The actionable side of Decision Frameworks, which covers where these tools come from and whether they hold up. Most decisions need none of this. The skill is spotting the few that do, then reaching for the right tool. This is what I reach for.
Triage first, every time
Before any analysis, sort the decision:
- Which door? One-way (hard to undo) or two-way (cheap to reverse)? A two-way door gets a minute, not a week. Lean toward just trying it.
- Stakes? Will this matter in a year? In ten? Or does it only feel big tonight?
- Is the clock real? Most “urgent” calls have slack. Ask what actually breaks if I decide tomorrow, or next week.
- Is it even mine? If it touches my partner or family logistics, I loop her in before the analysis, not after.
Reversible and low-stakes: decide and move on. Optimizing here is the trap. The rest is for the few that survive.
The walkthrough for the hard ones
Run in order. Skip what doesn’t fit.
- Write the question in one sentence. If I can’t, I don’t understand it yet. Watch for the fake binary (“should I do X?” when the real question is “what do I do about Y?”), and for arguing the tactic when the strategy is the real choice.
- Force a third option. Two options is usually a false choice. Ask what I’d do if neither existed.
- Name the load-bearing assumption. What has to be true for A to win? Which part am I guessing? Can I test the guess cheaply first?
- Premortem. It’s a year out and this went badly. Write how it happened. The failure modes I can name are the ones I can guard against. Putting it in past tense, as if it already failed, surfaces about 30% more of them.
- Check the gut against the analysis. When they disagree, that’s information. One is holding context the other isn’t. Don’t silence either.
- Decide, and set a review date. Write the call, the reason, and when I’ll revisit. The review date is how I learn whether the process works.
Specialty tools
Pull only when the case fits.
- Expected value — for bets I can quantify.
P(win)×value − P(lose)×value. Real when the numbers are real, dangerous when I fake them. - 10/10/10 — for emotionally loud calls. How will I feel in 10 minutes? 10 months? 10 years? It cuts recency bias.
- Regret minimization — for big life calls. Project to 80, ask which choice I’d regret not making.
- Reference class — for any estimate. Don’t guess from inside the project; look at how long similar things actually took. My estimates run optimistic by a steady multiple (the planning fallacy).
- Hell yes or no — for optional commitments. Anything short of a clear yes is a no. It guards against overcommitting, but can also dress up inertia, so I watch it.
Traps I keep hitting
- Analysis as procrastination. More research feels like work. Past a point it’s avoidance. Set a deadline.
- Sunk cost. What I’ve already spent doesn’t matter. Only the next step counts.
- Status quo bias. Not deciding is a decision, with the same consequences. A default no is still a no.
- Optimizing the measurable. Easy to chase salary or hours and miss energy, learning rate, time with family.
Log
YYYY-MM-DD — [decision] — [one-sentence reason] — [review date]
Checking past calls against how they turned out is the only thing that sharpens the process.
See also
- Decision Frameworks — the research behind the triage, the premortem, and reference classes
- Systems Thinking — the one-way/two-way door call is leverage-point reasoning: tuning a parameter versus changing the structure