Methodology

This page documents the method behind every number on TransformUnits: where each conversion factor comes from, which figures are exact by international agreement versus measured, and where this site’s own limits are. Every figure below can be checked against its cited source.

From SI base units to a conversion factor

The International System of Units (SI) defines seven base units — metre, kilogram, second, ampere, kelvin, mole and candela — from which every other unit is derived. A conversion factor between two units in the same category is the ratio between each unit’s size and the category’s base unit.

Worked example: how many metres are in one mile? A mile is defined as exactly 1,760 international yards. The international yard was fixed at exactly 0.9144 metres by the international yard and pound agreement, signed on 1 July 1959 by the standards bodies of Australia, Canada, New Zealand, South Africa, the United Kingdom and the United States. So 1 mile = 1,760 × 0.9144 m = 1,609.344 m — an exact value, because both steps are exact definitions, not measurements.

Every factor on this site is built the same way: trace the unit back to its category’s reference unit (metre for length and kilogram for mass — both SI base units; litre for volume — accepted for use with SI but not itself an SI base unit; and so on), then store the resulting ratio at full precision. Rounding, where it happens, is applied only when a result is displayed, never before storage (see “Rounding and precision” below).

Exact by definition vs. measured

Some conversion factors are exact by definition, fixed by international agreement with no uncertainty. Others are, or once were, empirical measurements with a margin of error.

Exact by definition: the international pound is exactly 0.45359237 kilograms and the international inch is exactly 25.4 millimetres, both fixed by the 1959 international yard and pound agreement. Since 20 May 2019, the kilogram has been defined by fixing the Planck constant at exactly 6.62607015 × 10^-34 J·s, replacing the physical platinum-iridium prototype that previously defined it.
Measured, not defined: before 2019, the kelvin was fixed by declaring the triple point of water to be exactly 273.16 K. After the 2019 redefinition, the kelvin is instead defined by fixing the Boltzmann constant at exactly 1.380649 × 10^-23 J/K, so the triple point of water is no longer an exact reference point — it is now a measured quantity with a small experimental uncertainty. What used to be the definition became the measurement, and vice versa.

A site that presents every number with equal confidence is misrepresenting its own data. Where a figure on this site is a snapshot of a real-world, changing quantity rather than a fixed definition — the electricity carbon-emissions factor is the clearest example — that is stated explicitly, with a source and a year, rather than presented as a constant.

Why temperature needs an offset, not just a ratio

Length, mass and volume conversions are pure ratios: multiply by one factor and the zero points already line up. Temperature scales do not share a zero point — 0°C is the freezing point of water, 0°F is a different point roughly 32 Fahrenheit degrees below that, and 0 K is absolute zero, a genuinely different physical state. A single multiplication cannot reconcile scales whose zero points do not match; the conversion needs an added or subtracted offset as well as a ratio (an affine transform, not a linear one). That is why temperature conversions on this site are computed through kelvin as an intermediate step, rather than a single stored ratio like every other category.

How a real thermometer is calibrated is defined by the International Temperature Scale of 1990 (ITS-90), adopted by the International Committee for Weights and Measures. ITS-90 specifies reproducible fixed points — such as the triple point of equilibrium hydrogen and the freezing point of silver — and standard interpolation instruments, such as platinum resistance thermometers, for the ranges between them, so thermometers calibrated in different countries agree with each other in practice.

Rounding and precision

Every stored conversion factor keeps full double-precision floating-point accuracy; nothing is pre-rounded in the database. What you see on screen is a display decision made when the page is generated:

Results are shown to up to 8 decimal places with trailing zeros removed, so a result reads “2.5”, not “2.50000000”.
Very large results (1 billion or more) or very small ones (under 0.000001) switch to scientific notation, since eight fixed decimal places would otherwise be unreadable or misleadingly precise at that scale.
A result of exactly zero always displays as “0”.

This is a display policy, not a claim about how many of those digits matter for your specific calculation. A result shown to eight decimal places is not more accurate than one shown to two if your starting value was only measured to two significant figures — judging significant figures for a specific calculation is left to the reader, since it depends on the precision of the original input, which this site cannot know.

Known limitations

Being specific about what this site does not do is part of the method, not separate from it:

Carbon emissions: the electricity-to-CO2e category uses a single global average emissions factor, cited with its source and year on the relevant pages. Real grid carbon intensity varies by country, generation mix and year, sometimes by a factor of ten or more between the cleanest and dirtiest national grids. Treat this category as an order-of-magnitude estimate, not a substitute for a country- or supplier-specific figure.
Cooking volumes: the volume category offers metric teaspoon (5 mL) and metric tablespoon (15 mL) — the measures used across most of the world — rather than a generic “teaspoon” or “cup”. A US customary teaspoon (≈4.93 mL) and a US legal cup (240 mL) are not the same as their metric namesakes, and conflating them silently would be worse than not offering them.
What this site deliberately does not offer: a heart-rate-variability conversion category (RMSSD ↔ SDNN) was removed rather than shipped with an invented factor. RMSSD and SDNN are different statistical measures of the same series of heartbeat intervals, and no fixed ratio relates one to the other; it depends on the shape of the underlying signal, which differs recording to recording. Publishing a numeric conversion between them would have stated something false. See the changelog for the removal.