One-Way vs Two-Way Door

This is a meta-decision tool: a triage that runs before any option comparison and asks one question - how reversible is this decision? - then matches the deliberation and sign-off to the answer. A two-way door is reversible (walk back cheaply), so decide it fast, low, and light. A one-way door is hard or expensive to reverse, so give it real rigor and senior sign-off. The load-bearing move is separating the reversibility judgment from the decision itself, which routes scarce deliberation toward the irreversible few and licenses speed on the reversible many. The output is a reversibility classification plus a matched deliberation level - it says how much machinery the choice deserves, never which option to pick.

When to Use

A decision is on the table and it is unclear how much process it deserves.
Someone is about to rubber-stamp something irreversible, or convene a committee over something trivially reversible.
A team or org is chronically slow, applying the same heavyweight approval to everything regardless of stakes.
You want an explicit, defensible reason - decided before deliberation starts - to either move fast or slow down.

When NOT to Use

When the decision is already known to be high-stakes and is being analyzed. Triage already happened; you are past this tool. Use a risk tool (premortem) or an option comparison, not a classifier that confirms what you know.
When you need to actually compare options against criteria. That is think-decision-option-review (a criteria-weighted option matrix). This skill triages before any comparison how much analysis the decision warrants; it never scores or recommends an option.
As a license for speed alone. If the irreversible-but-inconvenient consequences (trust, legal, path-dependence) get waved away, the classification is motivated, not honest.
As theater to bless a fast decision by labeling a one-way door a two-way door. The verdict has to be defensible.
For a routine, obviously reversible call with no meaningful reversibility question. Just decide; classifying it is its own small over-process.

Instructions

When asked to triage a decision by reversibility, follow these steps:

State the decision in one line. Name the specific choice being made. If it is already known to be high-stakes and under analysis, or if the request is to compare options, say so and point to the right tool (premortem or think-decision-option-review) instead of classifying.
Test reversibility against named dimensions. For each, state what it would cost to walk the decision back: money, time, trust/reputation, legal/contractual, and foreclosed future options (path-dependence). Do not accept the convenient label; a decision that feels reversible can carry one-way consequences.
Render the verdict. One-way door or two-way door. For borderline cases, state which way it leans and the single dimension that decides it. The verdict follows from step 2, not from how fast someone wants to move.
Match the deliberation level. Two-way door: who can decide it now, with what light analysis, and why slowing it is the real cost. One-way door: the rigor and sign-off it warrants, and a pointer to the heavier tool it should go to next (option comparison, premortem).
Emit the classification artifact and a one-line summary. Produce the artifact in references/TEMPLATE.md. It routes the decision; it does not make it.

Output Format

Use the template in references/TEMPLATE.md. The deliverable is the filled classification (verdict, reversibility dimensions, matched deliberation level) plus a one-line summary, not a prose essay and not a recommendation about which option to choose.

Quality Checklist

Before finalizing, verify:

Reversibility was tested against multiple named dimensions (cost, time, trust, legal, path-dependence), not asserted from the convenient label.
The verdict follows from those dimensions, and borderline cases say which way they lean and why.
The matched deliberation level is concrete: who decides, how much analysis, what sign-off.
A one-way door points to the heavier tool it should now go to (option comparison, premortem); the triage does not pretend to be the analysis.
The output recommends a level of process, never which option to pick.
No overclaiming: the skill claims better-calibrated effort, not a better outcome (see evidence/dossier.md).

Evidence

Tier P (practitioner). The framing comes from Bezos / Amazon shareholder letters and broad management practice; it is internally coherent and targets a real pathology (uniform heavyweight process applied to reversible decisions). There is no controlled evidence that the two-bucket classification improves decision outcomes, speed, or quality versus any other rule, and reversibility is often misjudged. The evidence is transferred from human practice, not AI-validated; the AI value is making the triage cheap, forcing the reversibility question explicit, and producing a durable classification. Full grading, sources, and caveats: evidence/dossier.md.

Examples

See references/EXAMPLE.md for a completed classification on a real decision.

Deep dive: worked example

A full worked run (the shared Northwind scenario)

Reversibility Classification - Worked Example

A completed run of the think-one-way-vs-two-way-door skill on a real decision. This is the quality bar a generated classification should meet.

Uses the shared recurring scenario (Northwind, a B2B SaaS weighing a self-serve free-tier launch) so examples across skills read as one coherent product. This skill runs first: it triages how much process the decision deserves, before any option comparison or premortem.

Decision being triaged

Decision: Launch a self-serve free tier of Northwind’s B2B SaaS to accelerate top-of-funnel growth.

Summary (top of the artifact)

One-way door. A public free tier is cheap to ship but expensive to retract - pulling it later damages trust and is path-dependent on pricing - so this gets full rigor and senior sign-off before committing, not a quick call. Route it next to an option comparison and a premortem.

