Contents
Use this to determine how much analysis a decision actually deserves. The Reversible vs. Irreversible framework sorts every decision into one of two categories — two-way doors you can walk back through, and one-way doors you cannot — then matches the decision process to the stakes. Most organisations over-deliberate on reversible choices and under-deliberate on irreversible ones. This tool fixes both errors simultaneously.
Section 1
What This Tool Does
Organisations don't fail because they make bad decisions. They fail because they apply the wrong decision-making process to the decision in front of them. A team spends six weeks building a business case, running financial models, and convening a steering committee — to choose a project management tool they can switch away from in a month. Meanwhile, the same organisation approves a ten-year exclusive distribution agreement after a single executive meeting because the partner "felt right" and the deal had momentum. The first decision got too much process. The second got almost none. Both errors are expensive, but the second one can be fatal.
Jeff Bezos articulated this asymmetry in his 1997 letter to Amazon shareholders and refined it in his 2015 letter, where he gave the framework its most memorable formulation. Type 1 decisions are one-way doors: once you walk through, you can't easily get back. Type 2 decisions are two-way doors: if you don't like what you find on the other side, you can walk back through. The language is deliberately simple. A child can understand it. That's the point — the framework needs to be fast enough to apply in real time, in any meeting, at any level of the organisation.
The cognitive gap it fills is specific and well-documented. Humans are poor at calibrating deliberation to stakes. Kahneman and Tversky's work on loss aversion explains part of it: we overweight the downside of any choice, which makes even trivially reversible decisions feel consequential. Organisational dynamics explain the rest. Process accumulates. Approval chains lengthen. What started as appropriate rigour for bet-the-company decisions metastasises into the default process for every decision, regardless of magnitude. Bezos called this "using a heavy-weight Type 1 decision-making process on Type 2 decisions" — and identified it as one of the primary causes of large-company slowness.
The mechanism is a triage step, not an analytical framework. You're not building a model or scoring criteria. You're asking one question — can we reverse this if we're wrong? — and letting the answer determine how much process the decision gets. Reversible decisions should be made quickly, by individuals or small teams, with minimal approval overhead. Irreversible decisions deserve the full apparatus: data gathering, scenario planning, devil's advocacy, senior review. The tool doesn't tell you what to decide. It tells you how much to invest in deciding.
What makes this deceptively hard is that most decisions aren't cleanly one type or the other. They sit on a spectrum. And the classification itself can be wrong — a decision that looks reversible may have hidden switching costs, reputational consequences, or second-order effects that make reversal far more expensive than it appears. The framework's power comes from forcing the classification question. Its danger comes from answering it too quickly.
Section 2
How to Use It — Step by Step
Instructions on the left. Worked example — "Should our B2B SaaS company switch from annual contracts to month-to-month pricing?" — on the right.
Step 1 — State
Articulate the decision as a specific, concrete action
Vague decisions can't be classified. "Should we change our pricing?" is unanswerable. "Should we move our enterprise tier from annual commitments to month-to-month billing starting Q2?" is classifiable. The more specific the action, the more accurately you can assess its reversibility. Write it down. One sentence. If you can't state the decision in one sentence, you're bundling multiple decisions together — unbundle them and classify each one separately.
Worked example
B2B SaaS pricing shift
"Move all new enterprise customers from mandatory 12-month contracts to month-to-month billing, effective April 1, while honouring existing annual contracts through their renewal dates." Clear action. Clear scope. Clear timing. Now it can be classified.
Step 2 — Classify
Ask: if this decision is wrong, what does reversal actually cost?
This is the core move. Don't ask "is this reversible?" in the abstract — that question is too easy to answer with a glib "sure, we can always change it back." Instead, enumerate the specific costs of reversal: financial cost, time cost, reputational cost, relationship cost, opportunity cost, and organisational cost (confusion, morale, credibility). If the total reversal cost is low relative to the value at stake, it's a two-way door. If reversal is impossible, extremely expensive, or would damage trust and credibility in ways that can't be repaired, it's a one-way door. Most decisions land somewhere in between — assign them a position on the spectrum rather than forcing a binary.
Worked example
Assessing reversal cost
Financial: Switching back to annual contracts means lost upfront cash — month-to-month customers who would have paid annually now pay monthly, reducing cash reserves by an estimated $2.1M over six months. Reversible, but expensive. Reputational: Customers who chose month-to-month will feel bait-and-switched if forced back to annual. Sales team credibility takes a hit. Partially irreversible. Competitive: Competitors will use the reversal as evidence of instability. Hard to undo. Organisational: Finance team has already modelled new cash flow projections; reversal means re-planning. Annoying but manageable. Classification: closer to a one-way door than it first appears. The pricing change itself is technically reversible, but the second-order effects — customer trust, competitive positioning, cash flow disruption — make reversal costly enough to warrant serious analysis.
Step 3 — Calibrate
Match the decision process to the classification
For clear two-way doors: delegate to the person or team closest to the information. Set a time limit — 48 hours, one meeting, whatever fits. Bias toward action. The cost of delay almost certainly exceeds the cost of a wrong decision you can reverse. For clear one-way doors: slow down. Gather data. Run scenarios. Assign a devil's advocate. Require sign-off from someone with the authority and context to own the outcome. For decisions in the middle of the spectrum: identify what would make the decision more reversible. Can you run a pilot? Test with a subset? Structure the commitment with an exit clause? Often the best move isn't to classify and decide — it's to restructure the decision so it becomes more reversible.
Worked example
Calibrating the pricing decision
This decision sits at roughly 70% irreversible. The team's response: don't make it fully reversible (that would mean not doing it at all), but reduce the irreversibility by restructuring the decision. Instead of switching all new enterprise customers to month-to-month, run a 90-day pilot with mid-market accounts only (lower ACV, shorter sales cycles, less reputational risk). Measure churn, cash impact, and customer satisfaction. Set a kill criterion: if monthly churn exceeds 8% or cash collections drop below $X, revert before expanding. The pilot converts a one-way door into a two-way door with a modest cost of passage.
Step 4 — Communicate
Make the classification explicit to the team
State the classification out loud. "This is a two-way door — I'm making the call and we'll adjust if it's wrong." Or: "This is a one-way door — we need another week of analysis before we commit." The classification itself is a leadership signal. It gives the team permission to move fast on reversible decisions without feeling reckless, and permission to slow down on irreversible ones without feeling indecisive. Teams that internalise this language start self-classifying their own decisions, which is the real payoff — the framework scales through vocabulary, not process.
Worked example
Communicating the pricing pilot
The VP of Product sends a one-paragraph Slack message: "We've classified the month-to-month pricing shift as a one-way door for the full rollout, but we've restructured it into a two-way door pilot. Mid-market segment, 90 days, clear kill criteria. The pilot decision is mine to make — no steering committee needed. The full rollout decision, if we get there, goes to the exec team with pilot data. Questions welcome, but we're starting April 1." Thirty seconds to read. Everyone knows the stakes, the process, and who owns what.
Section 3
When It Works Best
✓
Ideal Conditions for Reversible vs. Irreversible Classification
| Dimension | Best fit |
|---|---|
| Organisational context | Scaling companies where process is accumulating faster than judgment. The framework is most valuable between 50 and 5,000 employees — large enough that decision-making has started to slow, small enough that the vocabulary can spread organically. At Amazon, Bezos introduced it precisely when the company was growing fast enough that "Day 2" bureaucratic creep was becoming a real threat. |
| Decision volume | High-throughput environments where dozens of decisions compete for attention every week. The framework's primary value is triage — it prevents the most common resource misallocation in management: spending equal time on unequal decisions. Product teams, operations leaders, and founders making 20+ decisions per week benefit most. |
| Decision type | Operational and strategic decisions with identifiable reversal costs. Pricing changes, vendor selections, hiring plans, feature launches, market entries, partnership agreements — any decision where you can concretely enumerate what happens if you need to undo it. Less useful for interpersonal or cultural decisions where "reversal" isn't a meaningful concept. |
| Cultural problem | Organisations suffering from either excessive caution (analysis paralysis on routine decisions) or excessive speed (insufficient rigour on consequential ones). The framework diagnoses both pathologies simultaneously. If every decision goes through the same approval chain, you have a Type 1/Type 2 confusion problem. |
| Information environment | Situations where you have enough information to assess reversal costs but not enough to predict the optimal outcome with certainty. If you already know the right answer, you don't need a decision framework. If you can't even estimate reversal costs, you need a different tool — probably the Cynefin Framework to classify the problem domain first. |
| Leadership maturity | Teams where leaders are willing to delegate two-way door decisions downward and accept imperfect outcomes. The framework only works if the organisation actually changes its behaviour based on the classification. If every decision still requires the CEO's approval regardless of type, the vocabulary is decoration. |
Section 4
When It Breaks Down
⚠
Failure Modes
| Failure pattern | What goes wrong | What to use instead |
|---|---|---|
| Misclassification through optimism | Teams classify decisions as two-way doors because they want to move fast, not because reversal is actually cheap. "We can always change it back" becomes a mantra that ignores switching costs, sunk costs, reputational damage, and team fatigue from constant pivots. The classification becomes a permission structure for recklessness rather than a genuine assessment of reversibility. | Pre-Mortem to stress-test the "what if we need to reverse this?" scenario before committing |
| Hidden irreversibility | Some decisions look reversible at the surface but have irreversible second-order effects. Laying off 15% of your engineering team is technically reversible — you can rehire. But the institutional knowledge that walked out the door, the signal it sent to remaining employees, and the six months of reduced velocity are not reversible. The framework doesn't automatically surface these hidden costs; you have to look for them. | Second-Order Thinking to map downstream consequences before classifying |
| Binary thinking on a spectrum | Teams force every decision into "Type 1" or "Type 2" when most decisions sit somewhere in between. The binary framing is useful as a mental shortcut but dangerous as a rigid classification system. A decision that's 60% irreversible gets treated identically to one that's 95% irreversible, leading to either too much or too little process. | Use a 1–10 reversibility scale; calibrate process intensity to the score rather than a binary label |
| Weaponised speed | Leaders use "it's a two-way door" to shut down legitimate concerns and bypass necessary deliberation. The framework becomes a rhetorical tool for impatience rather than a genuine triage mechanism. Junior team members learn that raising concerns about a decision classified as Type 2 is career-limiting, so they stop raising concerns altogether. | Require that the person classifying the decision also states the specific reversal plan — not just "we can change it back" but how, at what cost, and by when |
| Cumulative reversible decisions | Each individual decision is reversible, but the aggregate of fifty small two-way door decisions creates a path dependency that is effectively irreversible. You chose this tech stack, this pricing model, this market segment, this brand voice — each one a two-way door. Together, they define a strategic position that would take years to unwind. The framework evaluates decisions in isolation; strategy happens in accumulation. | Periodically audit the portfolio of recent Type 2 decisions for emergent strategic commitments; use Scenario Planning to test whether the accumulated direction is sound |
| Emotional irreversibility | Some decisions are technically reversible but psychologically irreversible. Announcing a pivot to the entire company, publicly committing to a strategy at a conference, telling a co-founder you want to part ways — you can technically reverse all of these, but the emotional and relational costs make reversal practically impossible. The framework's rational cost-benefit lens misses the human dimension. | Hard Choice Model for decisions where identity, relationships, and values are the real stakes |
The most dangerous failure mode is misclassification through optimism, because it's self-reinforcing. A team that habitually classifies decisions as two-way doors moves fast, which feels good, which confirms the belief that speed is the right approach. The feedback loop only breaks when a supposedly reversible decision turns out to be irreversible — and by then the damage is done. The protection is simple but requires discipline: for any decision classified as a two-way door, state the reversal plan in concrete terms. Not "we can change it back" but "here is exactly what we would do, it would take X weeks, cost $Y, and require Z." If you can't articulate the reversal plan, you haven't actually established that the door is two-way. You've just assumed it.
Section 5
Visual Explanation
Reversible vs. Irreversible decision triage — applied to the B2B SaaS pricing worked example. Decisions are classified on a reversibility spectrum, then matched to an appropriate process intensity.
Section 6
Pairs With
The Reversible vs. Irreversible framework is a triage tool — it determines how much process a decision deserves. It says nothing about how to actually make the decision once you've calibrated the process. That's where the pairings come in.
Use before
Reframing
Before classifying a decision's reversibility, make sure you're classifying the right decision. Reframing challenges the problem definition itself. "Should we switch to month-to-month pricing?" might be the wrong question — the real decision might be "how do we reduce churn among enterprise customers?" which opens entirely different options with different reversibility profiles.
Use after
Pre-Mortem
For decisions classified as one-way doors, run a pre-mortem before committing. Imagine the decision has failed catastrophically — what went wrong? This surfaces hidden risks and hidden irreversibilities that the initial classification might have missed. It's the stress test for your "this is a one-way door" assessment.
Use after
Second-Order Thinking
The most common classification error is ignoring second-order effects. A decision looks reversible at the first order (you can change the price back) but irreversible at the second order (customers who left during the price change don't come back). Second-Order Thinking maps these downstream consequences before you commit.
Use after
Confidence Determines [Speed](/mental-models/speed) vs [Quality](/mental-models/quality)
Once you've classified the decision type, this companion framework helps calibrate execution. High confidence + two-way door = move immediately. Low confidence + one-way door = invest in reducing uncertainty before committing. The two frameworks together create a 2×2 that covers most decision situations.
Mental model
Regret Minimisation Framework
Bezos's other famous decision tool. Where Reversible vs. Irreversible classifies the decision's structure, Regret Minimisation addresses its emotional weight. For one-way doors where the data is ambiguous, ask: "At 80, will I regret not having tried this?" The two frameworks complement each other — one is structural, the other is existential.
Mental model
[OODA Loop](/mental-models/ooda-loop)
Boyd's Observe-Orient-Decide-Act cycle shares the same core insight: speed of decision-making is a competitive advantage, but only when matched to the situation. OODA provides the tempo; Reversible vs. Irreversible provides the filter for when to accelerate the loop and when to slow it down.
Section 7
Real-World Application
Amazon — the Fire Phone and the one-way door that wasn't treated like one
The scenario
In 2014, Amazon launched the Fire Phone — a smartphone with a 3D display, a feature called "Dynamic Perspective" that tracked the user's head movements, and deep integration with Amazon's shopping ecosystem. The device was a spectacular failure. Amazon took a $170 million write-down on unsold inventory in Q3 2014 alone. The phone was discontinued within a year. It remains one of the most expensive product failures in Amazon's history, and it's instructive precisely because it illustrates what happens when an organisation that preaches the Reversible vs. Irreversible framework fails to apply it to itself.
How the tool should have appliedBuilding a smartphone from scratch was, by any reasonable assessment, a one-way door. The investment was enormous — reportedly over $100 million in R&D across multiple years. The competitive landscape was dominated by Apple and Samsung with deeply entrenched ecosystems. The reputational stakes were high: Amazon was entering a consumer hardware category where it had no track record. And the decision had compounding irreversibilities: every month of development deepened the commitment, trained engineers on phone-specific skills rather than other projects, and narrowed the window for alternative uses of those resources. The reversal cost wasn't just the write-down — it was the years of engineering talent and leadership attention that could have been directed elsewhere.
What it surfacedBrad Stone's reporting in The Everything Store and subsequent coverage revealed that the Fire Phone was driven heavily by Bezos's personal conviction. The project reportedly had limited internal dissent — or at least, dissent that didn't alter the trajectory. The irony is sharp: the person who articulated the Type 1/Type 2 framework more clearly than anyone in business history pushed a Type 1 decision through with a process more appropriate for a Type 2 decision. The 3D display feature — Dynamic Perspective — consumed enormous engineering resources to solve a problem customers didn't have, and internal scepticism about its value reportedly didn't change the product direction.
The non-obvious factorThe deeper lesson isn't that Amazon made a bad bet. Companies that take big swings will sometimes miss. The lesson is about what Amazon did after the failure — and how the framework's real value shows up in the recovery. Amazon didn't abandon hardware. It redirected the Fire Phone team's work toward Echo and Alexa, which launched later that same year and became one of the most successful consumer hardware platforms of the decade. The hardware engineering, the supply chain relationships, the voice recognition research — much of it transferred. Amazon treated the Fire Phone's failure as information, not as a reason to retreat from the category. The framework's deepest application isn't just "classify before you commit." It's "when a one-way door leads somewhere you didn't want to go, salvage everything you can from the journey." The Echo team reportedly benefited directly from lessons and infrastructure built for the phone. A $170 million write-down became a down payment on a multi-billion-dollar product line.
Section 8
Analyst's Take
Faster Than Normal — Editorial View
The Reversible vs. Irreversible framework endures because it solves the right problem at the right altitude. It doesn't try to improve the quality of any individual decision. It improves the allocation of decision-making resources across all decisions — which, for any organisation making hundreds of choices per quarter, is the higher-leverage intervention. The framework's genius is that it's a meta-decision tool: a decision about how to decide. That recursive quality is why it scales. A CEO can use it. A product manager can use it. An intern can use it. The vocabulary is the technology, and vocabulary spreads faster than process.
The failure mode I see most often isn't misclassification — it's classification theatre. Teams adopt the language without changing behaviour. Every decision gets labelled Type 1 or Type 2 in the meeting, and then the same six-person approval chain processes it regardless. The label becomes a ritual, not a routing mechanism. The tell: if your organisation has adopted the framework but decision velocity hasn't measurably increased on Type 2 decisions, you have a vocabulary problem, not a framework problem. The fix is structural, not rhetorical. Identify the ten most common Type 2 decisions your team makes, pre-authorise them to specific individuals, and remove the approval requirement entirely. Make the classification consequential or don't bother making it.
The highest-leverage modification I've encountered: add a third question after "can we reverse this?" — namely, "can we restructure this decision to make it more reversible?" This is the move the worked example illustrates, and it's the move that separates operators who use the framework mechanically from those who use it creatively. Most one-way doors can be converted into two-way doors through pilots, phased rollouts, contractual exit clauses, or scope reductions. The classification isn't fixed — it's a design variable. The best decision-makers don't just sort decisions into buckets. They reshape the decisions themselves until the reversibility profile matches their risk tolerance. That's the advanced move, and it's worth more than the basic framework by an order of magnitude.
Section 9
Top Resources
01
Primary source
The definitive articulation of the Type 1/Type 2 framework in Bezos's own words. The letter is short, direct, and contains the one-way door / two-way door metaphor that has since entered the management lexicon. Read the original before reading anyone else's interpretation of it. The surrounding context — Bezos's argument about "Day 1" culture and the dangers of proxy decision-making — is as valuable as the framework itself.
02
Book
The most thorough account of Amazon's decision-making culture, including the internal dynamics that led to both brilliant bets (AWS, Prime) and expensive misses (Fire Phone). Stone's reporting shows how the Reversible vs. Irreversible framework operates — and sometimes fails to operate — inside the organisation that invented it. Essential for understanding the gap between the framework as articulated and the framework as practiced.
03
Book
Written by two former Amazon VPs who spent a combined 27 years at the company. The book details Amazon's decision-making processes from the inside, including how Type 1 and Type 2 classifications actually route through the organisation. Chapter coverage of the "single-threaded leader" model shows how Amazon structurally delegates Type 2 decisions — the operational implementation that makes the framework real rather than rhetorical.
04
Book
The scientific foundation for why the framework is necessary. Kahneman's work on loss aversion explains why humans systematically over-deliberate on low-stakes decisions (the potential loss looms larger than the potential gain) and why status quo bias makes irreversible commitments feel safer than they are. Read Part IV on prospect theory to understand the cognitive architecture the Reversible vs. Irreversible framework is designed to override.
05
Book
Grove's account of Intel's decision to exit the memory chip business and bet entirely on microprocessors — one of the most consequential one-way door decisions in technology history. The book is a masterclass in recognising strategic inflection points where irreversible commitment is the only viable path. Complements the Bezos framework by showing what disciplined one-way door decision-making looks like when the data is ambiguous but the stakes are existential.
Decision Tools Library — Browse by phase
FramingHard Choice ModelCynefin FrameworkReversibility TestReframingAbstraction LadderingSWOT Analysis
Why this matters next
mental modelsSecond-Order Thinking
Reframing applied the Second-Order Thinking mental model
mental modelsLeverage
Reframing applied the Leverage mental model
mental modelsCompounding
Reframing applied the Compounding mental model
mental modelsMomentum
Reframing applied the Momentum mental model
mental modelsOODA Loop
Reframing applied the OODA Loop mental model
mental modelsProxy
Reframing applied the Proxy mental model
Continue exploring
HMCFREALSA
Decision tool
Hard Choice Model
Determine what kind of decision you're facing (no-brainer, apples vs. oranges, b
Decision tool
Cynefin Framework
Determine what domain you're operating in — simple, complicated, complex, or cha
Decision tool
Reframing
Restate the problem from a different angle to ensure you're solving the right th
Decision tool
Abstraction Laddering
Move up (why?) or down (how?) levels to find the right altitude for your problem
Decision tool
SWOT Analysis
Map internal strengths/weaknesses against external opportunities/threats to unde
More like this, in your inbox
I send a newsletter every week — free, no spam, unsubscribe anytime.