Decision frameworks

Most bad decisions aren't bad judgment, they're the right judgment applied to the wrong kind of decision. A reversible, cheap choice deserves a different process than an irreversible, expensive one. Knowing which framework to use is half the skill. Executing it is the other half.

The first question: Type 1 or Type 2?

Bezos's framing. Two categories:

• Type 1, one-way door. Irreversible or nearly so. Hiring a senior exec. Selling the company. Signing a 5-year lease. Deliberate, analytical, high-conviction.
• Type 2, two-way door. Reversible with low cost. Launching a feature. Trying a new marketing channel. Hiring a junior. Fast, biased toward action.

The mistake most orgs make: treating Type 2 decisions like Type 1 (too slow), and Type 1 decisions like Type 2 (too careless). Get the type right first. Then pick the framework.

Framework 1. RAPID (for cross-functional decisions)

For decisions that span teams. Name each role explicitly:

• Recommend, proposes the decision, does the analysis
• Agree, people whose sign-off is required (regulatory, legal, etc.)
• Perform, who executes once decided
• Input, people consulted for information
• Decide, the single person who chooses

The critical thing is that D is a single person. Not a committee. Not a consensus. One name.

Framework 2. Expected value

When you can assign rough probabilities and values to outcomes:

EV = Σ (probability of outcome × value of outcome)

Example. Should we pursue this $2M deal?
- 30% chance we close at $2M → +$600K
- 40% chance we close at $1M (discounted) → +$400K
- 30% chance we lose → $0
EV = $1M.
Cost to pursue: ~$150K (sales engineering, exec time). EV/cost = 6.6x. Pursue.

The numbers are rough. The discipline is forcing yourself to actually assign probabilities instead of "it feels like a good opportunity."

Framework 3. Inversion

From Charlie Munger. Instead of asking "how do we succeed?", ask "how do we guarantee failure?" Then don't do those things. Often easier to generate a list of ways to fail than a list of ways to succeed.

Example. How do we guarantee our product launch fails?
- Launch without doing any customer discovery
- Target a segment we've never sold to
- Don't write positioning before building
- Assume virality
- Don't staff support
Now: are we doing any of those?

Framework 4. Regret minimization

When the expected-value math is too noisy to trust. Ask: "In 10 years, which choice will I regret more?" Particularly useful for career + life-scale decisions, but also for bet-the-company moves.

Framework 5. 70% rule

General Jim Mattis's rule: "If you have 70% of the information, go." Waiting for 90% information usually means you've missed the window. 70% + good judgment beats 90% + late every time. For Type 2 decisions especially.

Framework 6. Pre-mortem

Covered in detail: Pre-mortems. For high-stakes Type 1 decisions, always.

The decision memo

For any meaningful decision, write 1–2 pages before the decision. Template:

• Decision to be made (one sentence)
• Context (why now?)
• Options considered (usually 3)
• Recommendation + reasoning
• Risks + what would change the answer
• Reversibility (Type 1 or 2?)

The memo goes in the decision log. When you're in the situation six months later wondering "what were we thinking?", this is the answer.

What good looks like

• Every Type 1 decision produces a written memo before execution
• The decision-maker is always a single named person
• The team can distinguish Type 1 from Type 2 decisions and treat them accordingly
• Regular retros revisit past decisions against outcomes, not to blame, to learn

Related: Pre-mortems · What to kill · Decision logs