.5 to the nearest whole gives 100. The trailing zero in 10.0 is significant because it shows the precision of the rounding; dropping it to 10 changes the implied place.

Significant figures

Significant figures count the digits that carry meaning, starting with the first non-zero digit. All non-zero digits are significant. Zeros between non-zero digits are significant. Leading zeros are never significant. Trailing zeros after a decimal point are significant. Trailing zeros in a whole number without a decimal point are ambiguous: 1200 could have two, three, or four significant figures. Scientific notation resolves it: 1.2 × 10³ has two, 1.20 × 10³ has three, and 1.200 × 10³ has four.

So 0.004560 has four significant figures: the leading zeros only locate the decimal point, while 4, 5, 6, and the trailing 0 are meaningful. 100.0 has four. 0.003 has one. Exact counts, such as 12 eggs or 3 sides of a triangle, have unlimited significant figures and should not limit a calculation.

Rounding in calculations

For multiplication and division, the result carries the smallest number of significant figures among the inputs. 12.34 × 5.6 = 69.104, which rounds to 69 because 5.6 has two significant figures. For addition and subtraction, the result carries the smallest number of decimal places. 10.234 + 5.2 + 100.3234 = 115.7574, which rounds to 115.8 because 5.2 is known only to one decimal place. These are different rules because multiplication scales relative error, while addition aligns absolute positions.

Round once, at the end. Double rounding is a real error: 2.549 rounded to two decimals is 2.55, and then to one decimal is 2.6, but rounding directly to one decimal gives 2.5 because the hundredths digit is 4. Keep at least one guard digit through intermediate steps, then apply the rule to the final result.

Money, large numbers, and estimation

Money is usually rounded to the nearest cent, but the arithmetic should be done in exact decimal or integer cents, not binary floating point. The classic example is 2.675: in binary floating-point it is stored slightly below 2.675, so a naive round to two decimals can give 2.67 instead of 2.68. Some currencies round cash totals to the nearest 5 cents, which is a separate rule from rounding each line item. Apply the rule at the total, not to every price, or the sum will drift.

Large numbers carry the same ambiguity. A population of 1,200,000 is rarely known to the last person, so report it as 1.2 million or 1.2 × 10⁶ if two significant figures are intended. Front-end estimation uses the leading digits: 398 × 21 is about 400 × 20 = 8,000, while the exact product is 8,358, an error of about 4.5%.

Common mistakes and caveats

  • Looking at every digit to the right instead of only the immediate neighbour.
  • Rounding 5 down by reflex, or forgetting that a half-up tie goes away from zero for negatives.
  • Dropping trailing zeros that are significant, such as turning 10.0 into 10.
  • Counting leading zeros as significant, so 0.004 is read as three figures instead of one.
  • Confusing decimal places with significant figures when applying the multiplication rule.
  • Double rounding, or rounding each intermediate step in a long calculation.
  • Trusting binary floating point for money instead of decimal arithmetic.

Context decides the convention. A physics lab may keep three significant figures and report an uncertainty; a bank may be required to use a specific rounding rule for interest; a tax authority may specify rounding at each line.

When not to round

Do not round in the middle of a calculation. Keep full precision until the final step, then round once. Do not round exact counts: 1,247 people is not 1,200 people, and 3 sides is not an approximation. Do not round identifiers, account numbers, or codes, where every digit is meaningful. Do not round safety-critical values in a way that hides a threshold, and do not round a number before comparing it with a limit when the comparison is the whole point.

Rounding is also not the same as estimating. An estimate chooses a convenient value to simplify the work; rounding converts a known value to a stated precision. The two often travel together, as in 398 × 21 ≈ 400 × 20, but the estimate has no precision claim. Report the rounded result with its place or its significant figures, and the reader can tell how much of the original number survived.

Frequently asked questions

How do you round 3,847.265 to the nearest hundred?

Look at the tens digit, 4. Since 4 is less than 5, keep the hundreds digit 8 and zero the rest: 3,800.

What is the difference between half-up and half-even?

Half-up sends a tie away from zero: 2.5 → 3 and 3.5 → 4. Half-even sends a tie to the nearest even digit: 2.5 → 2 and 3.5 → 4.

Does 1200 have two or four significant figures?

It is ambiguous without context. Write 1.2 × 10³ for two figures or 1.200 × 10³ for four.

Why did 2.675 round to 2.67?

Binary floating point stores it slightly below 2.675. Use decimal arithmetic or integer cents for money to avoid that representation error.

Should I round before adding or after?

After. Add at full precision, then round the total to the least number of decimal places in the inputs.

Is rounding the same as truncating?

No. Truncation drops digits without adjusting the last kept digit, so 2.9 truncates to 2 but rounds to 3. Truncation always moves toward zero for positive numbers.

Further reading

Keep exploring

Read next Logarithm: Base, Product Rule, and Scale A logarithm is an exponent in disguise. Learn the definition, the product, quotient and power rules, change of… Read next Order of Operations: PEMDAS and Ambiguous Expressions PEMDAS is a shared convention, not a law of mathematics. Learn the real precedence rules, the exponent and… Read next Significant Figures: Counting, Rounding, and Precision Significant figures record how precisely a value was measured. Learn the counting rules, the addition and multiplication rules,…