Where does the one-way door framing come from?

Jeff Bezos described it in a letter to Amazon shareholders, using the terms Type 1 for irreversible decisions and Type 2 for reversible ones. The argument was that large organizations tend to apply heavyweight Type 1 process to Type 2 decisions and become slow as a result.

Can a decision be partly one-way and partly two-way?

Most are. The productive approach is to split the decision into its reversible and irreversible components and apply different process to each, rather than classifying the whole thing by its most permanent part.

What is the most common classification mistake?

Treating reversible decisions as irreversible, which shows up as approval chains and committee review for choices that could be tested in a week. The opposite error is rarer but more expensive, and usually happens when the cost of reversal falls on a team other than the one deciding.

One-way vs two-way door decisions: the framework that fixes slow companies

Comparison · founder · 4 min read · last verified 2026-07-21

Reviewed before publication Editorial board Independent commercial review

In shortA one-way door decision is costly or impossible to reverse; a two-way door decision can be undone cheaply. Most organizational slowness comes from applying one-way rigor to two-way decisions.

A one-way door decision is one that is expensive or impossible to reverse, while a two-way door decision can be undone cheaply if it turns out to be wrong. Jeff Bezos described this distinction in a letter to Amazon shareholders, labeling irreversible choices Type 1 and reversible ones Type 2, and warning that applying heavyweight process to reversible decisions makes organizations slow.

One-way vs two-way door decisions at a glance

One-way door: reversal is costly, slow, or impossible. Examples include multi-year contractual commitments, public pricing structure changes with grandfathered terms, and data model migrations that other systems depend on.
Two-way door: reversal is cheap and fast. Examples include page copy, campaign targeting, meeting structures, and most internal process changes.
Correct process for one-way doors: more evidence, more dissent, named decision owner, written reasoning.
Correct process for two-way doors: a fast decision by whoever is closest to the work, with a review date.
The common error: treating two-way doors as one-way, which produces committee review of reversible choices.
The rarer, costlier error: treating one-way doors as two-way, usually because the reversal cost sits outside the team making the call.

What a one-way door decision is

A one-way door decision creates commitments that others will build on. The test is not how important the decision feels but what it would cost to unwind after the fact, including costs paid by people outside the deciding team. A pricing model change is reversible on the website and irreversible in the contracts already signed under it. A database choice is reversible in month one and effectively permanent once a dozen services and a reporting pipeline assume its behavior.

Three properties tend to indicate a one-way door:

External commitment: customers, partners, or regulators can now hold the company to it.
Compounding dependence: other decisions will be made on top of it before the error is visible.
Delayed feedback: the signal that it was wrong arrives long after the point where reversing was cheap.

Decisions with these properties earn slow process: written proposals, explicit alternatives, someone assigned to argue against, and a record of the reasoning so that a later reversal can be evaluated on what was known at the time.

What a two-way door decision is

A two-way door decision can be tried and withdrawn within a short window at low cost. The right process is to name an owner, set a date to look at the result, and let them decide. The value of speed here is not just the time saved on the decision itself; it is the information generated by making it. Reversible decisions are experiments, and an organization that debates them instead of running them accumulates opinion rather than evidence.

The main hazard is misclassification by convenience. Teams sometimes label a decision two-way because they want to move quickly, then discover that customers were told about it. A useful check is to ask who would need to be informed of the reversal. If the answer includes anyone outside the team, the door is narrower than it looked.

How they relate

Most decisions are not cleanly one type. They are two-way doors with a one-way component, and the useful move is to separate the components rather than to classify the whole decision. Launching a new pricing tier is reversible; guaranteeing that tier's price for three years is not. Testing a new market segment is reversible; hiring a dedicated team for it is much less so, which is why hiring decisions deserve one-way process even though headcount is nominally adjustable.

The same separation applies to strategy. Choosing to compete in a segment is often reversible early, but the positioning commitments that follow are not. Deciding to define a new category, covered in category creation, commits the company to years of market education that competitors can harvest if it withdraws. Deciding how to attack an established segment, covered in entering a crowded market, usually leaves more room to change approach.

Which to use when

Classify before deciding, not after. The classification determines who decides and how much evidence is required, so doing it later means the process was chosen by default.
Ask what reversal costs, not what the decision costs. Expensive decisions with cheap reversals are two-way doors.
Ask who pays for the reversal. If the cost lands on customers, support, or a downstream team, treat it as one-way even if it is easy for the deciding team to undo.
Push two-way doors down, to the person closest to the work, with a review date rather than an approval chain.
Push one-way doors up, with written reasoning, named alternatives, and an assigned dissenter.
Re-examine classifications as the company grows. Decisions that were reversible with ten customers become one-way with a thousand, because the number of parties who built on them has changed.

The practical payoff is speed without recklessness. An organization that classifies well makes most decisions quickly, because most decisions genuinely are reversible, and reserves its slow, expensive deliberation for the small number that will still be constraining the company years later. Applying uniform rigor to everything is not caution; it spends the same scarce review capacity on choices that would have corrected themselves in a week.