A $10 million bet deserves a different process than a routine call, and the split starts with one question: can you reverse it? Irreversible, one-way-door decisions warrant slow, structured analysis — a pre-mortem, expected-value math, real debate. Reversible, two-way-door decisions warrant speed, cheap tests, and a bias toward action.

Quick answer: Sort every big decision by reversibility first. One-way doors — expensive or impossible to unwind — earn a pre-mortem, expected-value reasoning, and a slower, more deliberate process. Two-way doors — cheap to reverse — earn fast, low-ceremony experiments. Most directors invert this by default: heavy scrutiny on calls they could undo in a week, and a quick nod for the ones they can't.

Start With the Door: Sort the Decision Before You Analyze It

Sort every big decision by reversibility before you debate its merits: a one-way door is expensive or impossible to undo and deserves slow, structured process; a two-way door is cheap to reverse and deserves speed instead. Get the classification right first — it matters more than the dollar figure attached to the call.

Jeff Bezos described this split in Amazon's 1997 and 2015 shareholder letters as Type 1 (one-way door, hard to reverse) versus Type 2 (two-way door, cheap to reverse) decisions. His observation was that most organizations let heavyweight, committee-style process creep into Type 2 decisions that a small team should be able to make and move on from quickly.

The test isn't "how big is the number." A $10M marketing test you can kill in a quarter is a two-way door. A $2M decision to sunset a product line your sales team has already sold multi-year contracts against is a one-way door. Size and reversibility are correlated, not identical, and treating them as the same thing is where most process problems start.

DimensionTwo-way door (Type 2)One-way door (Type 1)
ReversibilityCheap and fast to undoExpensive, slow, or impossible to undo
Who should decideAn individual or small team, close to the workDirector/VP-level, with real debate
Ideal information levelRoughly 70% of what you'd like to haveAs close to complete as time reasonably allows
ProcessLightweight, fast, default to actionPre-mortem, expected value, documented reasoning
ExampleA pricing experiment, a feature flag rollout, a rubric tweakA platform migration, a market exit, an executive hire

Three tells that a decision is actually one-way, even when it arrives dressed up as routine:

  • Contractual or regulatory lock-in — customer commitments, compliance filings, or vendor terms that can't be unwound without real cost.
  • Reputation exposure — a public launch, a pricing change customers will remember, or a layoff that changes how the market and your best people see the company.
  • Compounding dependency — a technical or org decision other teams will build on top of within a quarter, so the true cost of reversal grows every week you wait.

Get the sorting habit right first. Everything else in this piece — the pre-mortem, the expected-value math, the after-the-fact grading — is process you apply once you've correctly named which kind of door you're standing in front of. Our complete guide to the director-level product role covers where this sorting instinct fits into the broader job; here we go deep on the mechanics of the call itself.

When It's Genuinely Unclear Which Door You're In

Some decisions resist a clean label on first pass — a new pricing tier, say, that's cheap to turn off but expensive to explain away to the sales team that just sold against it. When the door is ambiguous, three tie-breaker questions usually settle it:

  • Who else has to act on this before we could reverse it? The more teams that build downstream commitments, the more one-way the call actually is.
  • Does reversing it require admitting a public mistake? Reputational cost of reversal is a real cost, even when the technical undo is trivial.
  • What's the realistic time-to-reverse, not the theoretical one? A decision that's technically reversible in a sprint but would actually take two quarters to unwind politically is a one-way door wearing a two-way door's badge.

When in doubt, treat the call as one-way. The downside of over-preparing for a decision that turns out to be reversible is a wasted afternoon; the downside of the reverse is a multi-quarter mess.

Match Your Process to the Stakes, Not to the Org Chart

The single most expensive mistake directors make isn't a bad decision — it's applying the wrong amount of process to the right one. Reversible calls get slow-walked through steering committees while genuinely irreversible ones get waved through because the team is already excited. Matching process to stakes, not to habit, is the real skill.

Over-analyzing a two-way door has a cost that never shows up on a post-mortem: every week spent in review is a week a competitor spent shipping and learning instead. Under-analyzing a one-way door is worse, because the cost surfaces much later, after the organization has already built structurally on top of a bad call.

Failure modeWhat it looks likeWhat it costs
Over-analyzing a Type 2 callA six-week review cycle for a reversible pricing testLost learning velocity; competitors iterate faster
Under-analyzing a Type 1 callA platform re-architecture greenlit in one meeting because "engineering is confident"Multi-quarter lock-in to a choice made on incomplete reasoning
Right-sizing (the goal)A reversible test ships this week; an irreversible bet gets a pre-mortem and a written decision memoSpeed where it's cheap, rigor where it's expensive

A useful discipline before scheduling any decision review: ask how many weeks it would take to undo the call if it goes wrong, and what undoing it would cost. If the honest answer is days, cheaply, cancel the steering committee and let the team ship. If the honest answer is quarters, expensively, and touching other teams, the reasoning tools below are worth the time they take.

