Mars Climate Orbiter, 1999: the same number, twice
Lockheed Martin’s software calculated thruster impulse in pound-force; NASA’s navigation software read the same figures as newtons. No calculation error, no defective hardware — both teams were right inside their own system. The spacecraft entered the Martian atmosphere at 37 km instead of 57 km and burned up; mission cost $193.25 million. The investigation board named the cause explicitly: the failed handover of information between the teams. They shared the numbers — not their meaning.
Air Canada Flight 143, 1983: correct for every other aircraft
The “Gimli Glider” was Air Canada’s first metrically calibrated Boeing 767 in a fleet that calculated fuel in pounds. With the fuel gauge inoperative, the crew computed the load using a density figure of 1.77 from the fuel slip — a correct number, in pounds per litre, for every other aircraft in the fleet. Read as kilograms, it left the aircraft with 10,100 kg of fuel instead of the required 22,300 kg — 45%. Both engines stopped at altitude; the crew glided 17 minutes to a landing at Gimli with no fatalities. The inquiry’s conclusion generalises far beyond aviation: a fleet running two reference systems in parallel is more dangerous than either system alone — and the correct conversion had existed all along, as the private knowledge of individual people rather than as a procedure.
Laufenburg bridge, 2003: the misunderstanding was already solved
Germany measures height from the Amsterdam datum, Switzerland from a reference point at Geneva. When the two countries built a bridge across the Rhine at Laufenburg, the 27 cm difference was known, documented and included in the plans. A sign error applied the correction in the wrong direction — making the offset 54 cm, exactly double. It was caught at an inspection in December 2003, early enough to correct. The lesson is the uncomfortable one: the shared reference had been established, in writing — and then applied backwards. “We aligned on that” is not evidence that both sides did the same thing with the alignment.
What the three cases share
In none of them did anyone lack competence, diligence or goodwill. The failure lived between the parties, at the handover — where each side’s locally correct reference met the other’s. Organizations run such handovers every day, with softer words than newtons and kilograms: “done”, “approved”, “priority”, “acceptable risk”. The physics is more forgiving; the mechanism is the same.
Where this comes from. This page draws on The Hidden Cost of Speed by Christoph Svoboda — a book about what organizations lose when they optimize everything for velocity. About the book.
See it in your own answers. The Shared Reference Diagnostic measures the distance between how you actually operate and what your organization aspires to be — ten scenarios, about ten minutes, your own report, free.