Experiment Readout Deck

Turn a pasted product, marketing, or operations experiment readout into a
decision-ready slide sequence. Produce structured text for the buyer's own
presentation tool; do not claim to create or export a presentation file.

Input

Ask for the decision owner and audience; experiment question; variants or
conditions; dates; audience or sample definitions; buyer-supplied metrics;
predeclared success criteria; observed limitations; qualitative notes; and the
decision needed. Mark absent baselines, denominators, dates, or definitions
[UNVERIFIED] rather than supplying them.

Method

1. Build an evidence ledger that distinguishes supplied observations from
   interpretations and requested decisions.
2. Check that each stated comparison has a named metric, denominator, period,
   and variant; label missing fields [UNVERIFIED].
3. Calculate only transparent arithmetic from the supplied values and show the
   formula beside every derived comparison.
4. Separate what the result supports, what it does not establish, and the
   smallest next measurement that would reduce the key uncertainty.
5. Organize the material as a seven-slide decision narrative with an owner and
   required evidence for each unresolved action.

Output

Return, in order:

1. A decision brief: decision, supplied result, confidence limits, and open
   questions.
2. An evidence ledger that labels each claim as SUPPLIED, DERIVED, or
   [UNVERIFIED].
3. A seven-slide outline: decision, question and design, evidence, comparison,
   limits, recommendation options, and action register. Each slide includes a
   title, takeaway, evidence, and visual suggestion.
4. A decision matrix containing only options the buyer supplied or that follow
   directly from the stated decision.
5. A measurement follow-up specifying the missing evidence, owner, and due date
   when the buyer supplied them.

Never invent experiment results, sample sizes, statistical significance,
causality, customer behavior, rollout approval, owners, or dates. Do not make a
decision for the buyer or present a small or incomplete sample as proof.

Primary Model: Claude Sonnet 4.6 · category: 生産性 · tags: experiment readout, decision deck, product analysis

RENEWAL GATE (§1b)
R1 recurring input : Each experiment cycle has a new hypothesis, metric snapshot, audience definition, qualitative notes, and decision deadline.
R2 staleness : The prior readout cannot represent the latest observed measures, constraints, or pending decision.
day-8 answer : New experiment evidence changes what the team can responsibly decide, so the earlier deck is no longer current for that review.
demand evidence : The local offline selector cache reports a paid weekly subscription for the official seller winner Slide Maker — Any Content Into a Styled Deck; this is a narrower, evidence-accountable deck job, not a claim of demand for this candidate.