Most decisions do not need a framework.

Score six links of a decision and act on the weakest one, because quality is capped by the worst link rather than averaged across them.

Sort the situation into clear, complicated, complex or chaotic, because each one rewards a different response and punishes the other three.

Build a decision in five parts — problem, objectives, alternatives, consequences, trade-offs — so that each can be checked separately.

Break a question into sub-questions that do not overlap and leave nothing out, so that the parts can be answered separately and reassembled.

Restate the problem as its opposite and see whether the new statement is the one you should have been solving.

Follow the consequences of a choice two or three steps out, because the direct effect is rarely the one that decides whether it was right.

Remove every option currently on the table and ask what you would do then, which forces the set to widen in about two minutes.

Four counters to the four common errors: narrow framing, confirmation bias, short-term emotion, and overconfidence about the future.

Restate every option as what you give up by taking it, because the real alternative to a choice is the best thing you do not do.

Score options against a baseline as better, same or worse, then build new options by combining the parts that scored better.

Require at least three live options at every stage, because a set of one is a proposal and a set of two is usually a proposal with a decoy.

Score each option against weighted criteria and sum, then test whether the winner survives a small change in the weights.

Trade a gain on one criterion for a loss on another until criteria become equal across options, then delete them, until one option is left.

Delete any option that another option beats on every criterion, which shrinks the set for free and before any weighting.

Derive weights from pairwise comparisons rather than stating them directly, and check whether your comparisons were internally consistent.

Rank options by how close each is to the best possible profile and how far from the worst, rather than by a weighted sum.

Set a threshold in advance and take the first option that clears it, instead of searching for the best one.

Sort commitments by urgency against importance to decide what gets done, scheduled, delegated or dropped.

Multiply each outcome by its probability and sum, to get the average result of a choice repeated many times.

Draw the sequence of choices and chance events, then compute backwards from the outcomes to find the best first move.

Start from a base rate, then move it by how much more likely the evidence was under one hypothesis than the other.

Predict from what happened to similar cases rather than from the details of this one, because the details systematically mislead.

Decompose an unknown quantity into factors you can bound, so that errors partly cancel and the result lands within an order of magnitude.

Compute what a piece of research is worth by asking how often it would change your decision, and by how much.

Sample each uncertain input many times and record the distribution of results, rather than computing one answer from best guesses.

For each option, find the worst regret it could produce, then choose the option whose worst regret is smallest.

Look only at the worst outcome each option can produce and take the option whose worst outcome is best.

Build three or four internally consistent futures and test each option against all of them, rather than forecasting one future.

Assume the decision has already failed, then write the causes — which surfaces objections that nobody would raise as a prediction.

Ask how you would guarantee the outcome you do not want, then systematically stop doing those things.

Score each failure mode on severity, likelihood and detectability, then work the highest products first.

Assign someone to argue against the decision with real authority and real stakes, so the counter-case gets made by someone who benefits from making it well.

Size a repeated bet as a fraction of what you hold, to maximise long-run growth rather than the return on any single bet.

Compute the probability of losing enough that you cannot continue, and treat any non-trivial answer as disqualifying.

Put most of what you hold in something very safe and a small part in high-variance bets, avoiding the moderately risky middle.

Separate outcomes you recover from and outcomes that end the activity, and apply different decision rules to each.

Classify the decision by whether it can be undone cheaply, and size the process to that answer rather than to how important it feels.

Cycle through observe, orient, decide and act faster than the situation changes, treating orientation as the step that decides the rest.

Project forward to old age and choose the path you would regret least from there, which reweights short-term fear against long-term loss.

Ask how you will feel about this in ten minutes, ten months and ten years, to stop an immediate feeling from settling a long question.

Put a rate on what each week of not deciding costs, so that deliberation is compared against something rather than treated as free.

Pay a small amount now for the right to decide later, treating the delay itself as the thing being purchased.

Split effort between testing unknown options and using the best known one, shifting toward exploitation as the horizon shortens.

Observe a fixed share of the candidates without choosing, then take the first one better than everything seen so far.

Decide in advance what observation triggers a review, and commit to the review before the situation can argue you out of it.

Assign five distinct roles — recommend, agree, perform, input, decide — so that a decision has exactly one decider and no hidden vetoes.

Four roles — driver, approver, contributor, informed — with one driver who owns getting to a decision by a date.

Ask whether anyone has a reasoned objection rather than whether everyone agrees, which converts unanimity from a requirement into an accident.

Collect anonymous estimates, share the distribution without attribution, and repeat — so the group converges on evidence rather than on whoever spoke first.

Have everyone write silently before anyone speaks, then round-robin and rank, so the first speaker does not set the range.

Choose how much to involve others based on the decision’s quality requirement, your information, and whether acceptance is needed for it to work.

State the disagreement on the record, then execute the decision fully, so that dissent does not become passive resistance.

Name the conditions under which group decision-making degrades, so that involving people is a choice rather than a default.

Record the reasoning, the expected outcome and your confidence before the result is known, so that later review compares against what you actually thought.

Score the quality of the reasoning and the quality of the result separately, producing four cases rather than one verdict.

Recognise that outcomes look inevitable afterwards and that judging decisions by results teaches the wrong lesson in any domain with chance in it.

Measure whether your stated confidence matches your hit rate, then train it, because most people’s 90% intervals contain the truth about half the time.

Compare what was meant to happen with what happened, in a session structured so that the comparison can be honest.