Why a pilot that 'succeeds' still gets killed six months later

A pilot can hit every metric defined in the original proposal, accuracy, time saved, cost reduction, and still get quietly shelved within six months. This isn't a rare failure mode; it's common enough that it deserves its own explanation, separate from "the pilot didn't work."

The usual cause isn't the pilot's own performance, it's that the pilot was scoped narrowly enough to succeed in isolation but nobody planned how it would integrate into the surrounding workflow once it needed to scale: the champion who sponsored it moves to a different team, or a small tool that worked for fifty records a week hits real friction at five thousand.

The pilots that survive past six months are scoped with the scaling question answered upfront: who owns this after the champion leaves, what does the tooling and cost curve look like at ten times the current volume, and is there a second stakeholder besides the original sponsor who has a reason to keep it alive.

A pilot's own metrics answer "did this work." They don't answer "will someone still care in six months," and that second question is the one that actually determines whether a successful pilot becomes a permanent system or a slide in an old deck nobody opens again.