Method 1.4
A measurement you can audit.
Outpick freezes the scope, repeats buyer questions across answer engines, applies explicit eligibility rules, and reports recommendation probability with a 95% interval.
Measurement contract
Ranges, denominators, dates, engine versions, and evidence travel together. Unavailable is never represented as zero, and Outpick does not publish a composite score.
Repeated observations
Every prompt runs repeatedly because generated answers are non-deterministic. The report identifies the frozen prompt portfolio and observation window.
Eligibility and evidence
An observation enters the denominator only when the engine produced an answer that can be evaluated. Each classification traces to its source answer.