All lessonsBusiness & Leadership

The Two-Way Door: Deciding at the Speed the Decision Deserves

A landing page colour got six people and two rounds of feedback. A market exit got forty minutes and was reversed six weeks later. The effort was inversely related to the stakes because nobody had classified the door. Here is the reversibility test, the seventy percent rule, and the protocol that matches process to consequence.

2026-10-20 2,503 words 13 min read

Two Decisions, One Process

In the same week, the founder held two meetings that looked identical from the outside.

The first was about the colour of the new landing page. Six people attended. Two rounds of feedback were collected, a design agency was asked for a third option, and the decision was deferred to the following Tuesday.

The second was about exiting a market that had consumed nine months and produced no revenue. One person attended. The decision was made in forty minutes, announced the next morning, and reversed six weeks later at significant cost.

The effort was inversely related to the stakes. The reversible decision got a committee. The irreversible decision got a founder in a hurry.

Jeff Bezos described the distinction in his 2015 letter to shareholders. Some decisions are one-way doors: once made, they are difficult or impossible to reverse. Others are two-way doors: you can walk back through, reopen the option and try again. The error is applying a heavy, slow, one-way-door process to a decision that could have been tested cheaply (Bezos, 2015, 2015 Letter to Shareholders, Amazon).

The reverse error is worse. Applying a light, fast process to a one-way door produces the kind of decision that cannot be repaired.

The Reversibility Test

Before choosing a process, classify the decision. Four questions are enough.

What does it cost to undo? Not the cost of making the decision. The cost of reversing it. A landing page colour costs an afternoon. A market exit costs the relationships, the contracts and the nine months of learning that cannot be recovered.

How fast does the feedback arrive? A pricing change produces signal in two weeks. A senior hire produces signal in six months. A culture decision produces signal over years. The longer the feedback loop, the more the decision belongs in the slow category, because learning from the mistake will be delayed.

What is the blast radius? Who is affected if this is wrong? A decision that touches one team is different from a decision that touches customers, contracts or the company's reputation.

Does it preserve option value? A decision that keeps several paths open is worth making quickly, because the learning from trying it is worth more than the delay. A decision that closes paths deserves the slower process.

The four questions usually sort decisions into a clear majority of two-way doors and a small number of one-way doors. Most founders treat every decision as a one-way door and reserve no extra care for the ones that actually are.

Write the four answers down. A classification that exists only in the founder's head cannot be challenged by the team, and an unchallenged classification is a preference wearing a method.

Why Founders Over-Process

The slow default is not irrational. It is produced by several well-documented biases and one structural pressure.

Loss aversion. Kahneman and Tversky showed that losses weigh more heavily than equivalent gains, which makes the downside of a wrong decision feel larger than the upside of a right one (Kahneman & Tversky, 1979, Econometrica). The founder over-weights the possibility of being wrong and under-weights the cost of delay.

Accountability. A visible decision creates a record. A deferred decision does not. The founder who delays avoids a documented error and pays an invisible opportunity cost. The asymmetry favours delay even when delay is the more expensive choice.

Identity. Founders are often rewarded for being thorough, and thoroughness becomes part of the self-image. The identity makes a fast decision feel reckless, even when the decision is cheap to reverse.

Sunk cost. Arkes and Blumer showed that people continue investing in a failing course because of prior investment rather than expected return (Arkes & Blumer, 1985, Organizational Behavior and Human Decision Processes). The nine-month market exit was delayed by the nine months, not by the evidence.

Group process. A committee is a machine for adding perspectives and for diffusing responsibility. Both are useful for one-way doors and expensive for two-way doors.

None of these pressures is a character flaw. They are predictable, and predictable biases can be countered with a process that names them.

What Fast Decisions Require

Speed is not the same as haste. Kathleen Eisenhardt studied decision-making in high-velocity environments and found that fast decision-makers used a distinctive set of practices rather than deciding sooner without changing how they decided (Eisenhardt, 1989, Academy of Management Journal).

They tracked real-time operating information rather than relying on historical reports. They developed multiple alternatives simultaneously, which reduced the chance of falling in love with the first option. They consulted experienced advisers whose judgement they trusted, and they used a process of consensus with qualification, where the leader made the call after the team had been heard.

The practices explain why fast decisions can be better decisions. Multiple alternatives reduce anchoring. Real-time information reduces lag. Qualified advice improves the odds without requiring unanimity.

Herbert Simon's work on bounded rationality had already described the alternative to exhaustive optimisation. People satisfice: they choose an option that meets a defined threshold rather than searching for the best possible option (Simon, 1955, Quarterly Journal of Economics). Gigerenzer and Goldstein extended the argument, showing that simple heuristics can perform well in uncertain environments (Gigerenzer & Goldstein, 1996, Psychological Review).

The practical rule is to define the threshold before the search. What would a good enough decision look like? Once the threshold is met, decide. The search after the threshold is usually a search for permission rather than information.

The Seventy Percent Rule

Bezos described a related discipline in the same letter. Most decisions can be made with about seventy percent of the information you wish you had. Waiting for ninety percent means waiting too long, and the cost of the delay usually exceeds the value of the extra certainty.

The rule is not a licence to decide blind. It is a discipline for distinguishing information that would change the decision from information that would only make the decision feel safer.

Two questions separate the two. Would the missing information change the choice, or only the confidence? And can the decision be structured so that the information arrives after the commitment, through a test or a staged rollout?

The second question is the more useful one. A two-way door can be walked through in stages. Launch to a small segment. Hire on a trial project. Run the pricing change in one market. Each stage produces the information that the meeting was waiting for, at a fraction of the cost of waiting.

David Gilbert and Jane Ebert found that irreversible decisions were associated with more satisfaction than reversible ones in some contexts, because the reversible decision keeps the alternatives alive and invites continued comparison (Gilbert & Ebert, 2002, Journal of Personality and Social Psychology). The finding is a useful caution against over-engineering reversibility. The aim is to be able to reverse the decision, not to keep reversing it.

The Decision Protocol

Use this as a standing rule for the leadership team.

  1. Classify. State whether the decision is a one-way door or a two-way door, and why. Write the classification down so the team can challenge it.
  2. Set the process to the class. Two-way doors get a single owner, a deadline and a test. One-way doors get the full process: alternatives, advisers, a pre-mortem and a decision record.
  3. Name the threshold. Define what good enough looks like before the search begins. The threshold converts an open-ended discussion into a decision.
  4. Set a decision date. A deadline that does not depend on the arrival of more information. The date is the commitment that prevents drift.
  5. Name the tripwire. State the signal that would trigger a reversal, and the date the signal will be checked. A reversible decision without a tripwire is a decision that will be reviewed only when it becomes a problem.

The protocol takes ten minutes and it changes the texture of a leadership meeting. The team stops treating every item as equally consequential, and the founder stops reserving the same process for a logo and a market exit.