Week 7: The Handover Slide Deck Is Not the Handover

Two teams can both finish their half of a project and the thing still fails completely at the seam between them — a $125 million spacecraft is the textbook case

19 April 2027

NASA’s Mars Climate Orbiter is the cleanest case study this course has, because both halves of the handover it failed at were, individually, finished correctly. Lockheed Martin built the spacecraft and delivered thruster-force data exactly as their process specified. JPL’s navigation team wrote trajectory software and fed that data in exactly as their process specified. The catch, in NASA’s own account of the failure — documented in the Mishap Investigation Board’s Phase I report — was that “one team used English units…while the other used metric units for a key spacecraft operation,” a factor of roughly 4.45 that nobody on either finished team caught, because catching it was nobody’s explicit job. The $125 million spacecraft flew 57 kilometres above Mars instead of 140, hit atmosphere it was never built to survive, and broke apart.

That’s the specific, expensive thing about this domain’s last mile: unlike a solo software team’s shortcuts, or a construction site’s punch list, the failure here doesn’t live inside either team’s finished work — it lives in the boundary between them, where a document was traded for knowledge that never fully fit on the page. Explicit knowledge — formats, specs, file layouts — survives a handover reliably. Tacit knowledge — the assumption that “force” obviously means one particular unit, because that’s what everyone in this building has always meant by it — does not, and nothing in a standard handover checklist is designed to surface an assumption nobody thought to write down because nobody thought it was a choice.

The complication this adds to the thesis: “done” can be true twice and the project can still fail completely. Lockheed finished. JPL finished. Neither team’s retrospective would find anything to fix in their own process, because there wasn’t anything wrong with their own process — the defect only exists in the space between two teams’ definitions of done, which is precisely the space that neither team’s own quality checks were ever built to look at. NASA’s own review called it “an end-to-end process problem,” not a mistake by either team — which is the whole point. When the last mile is the seam between two finished things, no amount of finishing either thing better would have caught it; only someone explicitly responsible for the seam itself could have.