Why Unit Conversions Go Wrong (Even for Engineers)
Failure mode one: mixed systems in one expression
The Mars Climate Orbiter was lost because one team used pound-force seconds while another expected newton-seconds. Any calculation that mixes imperial and metric inputs without converting first inherits the error silently — no exception is thrown, the number just comes out wrong.
Failure mode two: affine units treated as ratios
Temperature is the classic trap: 20°C is not "twice as hot" as 10°C because Celsius has an offset, not just a scale. Length, mass, and energy multiply cleanly; temperature, sound level and similar offset-based measures do not. A converter that ignores offsets gets every temperature question subtly wrong.
Failure mode three: premature rounding
Rounding 1 inch = 2.54 cm down to 2.5 loses 1.6% per step — compound that across a multi-step calculation and the final answer drifts visibly. Convert with full precision factors and round only the displayed final result.
A self-check that catches all three
Convert forward, then convert back. If you do not land on your starting value within display precision, one of the three failure modes is hiding in the middle of your chain. It takes seconds and works for any pair of units.