Reversibility test

Dimension	Cost to walk it back	Reversible on this dimension? (Y / N / partial)
Money (spend / refunds / write-offs)	Build cost is sunk but modest; infra spend stops if the tier is killed	Y
Time (how long to undo)	Turning the tier off is fast technically; communicating a withdrawal is slow	Partial
Trust / reputation (customers, market, team)	Withdrawing a launched free tier reads as “they are struggling” and burns trust with users who adopted it; competitors cite it	N
Legal / contractual (commitments, regulation)	Free users on month-to-month terms are low-risk; any data-retention or migration promises made at sign-up could bind	Partial
Path-dependence (future options foreclosed, learning sunk)	A free tier reframes the market’s price expectation and the sales motion; unwinding the freemium position is very hard once set	N

Verdict

Classification: One-way door.
For borderline cases - leans: Not borderline. The money and time are reversible, but trust and path-dependence are not - and on this kind of decision the irreversible dimensions dominate. The “we can just turn it off” framing is exactly the convenient label this test exists to reject.

Matched deliberation level

Who decides: CEO + VP Sales + Head of Growth jointly, not the growth team alone.
How much analysis: Full. Compare the free-tier launch against alternatives (extended trial, sales-assisted PLG, no change) and stress-test the chosen path before committing budget.
Sign-off: Executive sign-off required before any external announcement; cross-functional alignment with Sales on comp and lead-routing first.
Next tool (one-way doors only): think-decision-option-review to compare free tier vs the alternatives against weighted criteria, then think-premortem to stress-test the chosen path before commit.
Why this level: Because the decision is hard to reverse on trust and pricing position, the cost of being wrong is high and durable. Speed here is a false economy; the rigor is warranted. (Contrast: the pricing copy on the free-tier page is a two-way door - let Growth A/B test it without sign-off.)

Note: the value is the routing decision made before any work starts. A naive prompt would jump straight to analyzing or even recommending the free tier; this skill first establishes that the decision is irreversible on the dimensions that matter, so it earns full rigor - and just as importantly, it flags the reversible sub-decisions (the page copy) that should stay fast.

Grounding: the full evidence dossier

What the research does and does not show, with graded sources

Evidence Dossier: One-Way vs Two-Way Door

The single source of truth for the one-way-vs-two-way-door skill. The SKILL.md, the sidecar (skill.meta.yml), and the eval cases all derive from this file. If a claim is not here, it does not belong in the skill. Author this FIRST.


Skill	`thinking-framework-skills.one-way-vs-two-way-door` (installable name `think-one-way-vs-two-way-door`)
Family	decision-and-option-evaluation
Evidence tier	P (practitioner; limited controlled evidence - see “What the evidence shows” below)
Confidence	Moderate that the triage move prevents real waste; low that the specific two-bucket framing is the optimal taxonomy or that it improves outcomes in controlled study
Status	draft (first authored 2026-05-31, against discovery corpus)

1. The mechanism (what actually does the work)

This is a meta-decision tool: a triage that runs before any option comparison and asks one question - how reversible is this decision? - then matches the amount of deliberation and the level of sign-off to the answer.

A two-way door is reversible: if the choice turns out wrong, you walk back through with little cost. Such decisions should be made fast, by the people closest to them, with light deliberation. Slowing them down is the real cost.
A one-way door is hard or expensive to reverse: walking back is costly, slow, or impossible. Such decisions warrant real rigor and senior sign-off before committing.

The load-bearing move is separating the reversibility judgment from the decision itself. Most decision processes apply one uniform level of care to everything, which over-deliberates the reversible many and under-deliberates the irreversible few. By forcing an explicit reversibility classification first, the method routes effort where it pays off and licenses speed where it does not. The output is a classification plus a matched deliberation level, not a recommendation about the decision’s content - this skill never says which option to pick; it says how much machinery the choice deserves.

The classification is deliberately coarse (two buckets, with a “lean” verdict for borderline cases) because its job is routing, not analysis. The moment you are doing the analysis, you are past this tool’s job.

2. Lineage

Bezos shareholder letters. The “Type 1 / Type 2 decision” and “one-way door / two-way door” framing was popularized by Jeff Bezos in Amazon’s 1997 and especially the 2015 and 2016 Amazon shareholder letters, arguing that as organizations grow they tend to apply heavyweight Type 1 (irreversible) process to Type 2 (reversible) decisions, producing slowness and risk-aversion.
Related decision-theory roots. The value of preserving reversibility connects to the economics of irreversibility and option value (the cost of foreclosing future choices), but Bezos’s framing is a practitioner heuristic, not a formalization of that literature.

No trademark. “One-way door / two-way door” and “Type 1 / Type 2 decision” are descriptive phrases in common business use; no attribution is required and none is claimed. We name the skill descriptively by its mechanism (reversibility-based triage) and cite the lineage here rather than branding it.

3. What the evidence shows, and what it does NOT show

This is the honest core of the dossier. The skill must not overclaim.

What is reasonably supported (the practitioner basis):

The framing is widely adopted in practice and is internally coherent: matching process weight to reversibility is a sensible allocation of scarce deliberation, and the failure it targets (uniform heavyweight process applied to reversible decisions) is a real and commonly observed organizational pathology.
As a forcing function, an explicit reversibility check reliably changes behavior: it gives teams permission to decide reversible things fast, and a defensible reason to slow down on irreversible ones. That behavioral effect does not depend on any outcome study.

What is NOT shown (the caveats that keep the skill honest):

There is no controlled evidence that this two-bucket classification improves decision outcomes, speed, or quality versus any other triage rule (or versus no triage). The support is practitioner testimony and a single influential source (Bezos), not experiment.
Reversibility is often misjudged. Decisions framed as two-way doors can carry one-way consequences (reputational, trust, legal, path-dependence, sunk learning) that the speedy bucket hides. The binary tempts people to under-classify genuinely irreversible calls as reversible because reversible is the convenient answer.
It is a triage, not an analysis. It says how much rigor a decision deserves; it does not perform the rigor, surface risks, or compare options. Treating the classification as if it resolved the decision is a category error.
The “fast = good” reading is a misuse. The method’s value is calibration, not speed for its own sake; speeding up a misclassified one-way door is exactly the harm it is meant to prevent.

Net grade: P. Useful practitioner method with a clear mechanism and a real failure mode it addresses; thin controlled evidence. The skill should claim the calibration/forcing-function value and explicitly disclaim any outcome-quality guarantee.

4. Transferred-evidence flag (required honesty for this library)

The basis for this skill is human organizational practice (Bezos / Amazon and broad management adoption), not controlled studies, and certainly not any study of AI-augmented use. There is no evidence on whether an AI agent classifies reversibility well, or whether an agent-produced classification improves a human’s decision routing. The evidence is therefore transferred from human practice, not AI-validated. This skill must say so. Treat the AI value as: the agent makes the triage cheap and habitual, forces the reversibility question to be answered explicitly before effort is spent, names the dimensions that make something hard to reverse (so reversibility is judged, not assumed), and produces a durable, inspectable classification artifact. None of that depends on the unproven outcome claim.

5. When it works / when it fails (drives the eval negative cases and “When NOT to Use”)

Works best when:

A decision is on the table and it is unclear how much process it deserves - someone is about to either rubber-stamp something irreversible or convene a committee over something trivially reversible.
A team or org is chronically slow, applying the same heavyweight approval to everything regardless of stakes.
You want an explicit, defensible reason to either move fast or slow down, decided before the deliberation starts.

Fails or misleads when (poor-fit / anti-patterns):

The decision is already known to be high-stakes and is being analyzed. Triage has already happened; you are past this tool. Use a risk tool (premortem) or an option comparison, not a classifier that will just tell you what you already know. (Anti-trigger.)
You need to actually compare options against criteria. That is think-decision-option-review (a criteria-weighted option matrix). This skill triages before any comparison how much analysis the decision even warrants; it never scores or recommends an option. (Near-miss anti-trigger against the overlapping neighbor.)
Reversibility is treated as a license for speed alone, with the irreversible-but-inconvenient consequences waved away. The method must test reversibility against multiple dimensions (cost, time, trust, legal, path-dependence), not accept the convenient label.
Used as theater to bless a decision someone already wants to make fast, by labeling a one-way door a two-way door. The classification has to be defensible, not motivated.
For a decision with no meaningful reversibility difference (a routine, repeated, obviously-reversible operational call) - just decide; classifying it is its own small over-process.

6. Output artifact

The skill must emit a reversibility classification plus a matched deliberation level, not prose and not a recommendation about the decision’s content. Concretely: the decision stated in one line; a verdict (one-way door / two-way door, or “leans” with the reason for borderline cases); the reversibility tested against named dimensions (what it would cost in money, time, trust/reputation, legal, and foreclosed future options to walk it back); the matched deliberation level (who decides, how much analysis, what sign-off); and, for one-way doors, a pointer to the heavier tool the decision should now go to (option comparison, premortem). The artifact is the deliverable; it routes the decision, it does not make it.

7. Sources

Bezos, J. - Amazon.com 1997 Letter to Shareholders (the original “Type 1 / Type 2” framing of decision reversibility).
Bezos, J. - Amazon.com 2015 and 2016 Letters to Shareholders (“one-way door / two-way door”; the argument that growing orgs over-apply Type 1 process to Type 2 decisions).

Verification status: the Bezos shareholder-letter attribution (citations 1-2) is well-attested and the framing is widely reproduced, but the exact letter-by-letter wording was drawn from secondary synthesis in the discovery corpus and should be confirmed against the primary letters before any public-facing claim quotes them directly. There is, by design, no outcome-effectiveness citation: section 3 states plainly that none exists, which is the honest position for a P-tier practitioner method.

Thinking Framework Skills v0.3.0 · 38 frameworks