One-Way vs Two-Way Door Decisions: Bezos's Decision Framework

By Tara Minh · Updated Oct 12, 2026

A one-way door decision is hard or impossible to undo, so it deserves slow, careful analysis and senior input. A two-way door decision can be reversed at modest cost, so a capable person or small group should make it quickly. Jeff Bezos called these Type 1 and Type 2 decisions. The framework's real job is matching process to reversibility.

Most slow organizations aren't slow because their leaders are indecisive. They're slow because every decision gets the same treatment: the same meetings, the same sign-offs, the same pre-reads. Picking a vendor for a pilot gets the process built for an acquisition. The door framework gives you a fast way to tell those two apart before the meeting invites go out.

What Are One-Way and Two-Way Door Decisions?

Key Facts

• Bezos introduced the Type 1 / Type 2 split in Amazon's 2015 letter to shareholders, filed with the SEC as an exhibit.
• In that letter he named the failure mode: a tendency to use "the heavy-weight Type 1 decision-making process on most decisions", which he said ends in slowness, unthoughtful risk aversion, and too little experimentation.
• The 2016 letter restates the idea as "two-way doors" that "can use a light-weight process" and adds the 70% rule: most decisions should be made with roughly 70% of the information you wish you had.
• The same letter says that if you wait for 90% of the information, you're probably being slow.
• Both letters are Bezos's own statements of a management practice. Neither is a controlled study, so treat the framework as a judgment tool, not a measured result.

The metaphor is simple. A one-way door locks behind you. A two-way door lets you step through, look around, and walk back if you don't like what you see.

In the 2015 letter, Bezos describes the first kind as decisions that are consequential and irreversible or nearly irreversible, and says those must be made methodically, carefully, and with consultation. He labels them Type 1. The second kind is everything else: choices that are changeable and reversible. Of those he writes that Type 2 decisions "can and should be made quickly by high judgment individuals or small groups."

Two things are easy to miss. First, the framework is about the process, not the topic. A pricing change can be either door depending on whether you can roll it back next quarter. Second, Bezos's warning is aimed at the opposite error from the one most people expect. He isn't mainly worried about rushing Type 1 calls. He's worried that growing companies pour Type 1 process over everything.

That connects to how Amazon's founder runs the rest of his playbook, which we cover in Jeff Bezos's leadership style.

Why Reversibility Is the Right Question

Most decision tools ask how big, how risky, or how important. Reversibility asks something sharper: what does being wrong cost, and how long does that cost last?

A decision can be large and still reversible. Launching a regional pilot of a new service tier costs real money, but you can end it. A decision can be small and still locked in. Signing a three-year contract with a termination penalty, or announcing a layoff, or publicly committing to a date, is hard to walk back no matter how little money changes hands.

Bezos makes the cost argument in the 2016 letter: if you're good at course correcting, being wrong may be less costly than you think, whereas being slow is going to be expensive for sure. That's the trade at the heart of the framework. Slowness has a price that shows up every week. A wrong two-way decision has a price that shows up once, and you can cap it.

One-Way Doors vs Two-Way Doors at a Glance

Notice the information bar row. The 70% rule belongs to the second column, not the first. Using it on a one-way door is a misreading, and it's the most common one.

How This Compares With Other Decision Frameworks

Several well-known frameworks answer neighboring questions. The door test doesn't replace them. It tells you which one to reach for.

A useful rule of thumb: doors tell you how heavy the process should be, Vroom-Yetton tells you how wide the circle of input should be, and risk-based analysis tells you how deep the homework goes. For the dialogue side of the same idea, the OODA loop is the natural companion for two-way doors, because the loop assumes you'll act, observe the result, and adjust.

How to Classify a Decision in Five Questions

You don't need a scoring model. Five questions, answered in a couple of minutes, sort most decisions.

1. Can we undo it? If the answer is "yes, in a week, for little money," it's probably a two-way door. If it's "not really," treat it as one-way.
2. What does reversal cost? Count money, time, customer trust, and morale. A reversal that burns public credibility is expensive even when the spreadsheet cost is small.
3. Who else gets locked in? A decision that traps customers, partners, or employees inherits their commitments too. Contracts, public promises, and roadmap dates all travel.
4. How fast does the window close? Some two-way doors have a cost for waiting. If a competitor will take the slot while you deliberate, speed is part of the price.
5. Can we shrink it? Often you can turn a one-way door into a two-way door by piloting, phasing, or adding an exit clause. Ask this before spending weeks on analysis.

That last question is the most useful one. Many "irreversible" decisions are only irreversible because nobody tried to design the exit.

