Risk-Driven Planning

Before detailing a product direction, design, architecture or plan, resolve unknowns whose answers could invalidate the next commitment. Do not try to eliminate every possible uncertainty before useful work can begin.

Classify uncertainty

• Verified/local: evidence is current enough for this decision.
• Carryable: uncertain, but either answer leaves the current bounded work viable; retain the assumption and its limits.
• Blocking for this commitment: a different answer could invalidate feasibility, architecture, task boundaries, dependency order, scope, or required acceptance evidence. Resolve it before the dependent commitment, not before all independent work.

Method

1. Identify the few assumptions this decision or proposed task relies on.
2. For each, explain what a different answer would actually change. A possible concern is not automatically a new requirement.
3. For blocking uncertainty, choose the smallest observation, research question, prototype or technical check that can distinguish the relevant answers. Give it a bounded outcome and an authorized owner before detailing dependent implementation.
4. Treat an authorized experiment as work that reduces uncertainty, not proof of its own hypothesis. Consciously accepted uncertainty about product value does not automatically block that experiment.
5. Place risky but executable work early when its result could invalidate substantial later effort. Preserve valid independent work rather than restarting everything.
6. Verify the risk where it becomes observable. Record actual evidence and update affected decisions through their existing owners. A proposed check is not an executed result.
7. If the same question returns without new evidence or a decision, do not repeat the same broad investigation. Save what is known and the remaining question, then narrow the check or identify the actual decision owner. A fresh conversation does not resolve the uncertainty, and a loop limit never waives a required guarantee.