Most advice about decisions is really advice about outcomes: how to pick the winner, land the job, buy the right house. But outcomes are only partly yours to control — luck, timing, and other people all get a vote. What you fully control is the process that produced the choice. The single most useful shift in decision-making is to start judging yourself on the quality of the process and let outcomes be what they are.
This guide builds that process step by step: how to separate decision quality from luck, how to size a decision before you invest in it, which reasoning tools to run on the ones that matter, and how to close the loop so your judgment actually improves with experience — which, for reasons covered below, it otherwise will not.
Decision quality is not outcome quality
Start with the foundation. A good decision can produce a bad outcome, and a terrible decision can get lucky. Whenever chance sits between your choice and its result — which is almost always — the outcome alone cannot tell you whether you chose well.
The former professional poker player Annie Duke calls the failure to keep these apart "resulting": judging a decision by its result. Poker makes the error vivid because chance is explicit — you can play a hand perfectly and lose it — but the same structure runs through hiring, investing, and product bets. A team that judges by results alone punishes sound reasoning that got unlucky and promotes reckless reasoning that got lucky, and so, over time, teaches itself to reason recklessly.
The practical rule: evaluate a decision by what was knowable at the time — the information available, the alternatives considered, the reasoning applied. Everything else in this guide is a way of making that process strong enough to defend.
Step 1: Size the decision before you work on it
Not every decision deserves a framework. The first move is triage, and the best tool for it is the distinction Jeff Bezos drew in his letters to Amazon shareholders: one-way doors and two-way doors.
- Two-way doors are reversible. Walk through, and if you dislike what you find, walk back — a trial subscription, a process change, most experiments. These decisions should be made quickly, by default, with about 70% of the information you wish you had. The cost of deciding slowly exceeds the cost of occasionally being wrong.
- One-way doors are hard or impossible to reverse — selling the company, quitting the career, going public with something. These deserve slow, deliberate, multi-model analysis.
The most expensive mistake is not choosing badly; it is applying the wrong speed — agonizing for weeks over reversible choices (burning time and attention that compound elsewhere) while waving irreversible ones through on instinct. Before any significant choice, ask one question first: what would it cost to undo this? The answer sets the process.
Step 2: Generate real alternatives
A decision with one option is not a decision; it is a ratification. Two upgrades:
Compare against the next-best alternative, not against nothing. This is opportunity cost, the economist's model: the true cost of any choice is the best alternative you give up. "Is this hire good?" is a weak question; "is this hire better than the next candidate, or than leaving the seat open and re-recruiting?" is the real one. Options evaluated in isolation almost always look better than they are.
Beware "whether or not" framing. Decisions posed as yes/no ("should we build this feature or not?") smuggle in the assumption that the listed option is the only move. Deliberately generating a third and fourth alternative — build a smaller version, buy instead, wait a quarter — routinely reveals that the original binary was the wrong question.
Step 3: Run the decision through more than one lens
For decisions that clear the triage bar, deliberately apply two or three reasoning tools and see whether they agree. (These are mental models — the full toolkit lives in our guide to mental models; here are the four that earn their place in almost every significant decision.)
Expected value. For each option, sketch the plausible outcomes, how likely each is, and what each would be worth — then weigh them together. The numbers will be rough; that is fine, because the discipline is the point: it forces you to admit that every option is a distribution of outcomes, not the single story you have been telling yourself. A choice with a small chance of catastrophe hiding behind an attractive average is exactly what expected-value thinking exists to expose.
Ruin check. Expected value has a boundary, and honesty requires stating it: it applies to repeatable bets, not to bets you can only survive once. An attractive expected value that includes a real chance of ruin — bankruptcy, health, reputation you cannot rebuild — is not attractive, because ruin ends the game and forfeits every future bet. Check the worst case for survivability before you check the average.
Regret minimization. For long-horizon personal decisions, project yourself to old age and ask which choice you would regret not having made — the framing Bezos has described using when he decided to leave a stable job to start Amazon. It cuts through short-term noise (awkwardness, temporary pay cuts, what colleagues will think) that dominates the present but vanishes at a distance. Its failure mode: imagined future regret is still imagination, and it inherits today's mood — use it as one lens, not a verdict.
The premortem. Before committing, assume the decision has failed and write down why. The exercise, developed by the research psychologist Gary Klein, legitimizes doubts that optimism and group politeness suppress. It is the cheapest insurance available: ten minutes of prospective hindsight, run before the commitment, while changing course is still free.
