At 2:00 in the morning, Pacific time, on September 23, 1999, a spacecraft the size of a refrigerator fired its main engine as it swung toward Mars. The Mars Climate Orbiter had spent nine and a half months crossing about 416 million miles of space. The burn was supposed to slow it just enough to be captured into orbit. A little over four minutes later, the orbiter slipped behind the planet, exactly as planned, and its radio went quiet. It was due to reappear about 21 minutes later. It never did.
A week later, engineers found the reason, and it was almost embarrassingly ordinary. One piece of software had been reporting thruster data in pounds. Another piece of software, written by a different team, assumed those numbers were in newtons. Nobody caught the difference for nine months. The result was a spacecraft that flew far too close to Mars and was lost, and one of the most famous cautionary tales in engineering. It is also not the only time a unit mix-up has nearly brought down a machine, or did.
A cheaper, faster way to Mars
The orbiter was a product of a new philosophy at NASA in the 1990s, summed up as “faster, better, cheaper.” After the expensive Mars Observer vanished in 1993, the agency switched to smaller, cheaper probes launched more often. The Mars Climate Orbiter was half of a two-spacecraft mission called Mars Surveyor ’98. It would study Martian weather, dust and water vapor, and it would relay signals for its partner, the Mars Polar Lander, which was due to touch down near the planet’s south pole in December 1999.
Lockheed Martin built the spacecraft in Colorado. NASA’s Jet Propulsion Laboratory in Pasadena, California, managed the mission and handled navigation. That split matters to the story, because the error lived exactly at the seam between the two organizations. Together, the orbiter and lander cost about $327.6 million, including launch and operations.
The orbiter left Cape Canaveral, Florida, on a Delta II rocket on December 11, 1998. It weighed 638 kilograms (about 1,400 pounds) at launch, much of it propellant, and it carried two instruments to map the Martian atmosphere.
Tiny pushes, added up for nine months
A spacecraft in deep space does not just coast. Sunlight exerts a faint but constant pressure on it, and the Mars Climate Orbiter had a single, lopsided solar panel sticking out to one side. That uneven shape meant sunlight kept trying to twist the spacecraft. Spinning wheels inside the orbiter, called reaction wheels, soaked up the twist, but every so often they had to be slowed down. To do that, the orbiter fired small thrusters, a maneuver called an angular momentum desaturation.
Each of those firings also nudged the spacecraft slightly off course, so the navigation team at JPL needed to know how hard each one pushed. That information came from ground software supplied by Lockheed Martin, which wrote the results to a file. The interface specification, the written agreement between the teams, said the numbers had to be in newton-seconds, the metric unit for impulse, or force multiplied by time. The software produced them in pound-force seconds instead.
One pound of force equals about 4.45 newtons. So every time the navigators fed those numbers into their trajectory models, they underestimated the effect of each thruster firing by a factor of 4.45. On its own, each error was minuscule. But because of the lopsided solar panel, the firings happened 10 to 14 times more often than the navigation team had expected, and the small errors piled up over the nine-month cruise.
The warning signs
The trouble did not go completely unnoticed. Through the spring and summer of 1999, navigators saw that their predictions did not quite match the tracking data. Solutions based on the Doppler shift of the spacecraft’s radio signal kept showing it heading closer to Mars than other methods did. The concerns were raised, but mostly informally, and they were never resolved. According to later accounts, at least two navigators who flagged the problem were told they had not followed the formal process for documenting it.
On September 15, the team fired the thrusters for the last planned course correction, aiming for a first pass about 226 kilometers (140 miles) above the surface. Over the following week, estimates of that altitude kept slipping, to 150 to 170 kilometers. In the final hour before the engine burn, fresh tracking data put the closest approach at about 110 kilometers. Engineers believed the orbiter could survive no lower than 80 kilometers. An optional fifth course correction was discussed, but it never happened.
The real number, as investigators later reconstructed it, was about 57 kilometers, or 35 miles. At that height the spacecraft would have plowed into the thick lower layers of the Martian atmosphere. It either burned up or skipped back out into orbit around the Sun, damaged beyond use. Nobody knows which.
Whose fault was it?
NASA’s Mishap Investigation Board released its first report on November 10, 1999. It named the root cause plainly: the failure to use metric units in the coding of a ground software file called “Small Forces,” which fed the trajectory models. But the board was just as interested in why such a simple mistake survived for so long. It pointed to inadequate testing of the software, poor communication between the spacecraft team and the navigators, a navigation team that was stretched thin, and a process that did not force anyone to settle the nagging discrepancies.
NASA officials did not pin the loss on Lockheed alone. The agency said its own job was to catch errors like this, and its checks had failed. The mistake was not that someone used pounds. It was that the system had no reliable way to notice.
Worse news followed. On December 3, 1999, the Mars Polar Lander also fell silent during its descent. A separate review concluded that it had most likely shut off its engines too early, after its software mistook the jolt of its legs swinging open for touchdown. Two missions lost in under three months led to a broad rethink of “faster, better, cheaper.” An independent review found the Mars Surveyor ’98 program had been underfunded by at least 30 percent.
The Mars Climate Orbiter now sits alongside disasters like the collapse of the Tacoma Narrows Bridge as a classic in engineering classrooms. The science it was meant to do was not lost for good, though. Versions of its instruments later flew on the Mars Reconnaissance Orbiter, which reached Mars in 2006 and is still working.
Why does America still have two systems?