The mismatch is common enough to show up in research on how executive teams actually spend their time. McKinsey's global surveys on organizational decision-making have repeatedly found that most executives report spending more time in decision meetings than they used to, while only around one in five believe their organization is genuinely good at making decisions quickly and executing on them. More meetings without better sorting doesn't buy better decisions — it just moves the same undifferentiated process onto more calendars.

This is also where org design intersects with decision speed. A lot of "run everything through committee" habits are actually a symptom of unclear team boundaries — nobody is confident which team owns the call, so everyone gets pulled in as a hedge. If that's the root cause on your team, it's worth reading how clear team boundaries change who gets to decide what — fixing the boundary problem often fixes the over-analysis problem for free.

Pressure-Test One-Way Doors With a Pre-Mortem

For decisions correctly sorted as irreversible, the highest-leverage single tool is a pre-mortem: before committing, imagine the decision has already failed spectacularly, and work backward to explain why. Psychologist Gary Klein popularized the technique in a 2007 Harvard Business Review piece, building on research showing that imagining an outcome has already happened — "prospective hindsight" — measurably improves people's ability to identify the real reasons behind it.

The mechanics are simple enough to run in one working session:

  1. Set the scene. State plainly: "It's 18 months from now, and this decision was a failure." No hedging language.
  2. Write independently first. Every participant privately lists reasons for the failure before any discussion — this is what defeats groupthink and deference to the most senior voice in the room.
  3. Pool and cluster. Share out loud, group similar reasons, and flag which ones are both likely and currently unaddressed in the plan.
  4. Assign an owner and a trigger. For each credible failure reason, name who watches for it and what signal would tell you it's happening.
  5. Decide anyway, or don't. A pre-mortem isn't a veto tool — it exists to make sure the bet is taken with eyes open, not to talk the room out of every irreversible move.

Invite people who will disagree with you. A pre-mortem run entirely by supporters of the bet mostly produces reasons that were already dismissed once; a skeptical engineer, a frontline salesperson, or a customer-support lead who sees the edge cases daily will usually surface risks the core team has stopped seeing.

Two inputs make the pre-mortem sharper than a generic brainstorm. First, ground the "why did it fail" reasoning in how customers actually experience the change — mapping the emotional highs and lows across a customer journey surfaces failure points a purely internal discussion tends to miss. Second, if the bet involves cutting or changing a product line, re-check it against the underlying jobs customers are hiring the product to do — a surprising share of one-way-door failures trace back to solving a job nobody actually asked you to touch.

Reason in Expected Value, Not Your Gut's Point Estimate

Argue a one-way-door decision in expected value — probability-weighted outcomes — instead of a single confident guess about what will happen. The math itself is simple; the discipline is forcing the room to state probabilities and outcome sizes out loud instead of arguing past each other in adjectives like "risky" or "promising."

Daniel Kahneman and Amos Tversky's research on prospect theory found that people systematically weigh losses roughly twice as heavily as equivalent gains — a bias that pushes risk-averse directors to kill good expected-value bets, and pushes overconfident ones to ignore the downside tail entirely. Naming the bias doesn't remove it, but it gives the room language for asking whether it's reacting to the loss column harder than the math justifies.

A simplified worked example for a hypothetical $10M platform bet:

ScenarioProbabilityOutcome (impact)Probability × outcome
Bet succeeds as planned40%+$32M value over 3 years+$12.8M
Partial success, needs rework35%+$4M value, delayed roadmap+$1.4M
Bet fails, fully sunk cost25%-$10M plus opportunity cost-$2.5M
Expected value≈ +$11.7M

The number itself matters less than the exercise of building the table. Once probabilities and outcomes are written down, disagreement gets specific: "I think failure is closer to 40%, not 25%" is a productive argument, while "this feels too risky" is not. RICE and Kano scoring — lighter-weight cousins of the same idea — force a gut call into stated, arguable components rather than a single unexamined number.

This is also a hiring lever. Directors who can build and defend a table like the one above under real disagreement are demonstrating exactly the judgment worth screening for in the first place — a theme covered in depth in how to interview for judgment when hiring PMs.

Judge the Call by the Reasoning, Not the Outcome

Judge a decision by the quality of reasoning available at the time it was made, not by how it turned out. Poker player and decision writer Annie Duke calls the opposite habit "resulting" — grading a decision by its outcome, which conflates skill with luck and teaches directors exactly the wrong lesson from both good and bad bets alike.

