When the right answer is stop.

The claim

Some pilots should be stopped, and the evidence has usually said so for a while. A kill verdict is not an accusation. It is capital reallocation with evidence.

How it fails

Pilots accumulate sunk cost and career risk. Stopping one feels like admitting failure, so the team keeps funding a path the evidence has already closed. Reviews quietly change their question from "should this continue" to "what did we learn", which is the question you ask when the answer to the first one is no.

Warning signs

• The value case was written once, at kickoff, and never revisited.
• Nobody can state the conditions under which the pilot would be stopped.
• The team has quietly stopped proposing production dates.
• The strongest argument for continuing is the money already spent.

What it costs

Budget and senior attention stay trapped in a closed path while the next opportunity waits. The eventual stop happens by erosion - a reorg, a quiet defunding - and nobody captures what was learned, so the next pilot repeats it.

What would change the answer

Pre-agreed termination criteria and a review with the authority to stop. And one honest test: is the constraint structural - the workflow, the data, the economics - or is it one fix away? Structural means stop. One fix away means name the fix and the date.

Where this maps

This failure is a Payoff gate failure. The PROOF Scorecard tests this gate in about four minutes; the full review proves it with evidence.