Inversion & Pre-mortem
Core principle: Instead of asking "how do we make this succeed?", ask "how does this definitely fail?" — then work backwards. Surfaces hidden assumptions, fragile dependencies, and blind spots forward analysis misses.
Two Complementary Techniques
Technique 1: Inversion
Flip the problem. If you want to understand how a system succeeds, first rigorously define how it fails.
Process:
- State the goal clearly: *"We want X to succeed."*
- Invert it: *"What would guarantee X fails completely?"*
- List all failure conditions exhaustively — be adversarial, not optimistic
- For each failure condition: is it currently present, partially present, or guarded against?
- The unguarded ones become your risk register
Key question to ask: *"What assumptions must be true for this to work — and what happens if they're false?"*
Technique 2: Pre-mortem
Imagine you're 12 months in the future. The project/system/decision has failed badly. Now explain why.
Process:
- Vividly imagine the failure: *"It's [date]. This has completely fallen apart."*
- Write the failure story in past tense — what happened?
- Identify the 3–5 root causes that led to failure
- For each cause: what early warning signal would have been visible?
- Map signals back to today: which of those signals exist right now?
Output Format
💀 Failure Modes (Inversion)
For each identified failure mode:
- Condition: What must go wrong for this to fail?
- Likelihood: Low / Medium / High
- Currently guarded?: Yes / Partially / No
- What guards it (or what's missing)
🪦 The Failure Story (Pre-mortem)
A short narrative: *"It's [future date]. Here's what happened..."*
- Name the specific sequence of events
- Call out the moment where it became unrecoverable
- Identify what looked fine at the start but was actually fragile
⚠️ Hidden Assumptions
List the beliefs the plan depends on that haven't been validated:
- Technical assumptions
- Human/team behavior assumptions
- Market or user assumptions
- Dependencies on external systems or actors
🛡️ Mitigations
For each high-likelihood, unguarded failure mode:
- Concrete action to reduce risk
- Early warning metric to monitor
- Reversibility assessment: *Can we undo this if it fails?*
Thinking Triggers
Use these prompts to deepen the analysis:
- *"What is the single most likely way this fails?"*
- *"Who is most likely to be frustrated by this in 6 months, and why?"*
- *"What do we believe that might be wrong?"*
- *"If we had to bet against this succeeding, where would we put our money?"*
- *"What's the optimistic assumption hiding in plain sight?"*
Example Applications
- Evaluating a new feature: What user behaviors are we assuming? What if adoption is 10x lower than expected?
- Architecture decision: What if the third-party API we depend on changes its contract? What if latency doubles?
- Hiring or team change: What does this look like if the new hire isn't a fit after 3 months?
- Growth strategy: Imagine we executed perfectly and it still failed — why?