The short answer
Unit conversion errors have caused some of the most expensive and embarrassing engineering disasters in history. The Mars Climate Orbiter (1999), the Air Canada Gimli Glider incident (1983), the Vasa warship sinking (1628), and several others share a common pattern: two teams using different measurement systems exchange data through an interface that does not flag the mismatch. The cost ranges from one warship to nine-figure space missions.
The single biggest lesson learned is that interface specifications must be explicit about units, and that runtime checks should validate unit consistency wherever possible. Modern engineering practice in aerospace, civil engineering, and medical devices includes unit-aware data types and automated checks. Most of these tools exist because of the specific disasters described in this guide.
The cases are also a useful illustration of why standardization matters. The 1959 International Yard and Pound Agreement that fixed one foot at exactly 0.3048 m, and the broader adoption of SI units in international engineering, exist precisely to prevent the kind of confusion that destroyed the Mars Climate Orbiter.
Case 1: Mars Climate Orbiter (1999)
The most famous unit conversion disaster of the modern era. NASA’s Mars Climate Orbiter was a $327 million spacecraft designed to study the Martian climate from orbit. On 23 September 1999, after a 286-day journey, it entered the Martian atmosphere at an altitude of about 57 km rather than the planned 226 km. The atmosphere destroyed it.
The investigation found that the cause was a software interface error. Lockheed Martin, which built the spacecraft, used English units (pound-force seconds) for the impulse calculations in the SM_FORCES software. The Jet Propulsion Laboratory, which operated the spacecraft, expected those values in metric units (newton-seconds). The conversion factor is 4.45 newtons per pound-force.
The accumulated error from many small trajectory corrections meant that by the time the spacecraft reached Mars, JPL’s navigation model thought it was on a course 100 km too high, while the actual trajectory was 100 km too low. The difference was caught by routine navigation checks, but the team’s confidence in the model meant they did not believe the discrepancy until the spacecraft was lost.
The official report identified two contributing root causes: the unit mismatch, and inadequate cross-checking between Lockheed Martin and JPL. The post-mortem led to NASA’s 2000 directive requiring SI units for new flight software interfaces.
Cost: $327 million spacecraft plus the lost science.
Lesson learned: every data interface must specify units. Most modern flight software uses unit-aware types (e.g., Python’s pint, C++‘s mp-units) that catch this class of error at compile time.
Case 2: Air Canada Flight 143, the Gimli Glider (1983)
On 23 July 1983, Air Canada Flight 143, a Boeing 767-200, ran out of fuel at 41,000 ft over Manitoba en route from Montreal to Edmonton. Both engines flamed out within minutes of each other. The crew successfully glided the aircraft 65 nautical miles (120 km) to a dead-stick landing at the abandoned Gimli air base, which had been converted to a drag racing strip. Two people were slightly injured during the evacuation. There were no fatalities.
The cause was a fuel-loading miscalculation. Air Canada had recently transitioned from imperial (pounds) to metric (kilograms) units for fuel quantity. The 767 was a newer-generation aircraft with all-metric documentation, but the ground crew on this flight used the conversion factor for pounds when they should have used the factor for kilograms. The aircraft was loaded with 22,300 pounds of fuel when 22,300 kilograms (about 49,200 pounds) were required.
The fuel quantity indicator system on the 767 was malfunctioning that day, so the crew relied on the manual calculation. The calculation used the wrong density value (1.77 pounds per liter instead of 0.803 kilograms per liter). The aircraft took off with roughly half the fuel it needed for the flight.
The crew’s airmanship in gliding the aircraft to a landing became one of the most-studied incidents in commercial aviation. Captain Robert Pearson was an experienced glider pilot, which directly contributed to his ability to fly the powerless 767. First Officer Maurice Quintal had been stationed at Gimli during military service and knew the abandoned runway was usable.
Cost: minor airframe damage, no fatalities, significant reputational impact for Air Canada. The aircraft was repaired and returned to service.
Lesson learned: unit transitions in operations require explicit retraining and double-checks. Modern fuel systems on Boeing aircraft display both kilograms and pounds, and the fuel-loading procedure includes mandatory cross-verification.
Case 3: Vasa warship (1628)
The Vasa was a Swedish warship commissioned by King Gustavus Adolphus. It was the most heavily armed warship in the world when launched on 10 August 1628 in Stockholm harbor. It traveled about 1,300 metres from the dock before a moderate gust of wind heeled it past its stability limit. The lower gun ports, which were open for the maiden-voyage salute, took on water. The ship sank in shallow water in front of the city.
The wreck was salvaged intact in 1961 and is now the centerpiece of the Vasa Museum in Stockholm. Modern analysis of the timbers, including unit-checking on the rulers used by individual carpenters, revealed something striking: the port and starboard halves of the hull were built using different foot units.
