Decision-Making Mental Models & The Scientific Method
Most decision failures come from sophisticated analysis on the wrong question, or from reaching for the same tool every time.
Use these as a latticework. The right model depends on the question, so scan the whole shelf rather than grabbing your favorite. Apply #1 first, always.
Clarify what you are actually optimizing for before anything else. Stripe asks "What problem are we solving?" Amazon asks "What customer outcome are we producing?" Without this, every other framework optimizes for the wrong thing.
For empirical questions where evidence-based testing is possible. The rigorous version lives in Part Two below.
Break a problem down to foundational truths and reason up. Strip away assumptions and analogies; derive from physics, math, economics, biology. Essential for novel problems where familiar analogies mislead.
Imagine the failure state and work backward to its causes. "What would make this fail?" is often more productive than "What would make this succeed?" Many failure modes are predictable; many success paths are not. Munger's primary tool.
Multiply each outcome's probability by its value, then sum. Powerful with asymmetric framing: a small downside plus a large upside means take the bet even at low probability.
Map feedback loops, second-order effects, and emergent properties. Complex systems (a body, an economy, an organization) do not respond linearly. Ask: "What changes when this changes? What propagates? What loops back?"
Default skepticism. Most novel ideas do not pan out; most patterns are noise. Ask: "What's the base rate for this kind of claim succeeding?" and "What would I have seen if this weren't true?"
Two-way doors (reversible) deserve less analysis: decide fast, iterate. One-way doors (irreversible) deserve much more: slow down, gather evidence. Treating a reversible decision as irreversible is the more common, more costly mistake.
The most underrated force in life. Knowledge, relationships, training adaptation, and capital all compound. Most long-horizon gains come from small consistent actions, not single big decisions.
Every yes is a no to the alternatives, especially for time, focus, and capital. Pair with the North Star to surface which alternatives actually matter: "What am I giving up by saying yes here?"
For rigorous empirical reasoning when a question deserves more than intuition, especially when being wrong has real cost.
// The Process · 11 Steps
Notice a phenomenon worth understanding.
Frame what specifically you want to understand.
Survey prior knowledge: literature, expert input, prior data.
A testable, falsifiable, specific statement.
"If H is true, then X happens at Y magnitude."
Setup, controls, sample size, threats to validity.
Collect data per the design.
Statistics, pattern recognition, interpretation.
Accept, reject, or refine. Never "prove": you can only support or refute.
Refine and loop back. Real science is cyclical, not linear.
Communicate for scrutiny and replication: peer review plus reproducibility.
Apply these as a disposition on top of every step. Different clusters matter at different phases.