How to Make Better Decisions: Strategies, Biases, and Frameworks That Actually Work

Here is a quietly damning pair of numbers from McKinsey's global survey on decision making: just over half of respondents reported spending more than 30% of their working time on decision making — and only 20% said their organizations excel at it. Read those together: the single largest consumption of managerial time, performed at a self-assessed excellence rate of one in five.

Now the encouraging part. When something that important is done that badly by people that capable, the cause is almost never the people. It is the absence of a designed process — decisions made the way they've always been made: in meetings shaped by habit, framed by whoever spoke first, decided by whoever spoke loudest, and forgotten by the time their consequences arrive.

This guide is the process. Five moves, in order, each one backed by its own deep-dive on this site — because this post's job is to give you the complete map, not to re-fight every battle on it. Run your next consequential decision through the five and you will feel the difference before the outcome ever lands.

The Premise: Better Process Beats Smarter People

Daniel Kahneman's Thinking, Fast and Slow gave the premise its vocabulary: System 1 — fast, intuitive, pattern-matching — makes most of our judgments, brilliantly and invisibly, and carries systematic biases with it; System 2 — slow, effortful, analytical — audits System 1 only when something forces it to. Left undesigned, group decisions run on System 1 plus social dynamics: anchoring on the first option, confirmation toward the early favorite, deference to seniority.

You cannot debias a brain by wanting to. What you can do is design the process so that System 2 gets forced into the loop exactly where it pays — which is what each of the five moves does at a different stage of the decision. None of them requires anyone to be wiser; all of them change what the room does by default.

Move 1: Sort by Reversibility

The highest-leverage move costs thirty seconds. In his 2015 shareholder letter, Jeff Bezos distinguished Type 1 decisions — consequential, irreversible or nearly so, "one-way doors" — from Type 2 decisions — changeable, reversible, "two-way doors." His observation: as organizations grow, they run the heavyweight Type 1 process on everything, buying slowness and risk-aversion for no protection in return. The following year's letter added the speed rule for the reversible kind: most decisions should be made with around 70% of the information you wish you had — waiting for 90% usually means being slow, and being wrong on a Type 2 is cheap because you can course-correct.

So the first question about any decision is not "what should we choose?" but "which type is this?" Two-way doors get delegated, decided fast at 70% information, and revisited if wrong. One-way doors get the full remaining four moves. Most teams discover they've been doing the exact opposite: agonizing over reversible tool choices while waving through irreversible commitments. We unpack the full Bezos decision system — including disagree-and-commit — in its own post. (For a full situational framework on exactly this — when to decide alone, consult, or hand the call to the group — see the Vroom-Yetton decision model.)

Move 2: Frame Before Options

Most bad decisions are doomed at the question. The framing discipline — the one Cassie Kozyrkov taught twenty thousand Googlers — is to write down, before any analysis: the actual decision being faced, the default action if nothing is decided, and what evidence would change your mind. If no realistic finding would alter the action, you are not deciding; you are decorating a foregone conclusion.

Framing's second half is criteria before candidates. Agree the three-to-five things that would make an option win before reviewing options — because once candidates are on the table, every criterion proposed afterward is suspected of being reverse-engineered for somebody's favorite. This single habit is the strongest known antidote to decision paralysis: the research on choice overload shows it strikes precisely when preferences are undefined and options are complex — both of which framing fixes in advance.

Move 3: Structure the Deliberation

An unstructured discussion is not a neutral container — it has physics. The first option anchors; the highest-paid opinion gravitates; doubters self-censor; silence reads as agreement. The groupthink countermeasures exist because exhortation ("speak up!") does not survive contact with those forces; structure does: the leader states their view last, dissent is collected written-first or anonymously, and someone is honestly free to attack the plan.

The strongest structural form is the explicit argument: every option carries its pro and con arguments, stated as claims with reasons, weighed on their merits rather than their authors. This is where collaborative decision-making earns its research pedigree — structured groups beat both individuals and unstructured groups — and it is the core of what this site exists to say: the quality of a decision is the quality of the arguments that survived it. It is also, not incidentally, how you earn stakeholder buy-in: people commit to decisions whose reasoning they can see — and helped build.

