A clerk’s office must approve model promotions using production evidence. Which practice supports that governance?

Reviewing monthly production metric packs with business owners before promotions is correct because a clerk’s office that must approve model promotions needs production evidence, not folklore. Regular metric packs—accuracy, drift, latency, error rate, and related civic KPIs—create a shared evidence trail that ties registry or endpoint promotion to observed live performance. Business-owner review embeds governance in the promotion cadence.

Reviewing monthly production cost summaries with business owners before promotions fails evidence-based governance because cost summaries show spending, not production evidence like accuracy, drift, latency, and error rate the clerk's office can check against live performance. Reviewing monthly production metric packs with data scientists before promotions fails the approval requirement because the packs are the right evidence but data scientists are not the business owners whose sign-off the clerk's office requires. Reviewing monthly production feedback excerpts with business owners before promotions fails governance because excerpts are selective anecdotes, not systematic metric packs that provide production evidence for approval.

Exam caveat: promotion governance stems that mention production evidence want scheduled metric review with business owners before promote—not monthly production cost summaries with business owners, monthly production metric packs with data scientists, or monthly production feedback excerpts with business owners.

Operational check: publish a monthly production metrics pack for the clerk’s office models, hold a short review with business owners, record go/no-go on candidates, and block registry promotion until that review is logged.