Document 2-3 pre-mortem failure modes with every…

For significant decisions, add a fifth field documenting pre-mortem risks (2-3 specific ways the decision could fail) to preserve concerns that hindsight bias will erase.

When This Fires

Documenting any significant decision (hiring, architecture, strategy, investment)
Writing ADRs (Architecture Decision Records) or decision logs
Committing team resources to a plan
Any decision that will be evaluated weeks or months later

Common Failure Mode

Writing vague risks: "It might not work" or "There could be resistance." These are so generic that they can't be evaluated post-decision. The failure modes must be specific enough to be falsifiable: "The latency increase from adding this middleware layer could push P99 above our 200ms SLA" or "The new hire's management experience is in B2C, and our B2B context may require different approaches." Specific risks can be checked; vague ones can't.

The Protocol

For each significant decision, add a pre-mortem field with 2-3 entries: (1) "This decision could fail if [specific mechanism]." (2) "The biggest uncontrolled risk is [specific factor]." (3) "The signal that would tell us this is failing is [specific observable]." Write these at the time of the decision, not after. Review the pre-mortem field when evaluating the decision's outcome. The gap between predicted and actual failure modes is where your risk-assessment skill improves.