One-way door decisions: meaning and examples

Cerno is a private AI boardroom that runs consequential decisions through a structured process and returns an Executive Decision Brief. Judging reversibility is one of the first things it does — here's the idea on its own, the doors that look two-way and aren't, and how to find a smaller one.

What does "one-way door" mean?

Jeff Bezos framed it in an Amazon shareholder letter. Some decisions are two-way doors: reversible, low-cost to undo, so the right move is to decide quickly, try it, and reverse if needed. Others are one-way doors: consequential and irreversible, deserving of genuine deliberation.

The failure mode he warned about is applying one-way-door caution to two-way-door decisions — treating everything as permanent, drowning reversible calls in process, and moving at a crawl. The opposite error is just as costly and, in a small business, more common: rushing a one-way door because it felt like a two-way one.

Why reversibility is the first question

Every decision has two costs: the cost of being wrong, and the cost of deciding. Deliberation is expensive — it takes your time, delays the outcome, and often doesn't change the answer. The only way to know how much of it a decision deserves is to know what being wrong would cost, and that is mostly a question of whether you can undo it.

Get this right and two good things happen. Reversible decisions get made in minutes, and you find out fast whether they worked. Irreversible ones get the attention that the reversible ones were stealing. Get it wrong in either direction and you are either paralysed or exposed.

Examples of one-way and two-way doors

The same kind of decision can sit on either side, depending on how it is set up. These are typical for a small business:

• ›Premises: a three-year lease with no break clause is one-way; a rolling monthly desk or a short licence is two-way.
• ›People: a permanent hire or a redundancy is one-way; a short contract or a freelancer for one project is two-way.
• ›Ownership: taking on a partner or selling equity is one-way; a joint venture on a single job is two-way.
• ›Pricing: a public price rise to every customer is one-way; a new price for new customers only, reviewed in a quarter, is two-way.
• ›Brand: changing the business name is one-way; testing a new tagline on one landing page is two-way.
• ›Customers: ending your biggest client is one-way; pausing new work from a difficult one is two-way.
• ›Tools: migrating every record to a new system is one-way; trialling the new system alongside the old one is two-way.

The two-way version is usually the "smaller door" from step 5 below: the same decision, staged so you can walk back through it.

When should you use it?

Use it as the first question about any decision: can I reverse this, and at what cost? The answer sets how much time, deliberation and evidence the decision deserves. It's the triage step before any deeper analysis — before a pre-mortem, before tracing second-order effects, before anything.

How to classify a decision in 5 steps

1. ›Ask: can this be undone? Genuinely — not "in theory". A decision that can be reversed only by an expensive, embarrassing climbdown is not reversible in any useful sense.
2. ›Cost the reversal. Time, money, reputation, relationships, data, and other people's lives. A door that costs a fortune to walk back through is closer to one-way.
3. ›Classify. Two-way (cheap to reverse), one-way (costly or impossible), or somewhere between — and most are somewhere between.
4. ›Match the process to the door. Two-way → decide fast, write down what you expect, set a review date, move on. One-way → deliberate properly; get it challenged; get the reasoning on the record.
5. ›Look for a smaller door. Often you can turn a one-way door into a two-way one — a pilot, a trial, a break clause, a staged commitment, a rolling contract instead of a fixed one.

Worked example

A pricing change looks permanent, but staging turns it into a reversible test:

A second example: the lease

A small business has outgrown its space. Two options are on the table: a three-year lease on a bigger unit at a good rate, or a rolling month-to-month arrangement at a serviced space that costs about a fifth more.

The three-year lease is a one-way door. If trade drops, if the location turns out wrong, if the business changes shape, you are still paying for three years — or paying to get out. The monthly arrangement is a two-way door: it costs more per month and you can walk out of it with a month's notice.

That doesn't make the monthly option right. It means the two options deserve different amounts of thinking. The monthly one can be decided this week and reviewed in six months. The three-year one deserves a pre-mortem, a look at what the base rate says about businesses that expand into fixed space, and an honest answer to "what would we do if turnover fell by a quarter?"

And there is usually a smaller door. A three-year lease with a break clause at eighteen months is a very different commitment from one without. The extra rent on the monthly space, for a year, may be cheaper than being wrong about the three-year one. Reversibility can be bought; the question is whether the price is worth paying.

Doors that look two-way and aren't

The dangerous decisions are the ones that feel reversible and aren't. Some regular offenders:

"We'll just try it" hires. You can end a probation period, but the person has left a job, told their family, and turned down other offers. Reversing it costs them a great deal and costs you your reputation as an employer. Closer to one-way than it feels.

Free trials and introductory offers. Customers who take a trial start relying on it. Withdrawing something people have built into how they work is a one-way door dressed as a two-way one.

Anything public. A price, a name, a position on something — once it is out, reversing it is a story. "We tried it and changed our minds" costs credibility even when the change is right.

Anything that involves other people's money or time. A supplier who tools up for you, a contractor who clears their diary, a customer who plans around a promised date. You can reverse your side; you cannot reverse theirs.

Data and access. Deleting, migrating, or granting access. The technical reversal is often impossible and always slower than the decision was.

The test in each case: who else has to reverse something for this to be undone, and would they?

Common mistakes

Costing reversal only in money. The expensive reversals are usually reputational and relational. A decision that costs nothing to reverse financially but requires telling a customer you were wrong is not free to reverse.

Forgetting the option value of waiting. Some one-way doors don't have to be walked through today. If waiting a month costs little and might bring information, the decision is not "yes or no" but "yes, no, or not yet".

Process theatre on two-way doors. Holding a meeting, writing a document and consulting three people about a decision you could reverse in a day is not caution; it is cost. The tell: nobody would have decided differently without the process.

Deciding two-way doors fast and then forgetting them. Fast is only safe if you find out whether it worked. A reversible decision with no review date is a slow decision in disguise — you'll only notice it was wrong when it's stopped being cheap to reverse.

Treating "we can always change it later" as a plan. You can, but will you? Reversal has a cost in attention and admission, and most reversible mistakes are never reversed. Cheap to undo is not the same as likely to be undone.

Which Cerno advisor uses this?