Two-Stage Review¶
Workflow — instantiates Bounded-Rationality Decision Design
Separates rapid provisional action from later verification, correction, or ratification.
A Two-Stage Review splits a single decision across time: a fast, reversible provisional action is taken now so that work can proceed, and a slower verification follows that confirms, corrects, or reverses it. The idea that makes it this mechanism and not a sibling is that it buys speed under time pressure without abandoning rigor — it defers the rigor. Stage one is deliberately allowed to be approximate because stage two is guaranteed to check it; the whole design rests on the provisional action being cheap to undo if stage two overturns it. It is one decision with two passes, not a fast method and a separate audit of the design.
Example¶
A software team ships a change to its recommendation service using a canary release.[n1] Stage one is provisional and fast: the new version is deployed to 5% of live traffic while the old version keeps serving the other 95%. No committee meeting, no week of pre-launch analysis — the change goes out under a rule that any single team can invoke. Stage two is verification: over the next few hours the system watches error rates, latency, and the business metric against the 95% baseline. If the canary holds, the change is ratified — rolled out to 100%. If it degrades, it is reversed — traffic snaps back to the old version automatically, and the change never reached most users. The provisional deploy was allowed to be bold precisely because the reversal was one click and the exposure was capped at 5%. The team gets the speed of "ship it now" with a real check standing behind it.
How it works¶
The workflow works by (1) taking a fast provisional action under a lightweight method, scoped so its consequences are bounded and recoverable; (2) recording what was decided, on what basis, and what would confirm or refute it; (3) running a heavier verification within a defined window — a second reviewer, a metric check, a confirmatory test; and (4) ratifying, correcting, or reversing on the verification's result, with the whole two-step trail left inspectable. Its distinctive move is decoupling the moment of action from the moment of confidence: the action commits early and cheaply, the confidence arrives later, and the reversibility of stage one is what makes acting-before-being-sure responsible rather than reckless.
Tuning parameters¶
- Provisional-action boldness — how far stage one is allowed to commit before verification; bolder provisional actions move faster but demand cheaper, faster reversal.
- Verification depth — how heavy stage two is (a glance, a second reviewer, a full confirmatory analysis); deeper checks catch more but delay ratification.
- Reversal window — how long the provisional action stays reversible; longer windows are safer but hold the decision open and consume attention.
- Auto- vs. human ratification — whether stage two clears automatically on a metric or requires a sign-off; automation is faster, human ratification is more accountable for consequential cases.
- Stage-one method — which lightweight method from the portfolio runs the provisional action; it must be fast and genuinely reversible for the pattern to be safe.
When it helps, and when it misleads¶
Its strength is that it dissolves the false choice between "act now, blind" and "wait until certain": it lets a reversible action go early and pays for the speed with a guaranteed later check. Under real time pressure with a recoverable move available, it is often the only responsible way to be fast.
Its characteristic failure is stage two silently not happening — the provisional action is taken, everyone moves on, and the verification is skipped, so a "temporary" decision quietly becomes permanent without ever being confirmed. The classic misuse is applying it where stage one is not actually reversible: a provisional action whose consequences can't be undone gives you all the risk of acting blind with none of the safety the pattern promises. The guarding discipline is to make stage-one reversibility a precondition, not a hope, and to instrument stage two as a commitment with an owner and a deadline — an unverified provisional action should expire or escalate, never coast.
How it implements the components¶
stakes_and_reversibility_profile— the pattern's precondition: it works only when stage one can be scoped to a bounded, recoverable action, so reversibility is assessed before anything ships.method_portfolio— it deliberately assigns two methods to one decision, a light provisional one and a heavier confirmatory one, drawn from the portfolio and sequenced in time.accountability_and_transparency_contract— the recorded provisional-then-verified trail keeps the decision inspectable, so a fast first action never means an unaccountable one.
It does NOT aggregate closed decisions to retune the design's thresholds via outcome_feedback_and_calibration_loop — that is Post-Decision Calibration Review. Two-Stage Review verifies one live case as it happens; its twin learns across the whole history after the fact.
Related¶
- Instantiates: Bounded-Rationality Decision Design — the workflow is the archetype's provisional-action-then-verification move, the way it stays fast under time pressure without going careless.
- Sibling mechanisms: Post-Decision Calibration Review · Algorithmic Escalation Gate · Default and Delegation Protocol · Satisficing Threshold Rule · Timeboxed Search · Progressive Option Screening · Cognitive Offloading Aid · Choice Architecture Simplification · Decision Method Triage Matrix
Editorial Notes¶
Form Classification¶
Form family: Protocol, Workflow & Routine
Rationale: Two Stage Review is defined in the frozen evidence as: Separates rapid provisional action from later verification, correction, or ratification. Its operative deployed or enacted form is therefore Protocol, Workflow & Routine.
Nearest alternative: Assessment, Review & Assurance — Assessment, Review & Assurance can support this mechanism, but the evidence centers the concrete operation described above rather than the alternative family's defining operation.
Review outcome: Adjudicated after independent review; medium confidence.
Origin Attribution¶
Primary origin: Organizational & Management Science
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Universal
Rationale: Two stage review is rooted in organizational coordination, workflow, and capability development; historically, that field developed the core operation described here: separates rapid provisional action from later verification, correction, or ratification.
Related originating lineages:
- Law & Governance — Legal doctrine, regulatory governance, and procedural accountability supplies a parallel or contributing lineage for the mechanism's defining operation: separates rapid provisional action from later verification, correction, or ratification.
- Statistics & Experimental Design — Statistics, experimental design, and measurement theory supplies a parallel or contributing lineage for the mechanism's defining operation: separates rapid provisional action from later verification, correction, or ratification.
- Systems Thinking & Cybernetics — Feedback, system boundaries, stocks, flows, and regulation supplies a distinct formative lineage for the mechanism's two stage review logic.
Review resolution: The blind reviewers agree that organizational_management is the primary origin and differ only on alternate origin disagreement, origin mode disagreement, encyclopedia synthesis disagreement. I preserve every independently explained alternate from both records rather than imposing a numeric cap. I retain cross_disciplinary_synthesis because the combined evidence shows material contributions from several lineages. The broader reach of universal records portability separately from historical provenance; encyclopedia_synthesis=true preserves the affirmative synthesis judgment where either reviewer identified one.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] A canary release deploys a change to a small slice of live traffic or users, watches its behavior against the unchanged baseline, and only then promotes it to everyone — or rolls it back. The name recalls the canary in a coal mine: a small, expendable early-warning exposure that catches trouble before it reaches the whole population. ↩