Step-by-Step: Choose the Process After You Pick the Door

Once you've sorted the decision, match the process to it.

For a two-way door

1. Name a single owner. One person decides. Ambiguity about who decides is the main thing that makes reversible choices slow.
2. Write the decision in two or three sentences. What are we choosing, and what would make us reverse it?
3. Set a review date. A two-way door is only two-way if somebody looks back. Put a date on it.
4. Decide at about 70%. Bezos's 70% rule is a permission to move before the picture is complete. Don't hold out for more data you can collect after you've acted.
5. Log it and move on. A one-line entry is enough (template below).

For a one-way door

1. Slow down on purpose. Say out loud that this is a Type 1 call so the room understands why the pace changes.
2. Gather broad input. This is where wide consultation belongs, and where a model like Vroom-Yetton helps you decide how much participation to invite.
3. Run the downside. Use risk-based decision-making: list what you don't know, the worst realistic outcome, and the early warning signs.
4. Invite dissent in writing. People often agree in the room and disagree in the hallway. Written objections are cheaper to read than a failed launch.
5. Decide, then commit. If the group stays split, the leader makes the call and the team commits to it (more on that below).
6. Record the reasoning. One-way decisions are the ones people want to revisit. A short record of what you knew and why you chose stops the same debate from restarting every quarter.

When you can't tell

Treat it as a two-way door and add a cheap safeguard: a trial period, a spending cap, or a rollback condition. If you can't add any safeguard, you've just learned something about the door.

Disagree and Commit: Closing the Loop

The 2016 letter pairs the door idea with a way to end debates. Bezos writes that if you have conviction on a direction even without consensus, it's helpful to say, look, I know we disagree on this, but will you gamble with me on it? Disagree and commit? He adds that he does it himself: "I disagree and commit all the time."

The mechanism matters. It works because everyone has had their say before it's invoked. Used before the debate, it's just a way to skip the debate. It also fits two-way doors especially well, since a committed team can test the choice quickly, and a reversible choice makes it easier for dissenters to say "fine, let's see."

Decision Log Template: Door Classification

Copy this into a shared doc or a decisions channel. One row per decision that someone asks about twice.

And a short checklist to fill before a meeting:

Keep the "Why this door" column honest. It's the one that catches teams quietly labeling a risky call as reversible to move faster.

Examples by Business Function

These are illustrative scenarios, not case studies.

Sales. Changing the discovery-call script is a two-way door: one owner, a two-week test, revert if conversion drops. Signing an exclusive reseller agreement that blocks other channels for three years is a one-way door. It needs legal review, finance input, and a written case.

Marketing. Testing a new ad channel with a capped budget is a two-way door. Rebranding the company name, domain, and visual identity is one-way in practice: customers, search rankings, and printed materials all carry the old name.

Product and engineering. A feature behind a flag is a two-way door. A public API contract that outside developers build on is a one-way door, because breaking it breaks other people's products. A data-model migration with no rollback script sits in the same category.

HR and people. Trialing a new 1:1 format for a quarter is a two-way door. Changing compensation structure, announcing a reorganization, promoting someone out of a 9-box talent review into a senior role, or ending a role are one-way doors, because trust, once lost, doesn't roll back with a policy revision. A RACI matrix helps here by making it clear who owns each part of the decision.

Finance and operations. Switching a low-risk software tool on a monthly plan is a two-way door. A multi-year capital commitment, a major supplier consolidation, or an acquisition is a one-way door.

Customer support. Trying a new macro or triage rule is a two-way door. Changing the refund policy company-wide is closer to one-way, since customers will quote it back to you.

For founders, where one person tends to sit on both kinds of door at once, founder decision-making at scale covers how to hand off the reversible ones. And the speed of decisions is its own subject, with more on how teams measure it.

Do's and Don'ts

Common Mistakes

• Mislabeling a one-way door as a two-way door. The speed feels good until the cost lands. Test the label by asking who's locked in.
• Treating everything as one-way. This is the failure Bezos warns about. If your approval chain is long for a $2,000 trial, the process is outsized for the risk.
• Skipping the review. A reversible choice nobody revisits is effectively permanent. Two-way doors need a date.
• Confusing reversible with free. A reversible decision can still cost money, time, and credibility. The point is that the cost is bounded.
• Over-trusting the 70% number. It's Bezos's rule of thumb from a shareholder letter, not a threshold from research. It's a nudge to stop waiting, not a formula.

The easiest way to start is small. Take the next three decisions waiting on your desk and write "one-way" or "two-way" next to each. For the two-way ones, name an owner and a review date today. For the one-way ones, spend the saved time on the part that deserves it: the downside, the dissent, and the written reasoning.