Picture two managers staring at the same blank approval form. One treats every choice as if it were carved in stone, so the easy ones crawl and the team waits. The other waves everything through at the same speed, and one of those quick "yeses" turns out to be a door that only opens one way. Both are making the same mistake, they're sorting decisions by how important they feel, when the question that actually matters is whether the decision can be undone.

The quick version

  • Two-way doors are reversible, if it goes wrong, you walk back through. Decide these fast, delegate them, and treat them as cheap experiments.
  • One-way doors are irreversible or close to it. These earn the slow, careful, "sleep on it and consult" process.
  • The classic failure is using the slow process on a fast decision, and occasionally the reverse, racing through something you can't take back.
  • Before deciding, ask three things: How reversible is this? How costly is the mistake? How much does waiting cost me?

The idea in depth

The cleanest articulation of this comes from Jeff Bezos. In the Amazon shareholder letter covering fiscal year 2015 (published in April 2016), he split decisions into two kinds. "Some decisions are consequential and irreversible or nearly irreversible, one-way doors, and these decisions must be made methodically, carefully, slowly, with great deliberation and consultation," he wrote. "If you walk through and don't like what you see on the other side, you can't get back to where you were before." Most decisions, he argued, aren't like that: "they are changeable, reversible, they're two-way doors," and they "can and should be made quickly by high judgment individuals or small groups." (You'll often see this framework dated to Amazon's famous 1997 "Day 1" letter, that's a common misattribution; the door language is from the FY2015 letter.)

Bezos's real warning was about organisational drift. "As organizations get larger," he wrote, "there seems to be a tendency to use the heavy-weight Type 1 decision-making process on most decisions, including many Type 2 decisions. The end result of this is slowness, unthoughtful risk aversion, failure to experiment sufficiently, and consequently diminished invention." Default to speed and delegation on anything reversible, then. Save your committees and your sleepless nights for the doors that don't swing both ways.

Why reversibility, not importance, is the right axis

The metaphor sounds folksy, but it rests on a fairly serious idea from economics: real options theory. In Investment under Uncertainty (Princeton, 1994), Avinash Dixit and Robert Pindyck showed that when a commitment is irreversible and the future is uncertain, the option to wait for better information has real, measurable value. Acting locks that option in; waiting keeps it open. The more irreversible and uncertain the choice, the more it's worth paying, in time, in analysis, in a small pilot, to keep your options open before you commit.

The flip side matters just as much. If a decision is genuinely reversible, the option value of waiting is close to zero, there's almost nothing to protect by stalling, because you can correct course cheaply once you learn more. That's the formal reason a two-way door should be quick: deliberation buys you very little, and the cost of delay is real. This is also where reversibility connects to risk vs uncertainty vs ambiguity, the deeper your uncertainty, the more an irreversible commitment should make you nervous, and the more a reversible, learn-as-you-go approach earns its keep.

flowchart TD
    A(["A decision lands on your desk"]) --> B{"Can I walk it back
cheaply if I'm wrong?"}
    B -- "Yes, two-way door" --> C("Decide fast.
Delegate. Treat as an experiment.")
    B -- "No, one-way door" --> D{"Is the cost of waiting
higher than the cost
of a mistake?"}
    D -- "No" --> E("Slow down. Gather more.
Consult. Pilot if you can.")
    D -- "Yes" --> F("Decide now, but make it
as reversible as you can first.")
The triage: reversibility first, then weigh the cost of waiting against the cost of being wrong. Leaders Loop

There's a behavioural reason the framework helps, too. In Thinking in Bets (Portfolio, 2018), former professional poker player Annie Duke names a trap she calls "resulting", judging the quality of a decision by how it happened to turn out, rather than by whether it was sound given what you knew at the time. Reversibility blunts resulting. When most of your decisions are two-way doors, a bad outcome isn't a catastrophe to be re-litigated; it's information you act on by reopening the door. Keep things reversible and you lower the stakes of being wrong, which frees you to decide quickly and learn from the result instead of dreading it.

