The roadmap review has one slide that never moves. Not because the feature is thriving — because nobody wants their name on the decision to end it. So it survives another quarter. Another line in the maintenance backlog. Another onboarding conversation where a new engineer asks why it exists, and nobody has a clean answer.
Kill decisions rarely fail from a lack of data. They fail from a lack of a rule. Everyone can point to the eight accounts still using it. Nobody can point to the threshold that says eight is too few.
The four signals that make a kill decision defensible, a worked example that shows all four converging on one call, and what has to already exist before you're allowed to pull the trigger.