Move 4: Decide by Protocol

Deliberation has to end, and how it ends should be decided before it starts. The lightweight standard is Gokul Rajaram's SPADE (developed at Square; the canonical write-up is on First Round Review): Setting — the what/when/why of the decision; People — who consults, who approves, and crucially one accountable decider; Alternatives — the real options with their trade-offs; Decide — the call, made explicitly; Explain — the reasoning communicated to everyone affected. Its cousin frameworks (RAPID, DACI) make the same core point: ambiguity about who decides and by what rule is where deliberation goes to die. Choose the rule deliberately, too — consent, consensus or single-decider each fit different decisions, a choice we compare in depth.

And before committing a one-way door: run Gary Klein's pre-mortem. The group assumes the decision failed and independently writes why. Fifteen minutes, and the doubts that politeness was suppressing arrive as assignments instead of confessions — the cheapest error-catching step ever designed, which is why it also anchors the groupthink toolkit.

Move 5: Record It, Then Review It

The four moves above produce a good decision. The fifth makes decisions compound: a half-page record — the question, options, arguments, decision, rationale, owner, review date — written when the decision lands, in one searchable place. The full practice, template included, is in how to document decisions, with its two boundary companions on minutes vs. the decision log and action items vs. decisions.

The review date is the move's engine: when it arrives, the team compares the outcome against the recorded reasoning and learns whether the process — not the luck — was good. That loop, run across many decisions, is the practical core of decision intelligence: decisions as organizational assets that improve the system that produces them. One scheduling note from the team-sequencing playbook: whatever else your agenda does, the consequential decision goes first, never in an overrunning meeting's final five minutes.

"Frameworks Are Procrastination — Judgment and Speed Win"

The strongest objection comes from the best practitioners: decision frameworks can become decision theater — SPADE documents for lunch orders, pre-mortems for reversible calls, process as a substitute for courage. Speed is a real competitive weapon; Bezos's letters are one long argument that large organizations die of decision heaviness, not decision haste. And seasoned judgment really does compress the five moves into instinct — expert intuition is pattern recognition, not recklessness.

The answer is that the five moves, correctly read, are a speed system. Move 1 exists to fast-track the majority of decisions — Type 2 calls should get less process than they do today, decided at 70% information by whoever owns them. Framing prevents week-long analyses of the wrong question. Named deciders and explicit rules end the slowest pattern in corporate life: the decision nobody is empowered to make. The heavyweight machinery — structured deliberation, pre-mortem, full records — is reserved for one-way doors, where "slow is smooth and smooth is fast" is not a cliché but arithmetic: re-litigating a botched irreversible call costs more than every framework ever will.

And the concession: if your team applies all five moves to everything, you have missed Move 1, which was the point. Process proportional to reversal cost — that is the entire philosophy in five words.

How Argumentree Runs the Five Moves

The five moves describe a process; Argumentree is that process as software. The framing is explicit — a question with its context and criteria. The deliberation is structured by construction — options with pro and con arguments in a rated tree, dissent invited as a branch rather than an act of courage. The decision lands under whatever rule the team chose, with the reasoning visible.

And Move 5 is free: the deliberation is the record — question, options, arguments, rationale, owner and review date preserved without a transcription step. The five-move process stops being a discipline someone must remember and becomes the shape of how the team already works.

One in Five Is a Choice

The McKinsey number is worth returning to: organizations spend a third of their time deciding and rate themselves excellent at it one time in five. Nothing in that gap requires better people. Every part of it yields to designed process — sorting, framing, structuring, protocol, memory — installed one move at a time, starting with the thirty-second one.

So pick the decision currently drifting through your calendar, and run it through the five. Type it. Frame it. Argue it in the open. Decide it by name and rule. Record it with a date. Your organization is already full of smart people — give them a decision process somebody designed.