The deeper question is why an American spacecraft project was juggling two measurement systems in the first place. The United States has had one foot in the metric world for a very long time. Congress made the metric system legal for trade in 1866, and the country was one of the original signers of the international Metre Convention in 1875. Since the Mendenhall Order of 1893, the foot and the pound have officially been defined in terms of the meter and the kilogram. In 1959, the United States and other English-speaking countries fixed the inch at exactly 2.54 centimeters.
For more than a century, the country’s official mass standard was a small platinum-iridium cylinder kept at what is now the National Institute of Standards and Technology. Called K20, it was a national copy of the international prototype kilogram in France. Every pound in America was, by law, a fraction of that metal cylinder. Since 2019, the kilogram itself has been defined by a constant of nature instead.
A real push to go fully metric came in 1975, when Congress passed the Metric Conversion Act. But conversion was voluntary, the public largely shrugged or pushed back, and the board set up to manage the change was abolished in 1982. Since 1988, the law has called metric the “preferred system” for U.S. trade and commerce, and federal agencies are supposed to use it where practical. In everyday life, though, Americans still buy gas by the gallon and measure height in feet.
Science is a different story. American laboratories work almost entirely in metric, and so does NASA. Much of the aerospace industry, however, grew up with inches and pounds, and many of its tools, drawings and habits still use them. That is exactly the kind of mixed environment where a single unlabeled number can slip through. Precision in space depends on getting the small things right, whether it is timing GPS signals to a few billionths of a second or tracking fragments of space junk in crowded orbits.
The jet that ran out of fuel at 41,000 feet
Sixteen years before the orbiter was lost, a mix-up between pounds and kilograms nearly killed 69 people in Canada. On July 23, 1983, Air Canada Flight 143, a brand-new Boeing 767, was flying from Montreal to Edmonton when both engines stopped over Ontario. It had run out of fuel.
Canada was in the middle of switching to metric, and the 767s were the first planes in Air Canada’s fleet to measure fuel in kilograms. This particular jet’s fuel gauges were not working, so the crew had to measure the fuel in the tanks by hand with a dipstick and calculate how much more to add. The fueler gave them the density of jet fuel as 1.77, a figure in pounds per liter, which is what the rest of the fleet used. The crew used it as if it were kilograms per liter. The plane took off with less than half the fuel it needed.
Captain Bob Pearson, who happened to be an experienced glider pilot, flew the powerless 767 for about 17 minutes. First Officer Maurice Quintal remembered a former air force base at Gimli, Manitoba, where he had once served. Neither man knew that part of it had become a racetrack, and that a car race was underway when they arrived. The nose gear failed to lock and collapsed on landing, and the friction from the nose scraping along the runway helped stop the plane short of the crowds. Nobody was seriously hurt.
The aircraft, registered C-GAUN, was repaired and flew for Air Canada for nearly 25 more years, nicknamed the Gimli Glider. Investigators blamed the airline’s procedures, training and manuals more than the pilots, and they recommended converting the whole fleet to a single system, since a mix of units was more dangerous than either one alone. In 1985, Pearson and Quintal received an award for outstanding airmanship. Other crews who tried the same scenario in a flight simulator crashed.
A royal warship and two kinds of feet
Unit confusion is far older than computers. In 1628, the Swedish warship Vasa sailed out of Stockholm harbor on her maiden voyage, leaned over in a gust of wind, and sank less than a mile from where she started. The main problem was that she was too tall and top-heavy for her narrow hull.
When the ship was raised in 1961 and studied, archaeologists found four rulers that the workmen had used. Two were marked in Swedish feet, divided into 12 inches. The other two were marked in Amsterdam feet, which were shorter and divided into only 11 inches. Different teams appear to have used different rulers, and researchers have linked this to the hull being heavier on one side. It was not the main reason Vasa sank, but it is a striking example of how two sets of measurements can quietly get built into the same structure. Today the ship, remarkably intact, is the centerpiece of the Vasa Museum in Stockholm.
How engineers keep units straight now
The lessons from these accidents are surprisingly practical. The first is simply to write the unit down, every time. A bare number like 1.77 or 4.2 is an accident waiting to happen. Many engineering teams now require units in every column of a data file and every field of an interface, and they test the connections between systems, not just each system on its own.
Software can help. Some programming languages and libraries let developers attach units to numbers, so that the computer refuses to add newtons to pounds, or quietly converts them correctly. The other lesson is human. In both the Mars Climate Orbiter and the Gimli Glider, someone noticed that something seemed off. What was missing was a system that took those doubts seriously before it was too late, the same kind of careful second look that engineers later used to stabilize the Leaning Tower of Pisa.
NASA itself has leaned further into metric since 1999, and in 2007 it said it would use metric units for operations on the Moon. Still, the old units have never quite gone away, and spacecraft are still built by many teams working with many tools. The Mars Climate Orbiter is a reminder that the most dangerous number in a project is not always the hardest one to calculate. Sometimes it is the one that everyone assumed someone else had checked.
Measure it both ways
NEIKO 01407A Electronic Digital Caliper, 0–6 Inches — A stainless steel caliper with a large LCD and one-button switching between inches, fractions and millimeters, so you can see the same part measured in both systems. Handy for 3D printing, woodworking and repairs.
Etekcity Digital Food Kitchen Scale, Grams and Ounces, 11 lb — A slim stainless steel kitchen scale that weighs up to 11 pounds and switches between grams and ounces with a tap, which makes metric recipes easy to follow.