Two gallons, three tons, two kilobytes, and a temperature you cannot double.
Most unit conversion is multiplication by a constant and nothing more. The trouble is concentrated in a small number of units where that assumption quietly fails — either because the unit means different things in different countries, or because the quantity is not the kind of thing that scales at all. These are the ones worth knowing.
Start with the one that is conceptually different rather than merely confusing.
Length, mass and volume sit on ratio scales. They have a true zero meaning "none of this", so ratios are meaningful: 10 metres really is twice 5 metres, and converting is a single multiplication.
Celsius and Fahrenheit are interval scales. Their zeros are arbitrary conventions, not an absence of temperature. Conversion therefore needs a multiplication and an offset:
F = C x 9/5 + 32
C = (F - 32) x 5/9Two consequences follow, and both catch people:
You cannot say 20°C is twice as warm as 10°C. The same two temperatures in Fahrenheit are 68°F and 50°F, a ratio of 1.36. A ratio that changes when you change units was never a real ratio. If you genuinely need proportional reasoning about temperature, use Kelvin, which has a true zero.
Temperature differences convert differently from temperatures. A rise of 10°C is a rise of 18°F, not 50°F — because the offset applies to a point on the scale, not to a gap between two points. "The temperature rose by 10 degrees" and "the temperature is 10 degrees" need different arithmetic, and conflating them is a classic bug.
The imperial and US customary systems share names for units of different sizes. This is not a rounding difference — it is a fifth.
| Unit | US customary | Imperial (UK) | Difference |
|---|---|---|---|
| Gallon | 3.785 L | 4.546 L | Imperial is ~20% larger |
| Quart | 0.946 L | 1.137 L | ~20% |
| Pint | 473 mL | 568 mL | ~20% |
| Fluid ounce | 29.57 mL | 28.41 mL | US is ~4% larger |
Note that the fluid ounce runs the other way from the larger units. That is not an error: a US pint is 16 US fluid ounces while an imperial pint is 20 imperial fluid ounces, so the smaller unit ends up slightly larger in the US system even though every unit above it is smaller. Recipes that cross the Atlantic go wrong for exactly this reason, and a 20% error in liquid quantities is the difference between a cake and a puddle.
A separate trap with the same word. An ounce is a unit of mass, about 28.35 grams. A fluid ounce is a unit of volume. They are not interchangeable, and there is no general conversion between them — that would require knowing the density of the substance.
The coincidence that makes this dangerous is that water is close to 1 gram per millilitre, so for water the numbers nearly line up. For flour, honey, oil or anything else they do not, and a recipe that says "8 oz" without specifying which is genuinely ambiguous. Honey is roughly 1.4 times as dense as water; measuring it by volume when mass was intended is a 40% error.
There is also a troy ounce, used for precious metals, at about 31.1 grams. Gold prices are quoted per troy ounce, so treating it as an ordinary ounce is a 10% error on something expensive.
| Name | Mass | Where used |
|---|---|---|
| Short ton | 2,000 lb = 907.2 kg | United States — usually just "ton" |
| Long ton | 2,240 lb = 1,016 kg | Imperial — shipping, older UK usage |
| Tonne (metric ton) | 1,000 kg | Everywhere else |
The spread between the largest and smallest is about 12%. In freight and commodity contexts the ambiguity is expensive, which is why careful writing specifies "metric tonne" rather than relying on spelling alone to carry the distinction.
This one causes more confused support requests than everything above combined.
The SI prefix kilo means exactly 1,000. But computer memory is organised in powers of two, and 210 is 1,024 — close enough to 1,000 that early practice borrowed the prefix. The two meanings then diverged and both stuck.
| Unit | Value | Typically used by |
|---|---|---|
| kB (kilobyte) | 1,000 bytes | Storage manufacturers, networking |
| KiB (kibibyte) | 1,024 bytes | Memory, many operating systems |
| MB / MiB | 1,000,000 / 1,048,576 | as above |
| GB / GiB | 109 / 1,073,741,824 | as above |
| TB / TiB | 1012 / 1,099,511,627,776 | as above |
The gap widens with each step: 2.4% at kilo, 4.9% at mega, 7.4% at giga, 10% at tera.
This is the "missing" disk space. A drive sold as 1 TB holds 1012 bytes, exactly as advertised. An operating system that divides by 1,024 three times reports 931 GB. Nothing is missing and nobody has cheated — the two numbers are the same quantity expressed with different meanings of the same prefix. Some systems now report in decimal units and show the full 1 TB, which is why the same drive can read differently on two machines.
A related confusion: network speeds are quoted in bits per second, file sizes in bytes. A 100 Mbps connection transfers at most about 12.5 MB per second. The lowercase b versus uppercase B is the only thing distinguishing them, and it carries a factor of eight.
A conversion cannot create accuracy that the input never had.
"About 10 miles" converted mechanically gives 16.09344 kilometres. The original figure was accurate to perhaps a mile; the answer claims accuracy to a centimetre. Every digit after "16" is noise dressed as information, and in engineering or scientific contexts it is actively misleading.
The rule of thumb: the result should carry roughly the same number of significant figures as the input. Ten miles is 16 kilometres. 10.0 miles is 16.1 kilometres. 10.00 miles is 16.09 kilometres.
The related trap is compounding. Converting inches to centimetres, rounding, then converting to metres and rounding again accumulates error at each step. Convert once, from the original value to the final unit, and round only at the end.