Learning needs an exit door. Without a kill condition, a bet becomes a habit, then a constituency, and eventually a permanent budget line.
This lane stays on product choices: which customer problems deserve capital, which bets deserve sequencing, and which roadmap items should lose funding. It does not drift into AI product design, growth loops, GTM motion, or generic execution cadence.
The common mistake with kill criteria is building documents that look complete but avoid forcing a decision. Teams describe the market, list stakeholders, and summarize data while ignoring the hard trade-offs. The more effective move is naming the decision boundary: what are we choosing, what are we refusing, who owns it, and what evidence would change our mind?
A roadmap item should carry an investment thesis: customer problem, target segment, expected behavior change, cost of delay, dependencies, learning milestone, and kill signal. If it cannot carry that weight, it is a task, not a strategy.
Operator artifact: build a kill-criteria sheet. Keep it small enough for a normal planning or review meeting. Include the decision and owner; evidence and tradeoff; next checkpoint and the condition that would force a change. If it doesn't fit on two pages, it is likely hiding weak thinking behind volume.
A productive review answers three questions: What did we learn? What will we stop doing? What decision changes now? If the meeting cannot answer those, the work might be useful background, but it isn't driving the business.
Field test: pick one current initiative and rewrite it through this lens before adding new process. If the rewrite exposes no tradeoff, no owner, and no next decision, the team has found the real work.
This is part 6 of 10 in Product Strategy That Actually Makes Choices.
