Compare consequence and correct the method

UNTESTED · claim_state: UNTESTED · outcome_maturity: UNTESTED

Candidate teaching projection; contextual recipe unproved. The JTBD Ledger keeps this route as a gap until a contextual owner accepts the method and retains decision evidence.

Workflow family: Analytics & Feedback.

GATE

• Authority: human_authority: Accountable outcome and method owner
• Beneficiary: people named before the action and the accountable method owner
• Permitted inputs and actions: the frozen prediction, baseline, mandate, method revision, permitted records, and observed consequences with provenance
• Prohibited: invented outcomes, moved baselines, excused boundary breaches, and another run without fresh authority
• Stop: stop_rule: Stop for missing provenance, incomparable measures, material confounding, harm, or absent authority.

CHECK

Classify each relevant claim as supported, contradicted, or unknown. Keep output separate from beneficiary consequence and preserve both the original prediction and the later observation. Never invent an outcome to close the loop.

NEXT MOVE

The accountable human chooses continue, revise, or stop, names the correction owner and review date, and returns the observation to the narrowest owner: Control, Beliefs, Values, or Purpose.

Complete the outcome receipt

Click ≠ proof. A pageview, prompt copy, generated artifact, or producer self-approval is not verified outcome evidence. Complete the checklist with observations after the authorised action; leave unknowns visible.

Put this to work

Copy the outcome receipt template

For the accountable outcome owner

Copy this prompt. Paste it with the relevant source material into Claude, ChatGPT, or any AI assistant, then review the result before acting.

OUTCOME RECEIPT

Intent named — question / job: [ ]
Beneficiary: [ ]
Mandate / authority boundary — who may act and what is prohibited: [ ]
Action taken — one sentence and time bound: [ ]
Prediction recorded before action: [ ]
Observation after action — what changed, for whom, harm or exception: [ ]
Human decision — continue / revise / stop: [ ]
Owner and review date: [ ]
recipe_id / ledger_address: business.analytics.compare-outcome-correct-method
method_revision: [ ]
Permitted inputs used: [ ]
Assistant draft or assessment: [ ]
Human assessment and corrections: [ ]
Stop events / boundary exceptions: [ ]
claim_state: UNTESTED

• Intent named: question or job and beneficiary
• Mandate and authority boundary: who may act and what is prohibited
• One time-bounded action taken
• Prediction recorded before action
• Observation recorded after action: change, beneficiary, harm, and exception
• Human decision: continue, revise, or stop
• Owner and review date
• recipe_id / ledger_address, method revision, and permitted inputs
• Assistant draft separated from human assessment and corrections
• Stop events and boundary exceptions
• claim_state remains UNTESTED until a separately authorised trial

Outcome

The beneficiary and method owner receive an evidence-bounded decision to continue, change, pause, or stop, plus the owner and input that must change.

Use When

Trigger this job at the frozen review point, after a stop, or when a material consequence arrives. The outcome owner controls interpretation, accepted trade- offs, retention, correction, further tests, and stop.

Inputs

Bring the original beneficiary and value standard, prediction and confidence, baseline, method revision, mandate, output, observed consequence with provenance, protected constraints, confounders, effort and cost, and review rule.

Recipe

1. Freeze the original prediction, standard, method revision, and decision rule.
2. Separate delivered output from observed beneficiary consequence.
3. Compare prediction with observation; name magnitude, timing, uncertainty, confounders, and protected-constraint effects.
4. Identify which value assumption, belief, control, or method step explains variance.
5. Let the accountable owner choose continue, change one input, pause, or stop.

Human and AI Control

An assistant may organise records, calculate variance, and surface alternatives. It may not invent missing outcomes, move the baseline, choose values, excuse a constraint breach, or authorise another run. Stop for missing provenance, incomparable measures, material confounding, harm, or absent decision authority.

Output and Handoff

Return expectation, observation, variance, confidence, costs, constraint effects, decision, correction, owner, and next review. Send a method correction to the page that taught the job; send authority correction to authorise bounded work.

Evidence and Stop

Retain original and observed records without silently rewriting either. Check that another reader can reproduce the comparison and distinguish output from outcome. Stop the claim when beneficiary evidence is absent, confounded, or below the frozen standard. A completed review does not prove the method improved.

Next

Change one decision-relevant input, state the next prediction and review point, and run again only with fresh authority. Stop permanently when protected costs outweigh the expected benefit or no safe discriminating test remains.