When Unit Conversion Goes Wrong: Four Expensive Failures
A lost Mars orbiter, an airliner out of fuel at 41,000 feet, a cancelled rocket launch and a fatal overdose — all traced to unit errors.
Unit errors are not a theoretical concern. They have destroyed spacecraft, grounded airliners and killed patients. What makes them dangerous is that a wrong-unit answer is usually a plausible number — it passes the eyeball test in a way that a genuinely corrupt value would not.
Mars Climate Orbiter, 1999
NASA lost a $327 million spacecraft to a units mismatch.
The orbiter used small thruster firings to manage angular momentum. Lockheed Martin's ground software produced impulse figures in pound-force seconds. NASA JPL's navigation software expected newton-seconds.
A pound-force second is about 4.45 newton-seconds, so every value fed into the trajectory model was too small by a factor of 4.45. The error accumulated over nine months of cruise.
The spacecraft was expected to reach periapsis at about 226 km altitude. It actually arrived at roughly 57 km — far inside the atmosphere — and was destroyed.
The Gimli Glider, 1983
Air Canada Flight 143 ran out of fuel at 41,000 feet over Ontario.
Canada was converting to metric, and the Boeing 767 was among the first Air Canada aircraft specified in kilograms rather than pounds. The fuel quantity system was inoperative, so crew calculated fuel load manually using a dipstick reading in centimetres.
Converting volume to mass requires fuel density. The figure used was 1.77 — pounds per litre. The correct figure for kilograms was 0.803 per litre.
The aircraft ended up with roughly half the fuel it needed. Both engines flamed out in cruise.
What followed was extraordinary. Captain Robert Pearson was an experienced glider pilot and flew the powerless 767 in a forward slip to lose altitude. First Officer Maurice Quintal identified a closed RCAF base at Gimli — unaware that part of it had become a motor racing circuit, in use that day. The aircraft landed on the strip, the nose gear collapsed, and the resulting friction helped stop it short of spectators. Nobody was seriously hurt.
Institute for Space and Astronautical Science, 1996 — and a broader pattern
Unit confusion in aerospace is persistent rather than rare. Documented incidents include mixed metric and imperial fasteners causing assembly errors, thrust figures misapplied between systems, and altitude clearances misread between feet and metres in airspace where conventions differ.
ICAO standardisation on feet for altitude was driven precisely by this: the risk of inconsistency between aircraft in shared airspace outweighs the metric argument.
Medication dosing errors
The most consequential unit errors happen in hospitals, one patient at a time.
Common patterns:
- Micrograms versus milligrams — a thousandfold error. The abbreviation "µg" handwritten can resemble "mg", which is why many health systems now mandate writing "microgram" in full.
- Pounds versus kilograms — paediatric doses are weight-based, and entering a weight in pounds where kilograms are expected produces a 2.2× overdose.
- Total dose versus dose per kilogram — not strictly a unit error but the same class of failure.
- Trailing zeros — "1.0 mg" misread as "10 mg" if the decimal point is faint. Standards now forbid trailing zeros and require a leading zero: "0.5 mg", never ".5 mg".
These patterns are well documented by medication-safety organisations, and the mitigations — standardised abbreviations, weight recorded only in kilograms, independent double-checks for high-risk drugs, computerised order entry with dose-range checking — exist because the errors were common enough to demand systemic fixes.
What the failures have in common
- The wrong answer looked reasonable. A fuel figure, a thrust value, a dose — each was in a range that did not obviously alarm anyone.
- The error crossed an interface. Between contractors, between systems, between a written order and its execution. Unit assumptions are usually implicit at exactly these boundaries.
- A transition was under way. Canada mid-metrication, aerospace mixing conventions. Transitions are the highest-risk period precisely because both systems are live.
- Individual competence was not the issue. These were capable professionals. The defence has to be in the system, not in vigilance.
Practical defences
- Label units on every value — in variable names, in spreadsheet
column headers, on every interface.
mass_kg, notmass. - Specify units in interface contracts and validate them at boundaries.
- Sanity-check magnitudes. Does this dose look like a normal dose for this drug? Does this fuel load look like a normal load for this route?
- Use dimensional analysis. Carry units through the calculation; if they do not cancel to the expected result, the formula is wrong.
- Independent double-check for high-consequence values, performed separately rather than by reviewing someone else's arithmetic.
This site exists to make conversion fast and correct. It cannot tell you that you selected the wrong unit — that check belongs to you, and for anything consequential, to a second person.
What happened afterwards
The Mars Climate Orbiter loss pushed space agencies to tighten how interfaces are specified. Review boards recommended explicit unit definitions in interface documents, better verification of software inputs, and clearer channels for engineers to voice concerns. Later missions put far more weight on end-to-end checks across organisational borders.
The Gimli Glider story has a quieter epilogue. The aircraft was repaired and returned to service, flying for Air Canada for years afterwards. Airlines revised fuelling procedures and cross-checks so that no single manual calculation could decide how much fuel was loaded.
Why transitions are dangerous
A fleet or hospital that is changing conventions has two sets of numbers in circulation at once. Staff trained on one system pick up equipment, labels or forms built for the other. The hazard peaks when the old and new coexist, which is why changeovers usually include temporary double labelling, staff training and a firm cut-off date.
Software and units
A bare number in code carries no unit. Some programming languages and libraries now attach units to values, so adding metres to feet either converts automatically or refuses to run. Even without such tools, naming variables with their units (thrust_newton_seconds) and checking them at boundaries removes a whole class of silent error.
A short checklist before trusting a converted number
- Write down the unit next to every figure, including the intermediate ones.
- Estimate the answer roughly in your head and compare it with the result.
- Check whether the conversion involves a hidden quantity, such as density, that you must supply.
- Have a colleague repeat the calculation without seeing your answer.
Everyday echoes
Most mix-ups are far smaller than a lost spacecraft: a recipe in the wrong cup size, a bank transfer in the wrong currency, a tyre inflated to bar when the gauge said psi. The mechanism is identical, which makes household examples good practice for the habits that matter at work.
Frequently asked questions
Was the Mars Climate Orbiter loss purely a maths error?
No. The arithmetic was easy; the failure was in how the interface was specified, checked and escalated.
Why did the Gimli Glider crew have to calculate fuel by hand?
The aircraft’s fuel quantity indication was not working, so the load had to be worked out from dipstick readings and density figures.
Which unit errors are most common in medicine?
Microgram versus milligram confusion, pounds versus kilograms for weight-based doses, and decimal-point misreadings are widely cited as examples.
Can software prevent all unit errors?
No, but unit-aware code, validation at interfaces and independent review make most of them much less likely to slip through.
What is the single best habit?
Always carry the unit with the number, in writing, at every step.