Guide · Published September 15, 2026 · By the site maintainer

How to round unit conversions without losing accuracy

Use the full conversion factor during the calculation, keep extra digits in intermediate steps, and round the final result to match the precision of the original measurement. A calculator may display ten decimal places, but those digits do not become meaningful just because they appear on screen.

Rounding too early can move the answer. Rounding too late can make a rough measurement look exact. The right result preserves what the input actually tells you and gives enough digits for the decision at hand.

The five-step method

1. Identify the input. Decide whether it is an exact count, a defined value, or a measured value.
2. Use an authoritative conversion factor. Keep all stated digits, especially when the factor is exact.
3. Calculate without trimming intermediate results. Retain guard digits through multiplication, division, powers, and offsets.
4. Choose precision from the input and purpose. Consider significant figures, measurement resolution, uncertainty, and any real threshold.
5. Round once at the end. Show the unit and enough digits to make the intended precision clear.

NIST gives this same core advice in its Guide to the SI conversion-factor appendix: converted values should be rounded consistently with the possible rounding error in the unconverted value. The goal is to avoid both discarding information and claiming information the original value never contained.

Exact numbers and measured numbers behave differently

Some conversion factors are exact definitions. One inch equals exactly 2.54 centimeters. One international pound equals exactly 0.45359237 kilogram. These factors do not limit the significant figures of the answer. The input measurement usually does.

Counts can also be exact. If a package contains exactly 12 identical one-foot tiles, the count 12 does not mean “approximately twelve.” By contrast, a board described as 12 feet long may have been measured only to the nearest foot, inch, or eighth of an inch. Those inputs deserve different result precision even though both display the number 12.

Ask what the final digit means. A length written as 5.2 m normally suggests resolution to the nearest tenth of a meter. A laboratory result written as 5.200 m communicates finer resolution. Adding zeros to the converted answer cannot improve the original measurement.

Significant figures are not decimal places

Significant figures count meaningful digits from the first nonzero digit. Decimal places count positions to the right of the decimal point. The two ideas coincide only in some numbers.

If the written input is ambiguous, do not invent certainty. A result based on “1,200 ft” should be labeled with a sensible approximation unless the source explains whether the last two zeros were measured. Writing the input as 1.2 × 10³ ft or 1.200 × 10³ ft removes the ambiguity.

Use the full factor and keep guard digits

Suppose a distance is 17 miles. The exact international-mile factor is 1.609344 kilometers per mile:

17 mi × 1.609344 km/mi = 27.358848 km

If 17 mi is a two-significant-figure measurement, report about 27 km. Do not replace the factor with 1.6 before multiplying. That shortcut produces 27.2 km, and another operation may amplify the difference. The full factor costs nothing in a calculator or spreadsheet.

Guard digits are the unreported digits you keep while working. If the 17-mile distance feeds into a speed, area, cost, or fuel calculation, keep 27.358848 internally and round only the final requested quantity. A displayed result can be short while the stored value remains precise.

Why rounding once matters

Consider a 5.25 mile route converted to kilometers and then used with a 42-minute time. The raw distance is 8.449056 km. If you first round it to 8 km, the calculated pace changes noticeably. Keep 8.449056 for the pace calculation, then round the final pace to a useful value.

Repeated conversions should also use the underlying value rather than the displayed text. If software converts miles to kilometers, shows a rounded number, and then converts that rounded display back to miles, the value can drift with every cycle. The display is for reading; it should not become the next calculation's source.

How the final digits should look

The 36-foot example deserves attention. NIST uses it to show the tradeoff between too few and too many digits. Reporting 11 m can lose more information than the original value implied, while 10.9728 m presents every calculator digit as meaningful. A result of 11.0 m can be the better representation. End use still matters.

What should happen when the next digit is exactly 5?

The familiar classroom rule says to increase the retained digit whenever the first discarded digit is 5 or greater. Technical work often uses round-half-to-even when the discarded part is exactly a 5 followed only by zeros. Under that rule, the retained digit stays unchanged if it is already even and increases if it is odd. Thus 3.45 rounded to one decimal becomes 3.4, while 3.55 becomes 3.6.

Half-to-even reduces systematic upward bias when a large data set contains many exact ties. NIST's official appendix specifies this method for its examples. Software does not all behave the same way, especially with binary floating-point numbers that only approximate many decimal fractions. For an audited calculation, state the rounding mode instead of assuming everyone used the same one.

Temperature conversions need the full formula

Temperature scales include an offset, so rounding the factor or applying the offset in the wrong order causes trouble. Convert Fahrenheit to Celsius with:

°C = (°F − 32) × 5/9

For 70.0°F, subtract first, multiply by the exact ratio, and keep the repeating decimal until the end: 21.111...°C becomes 21.1°C. Use the Celsius and Fahrenheit converter for the calculation. A temperature difference follows a different rule: a change of 10°C equals a change of 18°F, with no 32-degree offset.

Area and volume can magnify an early shortcut

Squared and cubed units require squared and cubed conversion factors. One foot equals 0.3048 m exactly, so one square foot equals 0.09290304 m². If you round 0.3048 to 0.3 before squaring, you get 0.09 m². That may look close for one square foot, but the shortfall grows across a large floor.

Convert the area directly with the exact area factor or square the unrounded length factor. Use the square meters and square feet converter for common floor and property measurements. Apply the same principle to cubic units: cube the factor before rounding, not a shortened display.

Measurement uncertainty is more informative than a digit count

Significant figures are a useful shorthand, but a stated uncertainty is clearer. If a rod measures 10.0 ± 0.1 inches, multiplying both the value and its uncertainty by the exact 2.54 cm/in factor gives 25.4 ± 0.254 cm. The uncertainty can be reported sensibly as about 0.3 cm, with the measured value rounded to the same decimal place: 25.4 ± 0.3 cm.

NIST Technical Note 1297 provides a full framework for evaluating and reporting uncertainty. Its measurement-uncertainty guidance is intended for technical results that need more than a simple significant-figures rule. Do not treat this short example as a complete uncertainty analysis when several measurements, correlations, calibration errors, or statistical effects are involved.

Match rounding to the decision

• Mental estimates: One or two significant figures may be enough to picture a distance or temperature.
• Recipes: Match the kitchen scale or measuring tool. Do not report thousandths of a gram from a scale that reads whole grams.
• Building and fabrication: Follow the drawing, tolerance, and instrument resolution. A nominal board size is not a measured finished size.
• Shipping and baggage: Keep enough precision to compare the actual measurement with a carrier or airline threshold, and follow its stated rounding rules.
• Science and calibration: Report uncertainty and the method, not only a chosen number of decimal places.
• Medication and safety calculations: Use the approved protocol, verified inputs, and qualified guidance rather than a generic rounding shortcut.

A spreadsheet checklist

1. Store the entered value in one cell and the full conversion factor in another or in a documented formula.
2. Calculate with the raw cells. Do not copy a formatted display value into the next formula.
3. Keep units in headings and document whether the factor is exact.
4. Use cell formatting for display when possible, leaving the stored number unrounded.
5. If a rounded value must feed another system, define and document the rounding rule at that boundary.
6. Test values just below, at, and just above any threshold.

For quick estimates, the mental conversion guide labels the error behind each shortcut. For exact factors, use the metric conversion chart and the consolidated converter that matches your units. Keep the math precise internally, then publish only the digits the input and purpose can support.