Scores are only as useful as the records behind them.

Manufacturer: Ingredients, guaranteed analysis, calories, AAFCO statement, feeding directions, brand disclosures.

Retailer: Current price, package size, availability, product images, and alternate label captures.

Regulatory: FDA recalls, enforcement reports, warning letters, and product safety notices.

Methodology: AAFCO/FDA/FTC/Google source anchors and published label limitations.

Public-beta scorecards can ship when product label fields, source confidence, regulatory disposition, unresolved gaps, and score-event math are visible.

Verified status waits for stronger Brand Trust closure, complete warning-letter coverage, fresher prices, and source artifacts that remove the remaining unknowns.

Source-capture holds stay out of the live product fixture set until deterministic score inputs can be captured from primary or account-level label evidence.

Recipe label: Use a recipe-specific manufacturer, package, customer-plan, or account-level label artifact. Brand-level marketing pages and third-party review captures are triage only.

Same-recipe fields: Ingredients, guaranteed analysis, calories, and AAFCO adequacy text must all tie to the same recipe before scoring.

Price scenario: Value needs a reproducible dog profile, recipe, portion or calorie basis, plan frequency, captured date, and calculation math.

Discounts: Do not use intro or subscription discounts as the default score input unless the page explicitly labels the scenario as discounted.

Public state: If recipe label or deterministic price evidence is missing, show a source-hold/not-scored record rather than a numeric score or ranking.

Move only after source gates and QA criteria are proven across 10-25 products.

Move only after Brand Trust record shape is stable.

Move only after score events can reconstruct every public score.

Move when manual JSON editing becomes the bottleneck, not before.