Mars Climate Orbiter: the units mix-up that lost a spacecraft
In September 1999, NASA lost the Mars Climate Orbiter as it arrived at Mars. The cause was not a broken engine or a bad launch. One piece of ground software produced numbers in English units, the software that used them expected metric, and nothing in the project caught the difference over a nine-month cruise.1
- Dec 11, 1998Launch
- 9 monthsCruise from Earth to Mars
- ~170 kmLower than planned at arrival
- 1 + 8Root cause, plus 8 contributing causes
Mars Climate Orbiter on a board: as it went, and with Kanbanchi
Switch between the two states to see the same five pieces of work, as they went and as they could have gone with an owner on every check.
As it went: the hand-off is marked delivered, and four checks around it sit in Open with no owner.With Kanbanchi: each check has a named owner, and the units check sits on the hand-off card where it could catch the mismatch.
This board is an illustration of how the same work could be tracked. It is not a reconstruction of NASA’s or Lockheed Martin’s actual tools or records.
The plan
NASA designed the Mars Climate Orbiter to be the first interplanetary weather satellite: it would orbit Mars, study its climate and act as a communications relay for the Mars Polar Lander, which was due to reach Mars in December 1999.1
The spacecraft launched on December 11, 1998.1 The work was split across organizations. Lockheed Martin Astronautics of Denver, the prime contractor, designed and developed the spacecraft and supported launch. NASA’s Jet Propulsion Laboratory (JPL) kept overall project management, systems engineering, mission design and navigation design, and a separate operations project flew the spacecraft after launch.2
The plan depended on a quiet, routine kind of teamwork: one group produces a data file, another reads it, and a written interface specification says what units each number is in. The specification for the file at the center of this story said metric: newton-seconds.2
What actually happened
Small thruster firings, small errors
During the nine-month cruise, the spacecraft periodically fired its thrusters to remove momentum that built up in its reaction wheels. Software called SM_FORCES (“small forces”) wrote the results of these events to a file, the Angular Momentum Desaturation (AMD) file, which the trajectory modelers used to predict where the spacecraft was going.1
The AMD file was required to be in metric units under the existing software interface documentation, and the trajectory modelers assumed it was. It was not: SM_FORCES used thruster performance data in English units.1 The report puts it plainly: the data came in pound-force seconds instead of the specified newton-seconds, so the navigation software underestimated the effect of each firing by a factor of 4.45, and an erroneous trajectory was computed.2
Ten to fourteen times more often than expected
The error was small for each event, but there were a lot of events. Because the orbiter’s solar array was asymmetrical, unlike the earlier Mars Global Surveyor’s, sunlight built up much more momentum on the spacecraft, and these momentum dumps happened 10 to 14 times more often than the navigation team expected.1 Each one added a little error to the trajectory estimate.1
Warning signs nobody resolved
Throughout the spring and summer of 1999, there were concerns at the working level about discrepancies between navigation solutions, but they were reported only informally. As the spacecraft approached Mars, solutions based only on Doppler tracking data consistently put the flight path closer to the planet. Those discrepancies were not resolved.2
Arrival
On September 23, 1999, the spacecraft began its Mars orbit insertion burn and passed behind the planet. It lost signal at 09:04:52 UTC, 49 seconds earlier than predicted, and the signal never came back.2 At insertion, its trajectory was approximately 170 kilometers lower than planned. The board concluded that it was either destroyed in the atmosphere or re-entered space around the Sun after leaving Mars’ atmosphere.1
Six days later, on September 29, the team found that the small forces data had been low by that factor of 4.45. Navigation estimates redone with corrected values put the lowest point of the orbit at 57 km, judged too low for the spacecraft to survive.2
Announcing the report in November 1999, board chairman Arthur Stephenson said the root cause was “the failed translation of English units into metric units in a segment of ground-based, navigation-related mission software.”3
What went wrong, in project management terms
The specification existed. Nobody owned the check against it.
This is the part that makes the story useful for any project manager. The interface documentation was right. The trajectory modelers did what the documentation said to expect. The failure was in the space between two teams, where the question “does this file actually match the spec?” belonged to nobody in particular.1 The board’s first recommendation for the Mars Polar Lander was to verify the consistent use of units throughout the design and operations, and the second was a software audit for specification compliance on all data transferred between JPL and Lockheed Martin Astronautics.1
A first-time hand-off between teams
NASA’s summary of the board’s findings says systems engineering was not robust enough, and that this was made worse by the first-time handover of a project from the team that built and launched it to a multi-mission operations team.3 The people who knew the spacecraft best were not the people flying it, and the navigation team was not fully informed about how the spacecraft was pointed in space.3 The report adds that the navigation team did not know until long after launch that the spacecraft itself calculated and sent to Earth velocity change data for these thruster events. An early comparison of that data with the tracking data might have uncovered the units problem.2
Concerns that stayed informal
The board found that communication among the engineering groups was too informal, and that training was insufficient, including training on how to file formal anomaly reports.3 One of its recommendations was to train the whole Mars Polar Lander team and encourage use of the Incident, Surprise, Anomaly process.1 The report is specific about what happened instead: when conflicts in the data were found, the team relied on e-mail to solve problems rather than the formal anomaly process, and failing to use the problem tracking system contributed to the problem “slipping through the cracks.”2 In plain terms: if something looks off, it should land somewhere official, with a name next to it.
A small team with no second pair of eyes
The operations navigation team was small, oversubscribed and had no independent peer review of its work.3 An optional fifth trajectory correction maneuver was considered but not performed.3 The board recommended independent reviews on all mission-critical events and an overall Mission Manager, a single person accountable for the whole.1
What Kanbanchi would have changed
No task board does orbital mechanics. What a board does well is make hand-offs and checks visible, give each one an owner and keep the spec next to the work. Here is what we would have put on that hand-off card.
