The truck meter counts litres or gallons. The aircraft counts kilograms or pounds, because mass is what the wing lifts and what the engine burns. Between the two there is a multiplication, and the number you multiply by is density.
Jet A-1 runs somewhere between 0.775 and 0.840 kg per litre depending on the batch and how warm it is. A US gallon comes out around 6.7 lb, near enough, most days. Nobody looks closely, because it has always worked.
Air Canada 143
In July 1983 a Boeing 767 left Montreal for Edmonton needing 22,300 kg of fuel. It was the airline’s first metric aircraft. The fuel quantity system was unserviceable, so the fuel already on board was measured by dripstick, in litres, and converted by hand.
The figure used was 1.77. That is pounds per litre. The correct one was 0.803, kilograms per litre. The multiplication was done correctly. The answer was 22,300 of the wrong thing, and the aircraft departed with less than half the fuel it needed.
Both engines stopped at 41,000 feet. The crew landed it at Gimli, Manitoba, on a runway that had been turned into a drag strip. Nobody died.
What I take from it is not that somebody was sloppy. Several people looked at that number, and each of them checked the arithmetic. None of them checked the unit. Arithmetic is easy to re-do, and re-doing it tells you nothing about whether you started in the right place.
The same thing happens in the hangar
Torque comes in inch-pounds, foot-pounds and newton-metres. The gap between the first two is a factor of twelve and it is available on every wrench in the building. Pressures come in psi and in bar. Manuals from different manufacturers, on the same aircraft, will not agree with each other.
NASA lost the Mars Climate Orbiter in 1999 the same way — one team supplied thrust data in pound-force seconds, the software expected newton-seconds, and the spacecraft flew into the atmosphere of Mars while doing exactly what it was told.
What actually helps
Write the unit down every time. “Uplift 3,000” is not an instruction.
Convert once, and put the density and the temperature on the paperwork. Not in somebody’s head, and not in a phone.
Check by expectation rather than by recalculation. Does this aircraft normally take this much? A unit error is almost never slightly wrong. It is wrong by a factor, so it shows up immediately if anyone asks whether the number looks like the usual number.
Decide which system your organisation works in, and hold it. Mixed fleets make this difficult, which is the reason to decide rather than let it happen.
And write the fallback procedure properly. When the fuel quantity system failed at Montreal, the manual method that replaced it was not covered the way the normal method was covered. That is the hole, and most maintenance organisations still have it somewhere.