Where the lenses agree, proceed with confidence. Where they conflict — expected value says yes, the ruin check winces — you have located exactly the part of the decision that deserves more thought.
Step 4: Decide how much certainty the decision is worth
Perfectionism in decision-making is a cost centre. The economist and polymath Herbert Simon coined the term satisficing — searching until you find an option that clears a well-chosen bar, then stopping — and argued that real decision-makers, bounded in time and information, cannot literally optimize. Later work by psychologist Barry Schwartz and colleagues found that habitual maximizers — people who must find the best option — tend to end up objectively better off but less satisfied, paying in search costs, second-guessing, and regret.
The synthesis is a two-tier rule: maximize on the few one-way doors; satisfice on everything else. Define "good enough" in writing before you start looking — the criteria you set in advance are honest; the criteria you improvise mid-search bend toward whatever option is currently charming you. That pre-commitment also blunts the biases that predictably distort this step: sunk costs arguing for the option you have already invested in, anchoring on the first number you saw, loss aversion double-counting the downside. (Those forces have their own field guide — see our companion piece on cognitive biases.)
Step 5: Close the loop with a decision journal
Here is the uncomfortable fact that makes this step non-optional: experience alone does not improve judgment, because memory rewrites itself. Once an outcome is known, hindsight bias makes it feel foreseen, and your recollection of what you believed beforehand quietly shifts toward what happened. Without a written record, you literally cannot audit your own reasoning.
The fix is cheap. For every significant decision, write down at decision time:
- The situation — what you knew, what you didn't.
- The alternatives — including the next-best option you rejected.
- Your prediction — what you expect to happen, with a rough confidence level.
- Your reasoning — which lenses you ran and what each said.
Review the journal when outcomes land. The payoffs compound: you catch systematic errors (always overestimating timelines, always underweighting people problems), you learn your real calibration — how often your "80% sure" is actually right — and you can finally distinguish bad luck from bad process, which is the entire foundation this guide started from.
The framework on one page
- Judge process, not outcomes — resist resulting in both directions.
- Size it — two-way doors fast with ~70% of the information; one-way doors slow.
- Generate real alternatives — compare against the next-best option, never against nothing; escape "whether or not" framing.
- Run multiple lenses — expected value, ruin check, regret minimization, premortem; disagreement between lenses marks where to think harder.
- Match certainty to stakes — satisfice by default, maximize only at one-way doors, set criteria before seeing options.
- Keep the journal — predictions written at decision time are the only feedback loop hindsight cannot corrupt.
FAQ
How do I make better decisions faster?
Triage first: most decisions are two-way doors, and for those, speed is quality — decide with roughly 70% of the information you wish you had, since waiting for more costs more than the occasional reversal. Reserve slow, multi-lens analysis for the few genuinely irreversible choices. Deciding everything at the same speed is the real time sink.
What is the best decision-making framework?
No single framework wins everywhere, and anyone claiming otherwise is selling something. The robust approach is layered: separate decision quality from outcome quality, size the decision by reversibility, then apply two or three lenses — expected value, opportunity cost, a premortem — and pay attention where they disagree. The disagreement, not the framework, is where the insight lives.
What is a decision journal and does it work?
A decision journal is a written record — situation, alternatives, prediction with confidence, reasoning — made at decision time and reviewed when outcomes arrive. It works for a specific, well-documented reason: hindsight bias rewrites your memory of what you believed, so an uncorrupted external record is the only way to audit and improve your own judgment.
How do emotions affect decision-making?
Emotions are information and interference at once: they encode real preferences a spreadsheet misses, and they also amplify biases — fear inflates loss aversion, excitement inflates overconfidence. The practical middle path is procedural: never make one-way-door decisions at emotional peaks, use pre-set criteria so feelings inform the inputs rather than move the goalposts, and let the journal tell you later how your emotional states correlate with your errors. (That is a reasoning practice, not clinical advice.)
Build the toolkit behind the framework
Each tool in this guide — opportunity cost, expected value, satisficing, the premortem, regret minimization — is a full entry on Build Mind, with its precise definition, a worked example, and the failure mode that tells you when to put it down. Browse the Decision-Making models on Build Mind and wire them into your next real decision — the framework only pays when it runs.