Swedish carpenters at the time used the Swedish foot of about 297 mm. Dutch carpenters working on the same project used the Amsterdam foot of about 283 mm, roughly 5 percent shorter. The mix of crews building each side of the hull resulted in subtle structural asymmetries. Combined with the high center of gravity from the heavy cannon and an undersized hull cross-section, the asymmetry contributed to the ship’s catastrophic stability problem.
The Vasa is therefore arguably the oldest documented engineering disaster caused by a unit confusion. The cause was diagnosed nearly 400 years after the fact, but the evidence is in the wood.
Cost: the ship and the crew of 150 (around 30 of whom died). The replacement warship took another decade to build.
Lesson learned: standardized units within a single project are essential, especially across crews from different countries. The shipbuilding industry largely standardized after the European metric reforms of the early 1800s.
Case 4: Korean Air Cargo Flight 6316 (1999)
On 15 April 1999, Korean Air Cargo Flight 6316, a McDonnell Douglas MD-11, crashed shortly after takeoff from Shanghai’s Hongqiao Airport. All three crew on board and five people on the ground were killed.
The investigation found that the captain mistakenly believed the controller’s altitude assignment in metric meters was in imperial feet. The Shanghai air traffic control system uses metric for altitude (consistent with Chinese practice), and the controller had assigned 1,500 meters. The captain read this as 1,500 feet and pitched the aircraft to descend instead of continuing the climb. The aircraft entered an unrecoverable nose-down attitude and crashed within seconds.
The crash highlighted the operational risk of the China-specific metric altitude system for international pilots used to ICAO standard feet. China subsequently improved its English-language radio phraseology to explicitly state “meters” or “feet” in altitude clearances, and pilot training for China operations now includes specific recognition of the metric altitude system.
Cost: 8 fatalities (3 crew, 5 on ground), one MD-11 aircraft.
Lesson learned: international aviation requires explicit unit identification in radio phraseology when operating in non-ICAO-standard airspace. The flight crew checklist for China routing now includes a specific reminder about metric altitudes.
Case 5: The medical case of vincristine overdose
Multiple medical incidents over the last 30 years have involved unit confusion between micrograms and milligrams. The most documented case involved the chemotherapy drug vincristine. The therapeutic dose is in milligrams per square meter of body surface area. A 10-times overdose, caused by a milligram-microgram mix-up in a hand-written prescription, was fatal to a teenage patient in 2001.
The error pattern is “milligram for microgram” or vice versa, often when a paper prescription is read in a hurry. Modern electronic prescribing systems flag dose units explicitly and require confirmation when the calculated dose differs from typical by more than a factor of three.
In the US, the FDA’s TALL-man lettering convention (writing “MILLIgram” and “MICROgram” in distinguishing case) is one of the operational interventions to reduce this class of error. Electronic medical records systems that store doses in unit-aware data types eliminate the error at the data layer.
Cost: individual patient deaths, no consolidated industry total. Estimated thousands of hospitalizations per year worldwide before electronic prescribing became standard.
Lesson learned: unit-aware data storage in critical systems is no longer optional. Modern hospital information systems require dose units to be stored as machine-readable codes, not free text, and the prescribing software validates the unit against the drug’s allowed range.
What changed
These disasters changed engineering practice in real ways:
- NASA and other space agencies now require SI units for new flight software interfaces. The exception is legacy ground systems where retrofitting is impractical.
- Aviation uses standardized radio phraseology with explicit unit identification when operating in non-standard airspace. Modern flight management systems display both imperial and metric units, with the active unit clearly indicated.
- Hospital prescribing systems validate doses against allowed unit ranges and require explicit confirmation of unusual values.
- Civil engineering now standardizes on the country’s official surveying system within any single project, with explicit conversion documentation at any interface to another system.
- Programming languages increasingly support unit-aware numerical types. Languages like F# (with units of measure) and libraries like Python’s pint or Boost.Units in C++ catch unit mismatches at compile or test time, before they reach production.
The general principle is the same one that motivated the 1959 International Yard and Pound Agreement: when systems built by different teams must interoperate, the units have to be either identical or explicitly converted at the interface, with the conversion documented and validated.
The cost of imperial-metric mix
The aggregate cost of unit conversion errors across the history of engineering is impossible to total exactly, but the documented disasters alone run into billions of dollars and dozens of fatalities. The recurring fix is always the same: be explicit about units, validate interfaces, and design systems so that a unit mismatch fails loudly rather than quietly.
For the everyday work of converting one meter to feet, the calculator on the homepage and the precomputed conversion pages give exact values without any of the human factor that has caused the disasters in this guide.
Sources and further reading: