Match the decision process to the cost of being wrong and the practical difficulty of changing course. For a bounded choice you can readily reverse, run a small test, name an owner, and decide in advance what signal would prompt a review or rollback. For a choice with lasting consequences or an expensive reversal, slow down and examine the risks before committing.
What makes an engineering decision reversible?
A decision is reversible when you can change it later without disproportionate cost or lasting harm. Think about more than whether the software has an undo button: reversal may also require restoring data, coordinating a team, notifying customers, or repairing lost trust.
Amazon Web Services describes an A/B test of a site detail-page or mobile-app feature as a basic reversible decision. By contrast, building a fulfillment or data center ties up capital, planning, and resources. AWS summarizes the distinction this way: “A two-way door decision, on the other hand, is one that has limited and reversible consequences: A/B testing a feature on a site detail page or a mobile app is a basic but elegant example of a reversible decision.” AWS Executive Insights, “Elements of Amazon’s Day 1 Culture”.
For an engineering team, classify the decision by its real-world effects, not just its technical implementation. A feature flag can make a deployment easy to turn off, but it cannot necessarily undo data loss, a safety incident, or a customer’s experience during the time the feature was live.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware match#1 Best Overall
Assess consequence and reversibility separately
Before choosing a process, write down what is being decided and who or what will be affected. Then assess the downside of a wrong choice separately from how feasible it would be to reverse. A reversible decision can still carry meaningful risk; a hard-to-reverse decision may be low consequence, though it still deserves a realistic assessment.
- Consequence: What could go wrong, how severe would it be, and who would bear the cost?
- Duration: How long would the effects last, including effects that persist after a rollback?
- Practical reversibility: Can the team reverse the choice technically, financially, operationally, and socially?
- Blast radius: How many systems, customers, teams, or other parties could be affected?
- Learning opportunity: How soon would a meaningful signal arrive, and could a smaller trial preserve options?
These questions are a practical way to apply the reversible-versus-irreversible distinction; they are not a formal checklist prescribed by Amazon or AWS. Their purpose is to expose hidden reversal costs before a team mistakes “we can roll back” for “nothing lasting can go wrong.”
Use a lightweight process for bounded, reversible choices
Jeff Bezos argued against using a single decision process for every choice: “First, never use a one-size-fits-all decision-making process.” In his 2016 Amazon shareholder letter, he says reversible decisions can use a lightweight process, while consequential, nearly irreversible decisions call for more methodical consideration. Amazon, “Jeff Bezos’ 2016 Letter to Amazon Shareholders”.
For a low-consequence choice that is genuinely easy to reverse, a compact process can keep analysis proportional:
Rank #3
- Set the boundary. Specify the change, its scope, and any limits that keep a mistake contained.
- Name an owner. Make clear who can make the call and who is responsible for monitoring it.
- Choose the smallest useful move. Prefer a limited rollout, prototype, or experiment over a broad commitment when it can answer the key question.
- Pick a signal and a review point. Decide what result matters and when the team will evaluate it.
- Define the response. State what would trigger a rollback, adjustment, or decision to continue.
This is a practical implementation for engineering teams, not a claim that Amazon prescribes these exact steps. The point is to make the choice easy to inspect and correct, rather than to remove all analysis.
Slow down when the cost of reversal is high
When a choice could cause serious harm, affect many people, commit substantial resources, or create effects that cannot be undone, spend more time before acting. Bring in relevant expertise, examine failure modes and second-order effects, and make assumptions and dissent visible. A smaller trial can help, but only if it genuinely limits exposure and does not create the very harm the team is trying to avoid.
Rank #4
In the shareholder letter, Bezos contrasts reversible decisions with decisions that are “one-way doors.” The more consequential and difficult a choice is to reverse, the stronger the case for careful review before commitment. That does not mean every such decision requires the same ceremony: the amount of analysis should still reflect the actual downside.
How much information is enough?
Bezos’s 2016 letter offers a rough management heuristic: make many decisions with “somewhere around 70% of the information you wish you had,” rather than waiting for complete information. He also emphasizes recognizing and correcting bad decisions. Amazon’s 2016 shareholder letter.
The 70% figure is not a validated probability of being right, a research-backed threshold, or a universal stopping rule for engineering teams. Use it as a prompt to avoid waiting for certainty when a decision is bounded and correctable—not as permission to skip information needed to understand a serious risk. For a high-consequence, hard-to-reverse choice, the missing information may be precisely what justifies further analysis.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.When should you stop analyzing and choose?
For a reversible decision, stop when you understand the meaningful downside, have a way to contain or detect it, and know what result would change your course. If a small experiment can answer the remaining question faster or more safely than additional debate, run it.
For a difficult-to-reverse decision, stop only after the people who understand the relevant risks have had a chance to surface them, the important assumptions are explicit, and the team has considered whether a smaller or staged commitment could preserve options. More discussion is not automatically better; it should reduce uncertainty that matters to the decision.
Review the outcome and update your judgment
If a reversible choice performs badly, correct it promptly and learn from the result. If a supposedly easy reversal proves costly, identify what made it difficult—such as data migration, dependencies, customer impact, or coordination—and use that information when classifying similar decisions. The goal is not to label every choice perfectly in advance, but to make the process responsive to actual consequences.
Bezos’s letter also discusses correcting bad decisions and the value of “disagree and commit.” That idea can help a team move after a decision has been made, but it does not replace careful review when the stakes are high or the consequences are hard to undo.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
