From Jeff Bezos's 2015 letter to Amazon shareholders (published 2016).
Sorts a decision by whether it can be undone at reasonable cost, then gives reversible calls a fast, delegated process and irreversible ones a slow, senior one.
Eight decisions from an executive backlog, sorted by what it would take to undo them
Select a decision, then the box for how costly and how slow it would be to reverse: the box lights and says what process it gets. Show the sort to compare yours. Illustrative.
- Not placedPlacedTwo-way doorOne-way doorSlow back: reshape itCostly back: reshape it
- Not placedPlacedTwo-way doorOne-way doorSlow back: reshape itCostly back: reshape it
- Not placedPlacedTwo-way doorOne-way doorSlow back: reshape itCostly back: reshape it
- Not placedPlacedTwo-way doorOne-way doorSlow back: reshape itCostly back: reshape it
- Not placedPlacedTwo-way doorOne-way doorSlow back: reshape itCostly back: reshape it
- Not placedPlacedTwo-way doorOne-way doorSlow back: reshape itCostly back: reshape it
- Not placedPlacedTwo-way doorOne-way doorSlow back: reshape itCostly back: reshape it
- Not placedPlacedTwo-way doorOne-way doorSlow back: reshape itCostly back: reshape it
Time to undo
Cost to undo
Your sort matched 0 of 8.Your sort matched 1 of 8.Your sort matched 2 of 8.Your sort matched 3 of 8.Your sort matched 4 of 8.Your sort matched 5 of 8.Your sort matched 6 of 8.Your sort matched 7 of 8.Your sort matched 8 of 8.
- Two-way door
- Cheap and quick to reverse: decide fast, close to the work, with one owner, a review date, and the undo plan written down.
- One-way door
- Costly and slow to reverse: decide slowly, with senior people, a written memo, and time for objections.
- Slow back: reshape it
- Cheap to reverse but slow: shorten the way back, with a pilot or a trial period, then decide it by what is left.
- Costly back: reshape it
- Quick to reverse but costly: cap the cost, with a test group or a limited release, then decide it by what is left.
The sort
- Price test on one plan: Two-way door. A week to switch back and a few lost sign-ups at worst. The product lead decides.
- Trial of an expense tool: Two-way door. A thirty-day trial with no contract. Finance runs it and reports back.
- Moving the all-hands: Two-way door. If Thursday fails, it moves back. Nobody needs a meeting to decide a meeting.
- Hiring a finance chief: One-way door. Months to find, months to unwind, and a wrong choice touches every number. Full team, written case.
- Leaving a country market: One-way door. Customers, staff, and licenses do not come back on request.
- Ten-year office lease: One-way door. Breaking it costs years of rent. This one deserves the memo and the week of objections.
- Public price cut: Costly back: reshape it. Reversible in a day, but raising a price customers saw cut costs trust. Test it on new sign-ups first.
- Merging support teams: Slow back: reshape it. Cheap to announce, months to settle, and as long to unwind. Pilot the new structure on one shift.
An example
At a three-hundred-person software company, every decision of any size went to the Tuesday executive meeting, and the backlog ran six weeks. A price test on one plan had waited a month; a ten-year office lease had been approved in twenty minutes because the meeting was running late.
The chief operating officer asked two questions of every item on the backlog: what would it cost to undo, and how long would undoing take? Most of the list was cheap and quick to reverse. Those went back to the people closest to the work, each with a named owner and a date to review it. Three were costly and slow to reverse, the lease among them, and those got a written memo, a week for objections, and the full executive team. Two sat in between and were reshaped first: the public price cut became a test on new sign-ups only. The backlog cleared in two weeks.
Why it works
Many organizations have one decision process and apply it to everything. If the process is slow and careful, reversible choices queue behind it, and the cost of waiting (lost learning, frustrated teams) usually exceeds the cost of an occasional wrong call that can be walked back. If it is fast, the few irreversible choices get the same quick treatment as the trivial ones.
Sorting by reversibility matches the effort to the stakes. A two-way door is cheap to get wrong, so speed and delegation win, and trying teaches more than analysis would. A one-way door is expensive to get wrong, so the slow care pays. The test also exposes false one-way doors: many choices that feel permanent can be made reversible with a pilot or a shorter contract.
The thinking behind it
Jeff Bezos Jeff Bezos, the founder of Amazon, set out the distinction in his 2015 letter to shareholders, filed with the SEC in April 2016. Type 1 decisions are consequential and close to irreversible, one-way doors that deserve slow and careful work; Type 2 decisions are reversible two-way doors that small groups should make fast. The test is whether a decision can be undone, and each kind gets its own process. The letter is a practitioner's account of how he ran a company, not research.
How to run it
- Name the decision and its current path. Write who would decide it today, and how long it has waited.
- Price the undo. Estimate what reversing would cost in money, customers, trust, and people, and how long it would take.
- Call the door. Cheap and quick to undo is two-way; costly and slow is one-way; a mixed case goes to the next step.
- Reshape the mixed cases. A pilot, a shorter contract, or a limited release often turns a one-way door into a two-way one.
- Push two-way doors down. Give them to the person closest to the work, with a review date and the undo plan written down.
- Slow the one-way doors. Write the reasoning, set a period for objections, and bring in the people who will live with the result.
- Check the calls each quarter. See which doors were misjudged, and tighten the questions.
Where it goes wrong
Every door gets called one-way. Owners who fear blame label their decisions irreversible to get senior cover, and the backlog returns. Ask for the specific cost of undoing; a vague answer usually means a two-way door.
The undo plan never gets written. A two-way door is reversible only if someone knows when and how to reverse it. Without a review date and a trigger, a quick call hardens into a permanent one.
One-way effects hide inside two-way decisions. A choice can be reversible on paper and still leave a mark, such as a price customers saw cut or a team told it would be disbanded. Count trust and reputation in the cost of undoing.
When to use something else
- Real Options Analysis. When a one-way door can be split into stages, and the right to wait is worth pricing.
- Premortem. When a one-way door is chosen and the team should hunt for how it fails before walking through.
- Decision Tree Analysis. When an irreversible choice has uncertain outcomes worth weighing with probabilities and payoffs.
Where it sits in SOLVE eX
Go to the source
- 2015 Letter to Shareholders, Amazon.com, filed with the SEC as exhibit 99.1 to a Form 8-K, 2016