One-way vs two-way door decisions: how to tell which kind you're making

What is a one-way vs two-way door decision?

A two-way door decision is a decision that is cheap or easy to reverse: you walk through, and if it is the wrong room, you walk back out at low cost. A one-way door decision is expensive or impossible to reverse: once you are through, you are committed, and undoing it costs real time, money, or trust. The framework is most associated with Jeff Bezos' 2016 letter to Amazon shareholders, though the underlying idea, that reversibility should set how much process a decision gets, predates the metaphor.

The point of naming the two types is not the metaphor itself, it is what falls out of it: a two-way door does not need a committee, a deck, or three weeks of deliberation, because being wrong is cheap. A one-way door deserves exactly the deliberation a two-way door does not, because being wrong is not.

How to tell which kind of decision you're making

Classifying a decision correctly comes down to one honest question: what does it actually cost, in time and money, to reverse this if we turn out to be wrong? Not how scary the decision feels, not how senior the room is, not how long the debate has already dragged on. Those three routinely mislead teams into over-processing decisions that are genuinely reversible.

A short checklist that holds up in practice:

• Reversal cost. Can you undo this next week for the cost of an afternoon, or does undoing it mean unwinding contracts, migrated data, or a public commitment?
• Blast radius.Does this affect one team's workflow, or every customer and every team at once?
• Exposure. Can you pilot it with a subset first, or does it have to ship to everyone on day one?

A decision that scores low on all three, cheap to reverse, narrow blast radius, pilotable, is a two-way door no matter how contentious the discussion around it got. A decision that scores high on any one of them, especially reversal cost, is a one-way door even if it feels routine.

How to run a two-way door decision

Move fast and keep the room small. A single owner, or an owner plus one or two people whose input actually changes the answer, is enough; you do not need a full DACI exercise for a decision you can reverse next sprint. Set a short deadline, pick, ship, and set a revisit date rather than a review committee.

The record can be brief. A two-way door decision still deserves a one- or two-sentence entry in a decision log, question, decision, owner, date, but it does not need documented alternatives or a synthesis of dissent. If it turns out wrong, the fix is walking back through the door, not an inquest.

How to run a one-way door decision

Widen the input and slow down deliberately. This is where a heavier role framework earns its keep: the RAPID framework, which separates who recommends from who ultimately decides, suits exactly this kind of high-stakes, hard-to-undo call. If there are several genuinely viable options, a decision matrix forces the competing criteria into the open before anyone argues for a favorite.

The record needs to be complete, not brief: the options that were considered and explicitly rejected (so they do not get re-proposed as new ideas in six months), who stood where, and the reasoning in enough detail that someone who was not in the room can understand why. This is what a decision record is for, and it is worth the extra ten minutes on a decision you cannot cheaply undo.

The mistake most teams make with this framework

Teams routinely classify by how a decision feels rather than what it costs to reverse. A decision that feels politically loaded, even one as reversible as which internal tool a five-person team pilots for a month, gets a deck and a committee. Meanwhile a decision that feels routine, a pricing change, a policy update, a vendor contract with a two-year lock-in, ships fast because nobody stopped to ask whether it could actually be undone. Reversal cost is a fact about the decision, not a feeling about the meeting.

The second mistake is treating “it's a two-way door, move fast” as a reason to skip the record entirely. Speed and documentation are not in tension: a fast decision with no record is exactly how decision debtaccumulates, quietly, one cheap-to-reverse call at a time, until nobody remembers which of last quarter's ten small calls this new problem traces back to. The door type changes how much deliberation a decision earns. It does not change whether the decision gets written down.