The limitation: reversibility is a judgement, not a fact

Here's where honesty matters. Whether a door is one-way or two-way is rarely stamped on it, you have to judge it, and people get that judgement wrong in predictable directions. Some choices are reversible on paper but not in practice: you can technically un-launch a product or re-hire for a role, but the reputational damage, the lost trust, or the sunk months don't reverse with it. Switching costs, public commitments, and momentum can quietly weld a two-way door shut. And the framework says nothing about whether a decision is good, only about how much process it deserves. A fast, reversible decision can still be a bad one. Use the door test to set your speed, not to outsource your judgement. The practical guard is to ask not just "can this be undone?" but "undone by when, at what cost, and would anyone actually do it?"

A worked example

Maya runs a 12-person support team. Two decisions hit her inbox in the same week.

Decision one: trial a new layout for the team's internal help-desk dashboard. She's tempted to schedule a review, loop in two other team leads, and build a business case. But it's a two-way door, if the new layout flops, she switches it back in an afternoon. So she runs it as an experiment instead: turns it on for half the team for two weeks, agrees up front what "working" looks like (faster ticket triage, no rise in complaints), and decides she'll keep it or revert based on what she sees. Total deliberation: one short message to the team. (Figures here are illustrative.)

Decision two: migrate the whole team off their current help-desk software onto a new platform on a two-year contract, retraining everyone and re-pointing every workflow. This one she's tempted to wave through because a slick vendor demo made it feel obvious. But it's much closer to a one-way door, the contract, the retraining, and the data migration are expensive to unwind. So she slows down: she runs a 30-day paid pilot with three agents before signing anything, asks the vendor for an exit clause, and consults the people who'll live with it daily. Same manager, same week, opposite processes, because she sorted by reversibility, not by which decision felt bigger.

flowchart LR
    subgraph TWO ["Two-way door · dashboard layout"]
        direction TB
        A1("Switch it on for half the team") --> A2("Watch for 2 weeks") --> A3("Keep it or revert in an afternoon")
    end
    subgraph ONE ["One-way door · platform migration"]
        direction TB
        B1("Run a 30-day paid pilot") --> B2("Negotiate an exit clause") --> B3("Consult, then commit")
    end
Same manager, same week: the reversible choice becomes an experiment; the irreversible one earns a slower, hedged path. Leaders Loop

The sticky version, worth repeating to a colleague: don't ask how big it is, ask whether you can walk it back. And notice the second move inside the second decision, the pilot, the exit clause. A lot of leadership skill is quietly turning one-way doors into two-way ones before you walk through. Closely related is whether you reason from first principles or a fast heuristic: the door test is itself a heuristic, a quick rule that's right often enough to be worth using, as long as you know when to drop it.

Frequently asked questions

Isn't this just "think before you act"?

The opposite, actually. The point is to think less about the majority of decisions, the reversible ones, so you have the time and attention to think harder about the few that are irreversible. The error it targets isn't recklessness; it's over-deliberating on choices that don't deserve it.

How do I tell a one-way door from a two-way door in real time?

Ask three quick questions: If this goes wrong, what does it cost to undo? How long until I'd find out it was wrong? And would I actually reverse it, or would sunk cost and pride keep me walking forward anyway? If undoing is cheap, fast, and realistic, it's a two-way door. If any of those answers is "ugh," treat it as one-way.

Can I make an irreversible decision more reversible?

Often, yes, and it's one of the highest-leverage moves a leader has. Pilots, trials, staged rollouts, exit clauses, and "we'll revisit this in 90 days" all convert a one-way door into something closer to two-way. You pay a little extra up front to keep the option to back out.

Doesn't moving fast on reversible decisions just create chaos?

Only if you skip the cheap discipline of saying, in advance, what would make you reverse. A fast two-way-door decision still has a tripwire, "if complaints rise, we switch back." That's not chaos; it's a controlled experiment with a defined off-ramp. The chaos comes from fast decisions with no agreed signal for when to undo them.

Where does this framework break down?

When reversibility is theoretical rather than real, a decision that could be undone but never will be because of cost, politics, or momentum. It also can't tell you whether a decision is wise, only how much process it warrants. It sets your speed, not your direction.

Related in the Toolkit

Where to go next