Resulting produces two specific failure patterns worth naming out loud:

  • A well-reasoned bet that loses to bad luck gets punished, teaching the team to avoid defensible risk in the future even when the process behind it was sound.
  • A poorly-reasoned bet that wins by luck gets celebrated, reinforcing exactly the sloppy process that should have been caught, and making it harder to challenge next time.

The fix requires a habit most teams don't have: write the reasoning down before the outcome is known. A short decision memo — the bet, the probabilities assigned, the pre-mortem's top failure risks, and the information available at the time — turns a future outcome into a graded exam instead of a war story.

Reviewing that memo on a fixed cadence, not only when something goes wrong, is what actually builds the muscle. It's one reason a deliberate operating cadence for directors matters as much for decision quality as it does for team rhythm.

The core idea behind Annie Duke's Thinking in Bets: a decision earns its grade from the reasoning available at the time, not from how things happened to turn out. It's also the single hardest habit to actually install in a leadership team.

Rehearse the Bet Before It's Real

The best time to pressure-test a $10M decision is before you're actually staring at one, under deadline, in a room that's already half-committed. Rehearsing named high-stakes scenarios — a platform migration, a pricing overhaul, a market exit — in a low-stakes setting builds the reflex to sort reversible from irreversible before the adrenaline kicks in.

This is the problem Prodinja's Decision Dojo is designed to address: it walks directors through exactly these named high-stakes scenario types, so the one-way/two-way sorting habit and the pre-mortem discipline are already reflexive before a real bet lands on your desk.

Paired with it, the Decision Journal lets you record the reasoning, probabilities, and pre-mortem risks at the moment you commit. Months later, you can grade the call against what you actually knew at the time — not against how it happened to turn out — closing the loop this whole framework depends on.

Neither tool replaces judgment. Both exist to make the judgment you already have easier to apply consistently, and easier to improve deliberately rather than by accident.

Key Takeaways

  • Sort before you analyze. Classify every big call as reversible (Type 2) or irreversible (Type 1) before deciding how much process it deserves — the classification matters more than the dollar amount.
  • Match rigor to reversibility, not habit. Over-analyzing a cheap-to-reverse decision costs speed; under-analyzing an expensive-to-reverse one costs quarters of lock-in later.
  • Pre-mortem every one-way door. Imagine the decision has already failed, write reasons independently before discussing, and assign an owner and a trigger to each credible risk.
  • Argue in expected value, not adjectives. State probabilities and outcome sizes explicitly; it turns vague disagreement ("this feels risky") into a specific, resolvable one.
  • Watch for loss aversion. People weigh losses roughly twice as heavily as equivalent gains — name the bias when it's driving the room, not the math.
  • Grade the reasoning, not the result. A good decision can lose to bad luck and a bad decision can win by it — write the reasoning down before you know the outcome.
  • Build the rehearsal in before the stakes are high. Rehearsing named high-stakes scenarios and journaling reasoning are what turn disciplined-nerve decision-making into a repeatable skill instead of a one-time act of courage.

Frequently Asked Questions

What is the difference between a Type 1 and Type 2 decision?

A Type 1 (one-way door) decision is expensive or impossible to reverse — a market exit, a platform migration, a senior hire. A Type 2 (two-way door) decision is cheap and fast to undo, like a pricing test or a feature flag rollout. The distinction, popularized by Jeff Bezos, determines how much process a decision deserves.

How do you know if a decision is actually reversible?

Ask what it would cost in time and money to undo the decision six months from now, not today. If contracts, regulatory filings, public commitments, or other teams' work will already be built on top of it by then, treat it as irreversible even if it looked routine at the start.

What is a pre-mortem in decision-making?

A pre-mortem is a technique, popularized by psychologist Gary Klein, where a team imagines a decision has already failed and works backward to list the reasons why, before committing. Writing reasons independently first, before group discussion, is what makes it more effective than a standard risk brainstorm.

Why shouldn't you judge a decision by its outcome?

Judging decisions by outcome alone — what Annie Duke calls "resulting" — confuses luck with skill. A well-reasoned bet can still lose, and a poorly-reasoned one can still win; grading the reasoning available at the time is the only reliable way to improve decision quality going forward.

How much analysis is too much for a big product decision?

If the decision is genuinely reversible at low cost, any review cycle longer than it takes to run a quick test or launch a small experiment is likely too much. Reserve deep analysis — pre-mortems, expected-value tables, multi-week reviews — for decisions that are expensive or impossible to undo.