Checklists
Put “Units match the interface spec” on the hand-off card as a checklist item with a named person, so the card cannot look finished while the check is open.
How to create a checklistCard templatesPremium
Save the hand-off card as a template, so every data delivery between teams arrives with the same checks, including units.
How to use card templatesGoogle Drive attachments
Attach the interface specification to the card straight from Google Drive, so both teams open the same file, not their own copy.
How to attach filesAssignees
Assign each check to one person. On the board, an unassigned check is easy to spot, which is the point.
How to assign a cardMentions in comments
When trajectory numbers look odd, @mention the other team in the card’s comments. The question gets to a named person and stays on record.
How to mention a colleagueTry it: Mars Climate Orbiter on a Kanbanchi board
Drag cards between lists with a mouse or a finger (press and hold on touch screens), or focus a card and use the arrow keys. Open a card to try the features from this article: Checklists, Card templates, Google Drive attachments, Assignees, @mentions in comments.
Press Enter to open the card, Left or Right arrow to move it to the next list, Up or Down to go through the list, Shift F10 for the move menu.
To Do
1- Operations
Orbit insertion
MG
In progress
2- Testing
Check trajectory numbers
- Operations
Plan trajectory correction
NV
Review
1- Development
Hand-off: thruster data file
0/2SP
Done
0What just happened
Nothing yet. Move a card or open one, and this box explains what Kanbanchi would show.
Try this
- Drag the hand-off card to Done with the units check open.
- Attach the interface spec from Drive so both teams open one file.
- Assign the trajectory check and post the navigation team’s question.
A simplified copy of the Kanbanchi board for this story. Cards, names and dates are illustrations, not a historical record. Nothing is saved: reload the page and the board starts over.
To be fair to the engineers: the board’s findings were about process, not about any single person, and no tool replaces systems engineering or a verification matrix. A board helps by making one simple thing hard to miss: a check with no owner.
Lessons for project managers
- Name an owner for every interface. A written specification is not a check. Someone has to own comparing the delivered file to it, and that person’s name should be on the card.
- Treat first-time hand-offs as high risk. When work moves from the team that built it to a team that runs it, expect gaps in context. The board pointed to exactly this kind of handover.3
- Make it easy to raise a concern formally. The board found the team relied on e-mail instead of the formal anomaly process.2 If the only way to flag a problem is a message thread, problems stay in threads. Give concerns a place to land, with a priority and an owner.
- Small errors repeated often are big errors. Each momentum dump added only a little error; there were 10 to 14 times more of them than expected.1 Watch for small issues that repeat.
- Get a second pair of eyes on critical work. The board recommended independent reviews on all mission-critical events.1 A review is a task. Put it on the board.
Mars Climate Orbiter by the numbers
| What | Figure | Source |
|---|---|---|
| Launch date | December 11, 1998 | 1 |
| Cruise to Mars | About 9 months | 1 |
| Loss of signal | 09:04:52 UTC, September 23, 1999, 49 seconds earlier than predicted | 2 |
| Lowest point of orbit, recalculated | 57 km, too low to survive | 2 |
| Momentum dumps vs. expected | 10 to 14 times more often | 1 |
| Unit error | Pound-force seconds instead of newton-seconds, a factor of about 4.45 | 2 |
| Trajectory at arrival | About 170 km lower than planned | 1 |
| Contributing causes found | 8, plus 1 root cause | 1 |
| Recommendations for Mars Polar Lander | 16 | 1 |
Mars Climate Orbiter: frequently asked questions
Why did the Mars Climate Orbiter fail?
NASA’s investigation board found that ground software called SM_FORCES produced thruster data in English units, while the interface specification and the navigation team expected metric units. Small errors built up over the nine-month cruise, and the spacecraft arrived about 170 km lower than planned.
Was it really just a units conversion error?
The units error was the root cause, but the board also listed eight contributing causes, including a first-time hand-off from the development team to the operations team, informal communication, a small navigation team without independent peer review and inadequate verification of ground software.
Who was responsible for the Mars Climate Orbiter loss?
The board framed it as a failure of process, not of one person. NASA’s Jet Propulsion Laboratory ran navigation, and Lockheed Martin Astronautics was the prime contractor that designed and developed the spacecraft. The mismatch sat in the interface between teams: a file that did not follow its own specification.
What is the project management lesson from the Mars Climate Orbiter?
Every interface between teams needs a named owner who checks the delivered work against the specification. A specification that nobody checks against does not protect the project.
How would Kanbanchi have helped?
Kanbanchi cannot do navigation, but it can put a hand-off on a card with a checklist item for units, an assignee, the spec attached from Google Drive and @mentions for questions. Card templates make the same checks appear on every hand-off.
Sources
- NASA Lessons Learned, Lesson 641: Mars Climate Orbiter Mishap Investigation Board, Phase I Report, NASA, November 1999. Accessed November 2, 2026.
- Mars Climate Orbiter Mishap Investigation Board, Phase I Report (PDF), NASA, November 10, 1999. Accessed November 2, 2026.
- NASA press release 99-134: Mars Climate Orbiter Failure Board Releases Report, NASA, November 10, 1999. Accessed November 2, 2026.
Figures come from NASA’s investigation board and NASA’s own press release. We have not added cost figures, because the sources above do not give one for this spacecraft alone.
Plan your next hand-off like it matters
Kanbanchi is a task board for Google Workspace and Microsoft 365. Give every check an owner, starting today.
Start your free trial