One-way and two-way door decisions: sort by how hard they are to undo

By Meet Patel · 2026-10-03 · 6 min read

Take a 40-person company (a hypothetical) where a $300-a-month analytics tool needs sign-off from a team lead, the head of finance and the COO. The subscription can be cancelled next month. The same company signs a three-year office lease after one meeting and a single signature.

The routing is inverted. The cheap, reversible decision collects three approvals, and the expensive, nearly permanent one collects one. A plausible cause is that the approval process was built around the worst decision the company ever made and then applied to everything that followed.

Jeff Bezos described this pattern in his 2015 letter to Amazon shareholders: “As organizations get larger, there seems to be a tendency to use the heavy-weight Type 1 decision-making process on most decisions, including many Type 2 decisions.” He listed the consequences as “slowness, unthoughtful risk aversion, failure to experiment sufficiently, and consequently diminished invention.”

What the two types are

The letter separates decisions into two kinds. Type 1 decisions are “consequential and irreversible or nearly irreversible,” which Bezos calls one-way doors. If you walk through and dislike what you find, you cannot return to where you were. Those decisions, he wrote, must be made “methodically, carefully, slowly, with great deliberation and consultation.”

Type 2 decisions are changeable and reversible. They are two-way doors, and a suboptimal choice does not have to be lived with for long. His instruction is that Type 2 decisions “can and should be made quickly by high judgment individuals or small groups.”

His 2016 letter added a companion rule on speed: “most decisions should probably be made with somewhere around 70% of the information you wish you had.” The two ideas fit together. If a decision can be reversed, waiting for the last 30 percent of information costs more than it protects.

Two-way doors still matter. The amount of process should follow the cost of being wrong, and that cost depends heavily on whether you can undo the choice.

Classify with three questions

A team can sort a decision in about two minutes by asking the same three questions each time.

1. What does undoing it cost in money? Cancelling a monthly subscription costs one month's fee. Exiting a three-year lease costs the remaining rent, or a negotiated penalty.
2. How long does undoing it take, and does it need anyone else? Changing copy on your own website takes minutes and needs nobody. Reversing a hire, a contract or a regulatory filing takes weeks or months and depends on other parties.
3. Does it leave a mark that cannot be removed? Customers who saw a price rise remember it. Deleted data stays deleted. A public promise stays public.

Rate each answer low, medium or high. If all three are low, the decision is a two-way door. If any one is high, treat it as a one-way door. Everything in between needs a third class, which I would add to the Bezos framing: a heavy two-way door, which can be reversed but at a real price. Run these as time-boxed trials with a stated review date, so that the price of reversal is accepted deliberately.

Route each class differently

Classification only helps if it changes who decides and how. This is the routing I would use.

• Two-way door. The owner decides alone, within a spend cap the company sets in advance. The owner tells the team afterward and writes one line in the decision record.
• Heavy two-way door. The owner decides after consulting the one or two people most affected. The decision states the trial length, the measure of success and a review date. One named person is informed before it starts.
• One-way door. The owner writes a one-page memo covering the options, the evidence and the cost of reversal. Named approvers respond in writing within a fixed window. The team runs a pre-mortem before committing.

Here is the 40-person company again, with each decision sorted. The ratings are my illustrative judgments.

• $300-a-month analytics tool, cancellable monthly. Cost to undo: one month's fee. Time: minutes. Mark: none. Two-way door, owner decides.
• New pricing-page copy. Cost to undo: nothing. Time: minutes. Mark: low. Two-way door, owner decides.
• A price increase announced to existing customers. Cost to undo: low in money. Time: weeks. Mark: high, because customers remember it. Heavy two-way door at best.
• Hiring a senior salesperson. Cost to undo: months of pay plus recruiting. Time: months. Mark: affects the whole team. Heavy two-way door, handled with one-way care at the offer stage.
• Three-year office lease. Cost to undo: the remaining rent. Time: years. Mark: a signed obligation. One-way door.

Under the old process, six tool purchases a month at $300 each carry three sign-offs apiece. If each approver spends 10 minutes, that is three hours of approver time, about $150 at a loaded $50 an hour. The money is small, and the larger cost is the waiting, which in this hypothetical averages four days per request. Under the new routing, the worst case across all six purchases is $1,800 for one month before cancellation, and the owner acts the same day. The company has also freed its approvers' attention for the lease.

The same logic sits behind the asymmetric bet framework: when the downside is capped, the right amount of deliberation is small.

Set the cap and handle disagreement

A two-way door still needs a limit, because reversibility has a price. A simple way to set the spend cap is the amount the company would be willing to lose without discussing it, such as a week of the owner's loaded cost or a small fixed sum in the budget. That is the rule of thumb I would use, and the number should come from your own finances. Below the cap, the owner decides and nobody is asked.

Speed also depends on what happens when colleagues disagree. Bezos's 2016 letter offers a phrase for it: “use the phrase 'disagree and commit.' This phrase will save a lot of time.” On a two-way door, a person who disagrees states the objection once, in writing if it matters, and then supports the decision for the length of the trial. The review date is where the objection gets tested against results, which gives the dissenter a fair hearing and gives the owner room to move.

Make the classification visible and contestable

A routing scheme fails quietly when the class lives in someone's head. Put it on the request itself. Every request for a decision carries five short fields: the decision, its class, the cost to undo, the owner and the review date. A request for the analytics tool reads in full: “Adopt the analytics tool. Two-way door. Undo cost: one month's fee, $300. Owner: head of marketing. Review: 30 days.” That is a request a busy approver can read in ten seconds, and it makes the owner commit to a classification that others can challenge.

Disagreements about class will happen, and the useful tie-break is cheap. The person who would carry the cost of reversal has the strongest claim to name the class. If two people still disagree, route the decision as the higher class for one cycle and record who was right afterward. After a few months the record shows which kinds of decisions your team tends to overrate, and the process can be loosened where the fear turned out to be unfounded.

Three mistakes to avoid

Sorting by size. Size and reversibility are separate axes. A $40,000 advertising test that you can stop on any day is a two-way door. A single written promise to a customer, with no price attached, can be a one-way door. Ask what undoing costs, whatever the headline number.

Accepting doors that are one-way by default. Many decisions are irreversible only because nobody built a hinge. A pilot, a break clause, a staged rollout or a feature flag can turn a one-way door into a heavy two-way door. A three-year lease with a break option after twelve months is a different decision from the same lease without one. Before approving a one-way decision, ask whether a cheaper version of the commitment exists.

Treating a two-way door as permission to be careless. Reversibility lowers the cost of a mistake only if someone notices it. Record the decision and a review date. If the same two-way decision gets reversed twice, the company has a problem with how it decides, whatever class the door was given. The decisions a company never makes belong in this review too, because avoiding a two-way door has a cost that accumulates quietly.

A rule to keep

Before routing any decision, ask what it would take to undo it. If the answer is a month's fee and an afternoon, give it to one person and let them move. If the answer is a lease, a launch or someone's livelihood, slow down and write it out. An approval process should never cost more than the mistake it exists to prevent.

Written by Meet Patel — startup operator and growth strategist in Dubai.

Read on themeetpatel.com