Think like Munger

Vice chairman of Berkshire Hathaway for decades. He collected models from many fields into a "latticework" and leaned on a few habits: invert hard problems, never underestimate incentives, and use checklists of predictable human misjudgments before deciding.

When to use

• A decision that is expensive, slow to reverse, or affects many people
• A plan with no visible opposition (suspicious)
• Choosing vendors, tools, or partners whose incentives differ from yours
• Anything where your team is outside its usual expertise

Method

1. Invert. Instead of "how do we make this succeed?", ask "what would guarantee this fails?" List the ways, then check the plan against each.
2. Follow the incentives. For every party (vendor, team, manager, user, you): what are they rewarded for? Where do their incentives point away from the outcome you want? Expect behavior to follow incentives, not promises.
3. Run a bias checklist against the reasoning so far:
   • Consistency and commitment (defending an earlier choice)
   • Social proof ("everyone uses it")
   • Authority (a senior person said so)
   • Availability (the last incident dominates thinking)
   • Sunk cost (we already spent three months)
   • Overconfidence and excessive optimism
   • Envy or fear of missing out (competitors did it)
4. Circle of competence. Do we actually understand this domain well enough to judge? If not, either learn the key parts, get someone who does, or choose the option that needs less expertise.
5. Margin of safety. Prefer options that still work if your estimates are wrong by a lot.

Output

• Inversion: ways this fails, each with "covered / not covered by the plan"
• Incentives: party → what they're rewarded for → risk to us
• Biases spotted: which ones, and where they show up in the reasoning
• Competence: inside or outside the circle, and what to do about it
• Decision: recommended option with its margin of safety
• Write in the user's language.

Anti-patterns

• Only asking how to succeed
• Trusting a vendor's roadmap over its pricing model
• Treating "we already invested" as a reason
• Deciding outside your expertise without saying so

Example (illustration)

User: Should we move our whole backend to a new serverless platform? Their sales team offered a big discount for a 3-year deal.

Inversion: fails if cold starts break our 300 ms SLA (not covered), if the vendor raises prices after year 3 (not covered), if our team can't debug it at 2 a.m. (not covered).

Incentives: the sales rep is paid on contract length; the discount rewards lock-in. Our platform team lead is excited to learn it (career incentive).

Biases spotted: social proof ("big companies use it"), envy (a competitor announced their migration).

Decision: pilot one non-critical service for 3 months, monthly billing, measure p95 and on-call pain. No 3-year contract until then.

Sources

• Charlie Munger, Poor Charlie's Almanack (Stripe Press edition, 2023), including "The Psychology of Human Misjudgment" and the 1986 Harvard School commencement speech on inversion: https://www.stripe.press/poor-charlies-almanack
• Berkshire Hathaway shareholder letters and annual meetings: https://www.berkshirehathaway.com/letters/letters.html