Claim
Borrowed from Jacobi, Munger's inversion principle replaces the standard success question with its negative: list everything that would guarantee failure, then systematically avoid those conditions. Inversion catches failure modes that goal-orientation misses, especially in compounding domains (investing, hiring, marriage) where avoiding ruin matters more than optimizing.
Mechanism
Forward planning starts from the goal and selects positive moves. Inversion starts from disaster and selects which to avoid. The two are not symmetric: the failure list is usually shorter and more concrete (don't lose principal, don't bet the company on one customer, don't hire someone you don't trust), while the success list expands into infinite tactical decisions. Avoiding the failure list keeps you in the game long enough for compounding to work.
Conditions
Holds when:
- The domain has long horizons and asymmetric downside (capital allocation, leadership, health).
- The failure modes are namable in advance.
Fails when:
- Greenfield innovation work where survival is not the constraint, optionality is.
- Highly time-sensitive decisions where exhaustive failure-listing wastes the window.
Evidence
"Invert, always invert. Instead of asking 'How do I succeed?' you ask 'What would guarantee failure?' and then avoid those conditions."
"Knowing what you don't know is more useful than being brilliant."
— Charlie Munger (synthesized from operator's published work)
Signals
- Strategy reviews include a written "what would guarantee failure" list.
- Hiring loops have an explicit "what would make this a disaster hire?" filter.
- Capital decisions name the downside scenario before the upside.
Counter-evidence
Optimization cultures (high-frequency trading, growth marketing) argue that inversion biases toward conservatism — when the cost of action is reversible and small, forward optimization beats failure-avoidance. Reid Hoffman's "blitzscaling" thesis treats some categories as winner-take-all where avoiding failure is the surest way to lose.
Cross-references
- ins_latticework-of-mental-models — same operator