Why we version the rubric.

Public scores can move when source records improve, conflicts get resolved, or the rubric itself changes. This page is the public record of every change — and why we make it.

Why we version

Stripe versions its API. Consumer Reports versions its test methods. We version the rubric.

A score has two parts. There is the data on the bag (the label, the nutrition panel, the price, the recall record) and there is the rubric we use to score it (the rules in /methodology/dog-food). If the rubric changes, the score changes. We do not silently improve our methods and quietly bump everyone’s rating — we publish the diff.

Every product scorecard shows the rubric version it was scored on. Every editorial byline names the rubric. When a version bumps, every page that depends on it recomputes on the same day, and the new version lands here with a note explaining what moved.

Versions, in order

What each version is doing

  1. April 28, 2026

    Dog food rubric · April 2026

    Active

    What it is. The first public PetScored rubric. Five sub-scores (Nutrition Fit 30%, Brand Trust 25%, Ingredient Clarity 15%, Sensitivity Risk 15%, Value 15%) compute a 0–10 composite. Each score is reconstructed from a deterministic rule-event trail visible on the product page.

    Why we wrote it. We needed a rubric that was honest about what a label can prove and tough on the rest. v1.0 is conservative on purpose. Most foods land between 6 and 8, which the rubric is designed to do.

    Identifier: v1.0

  2. April 28, 2026

    Six-state source labels · April 2026

    Active

    What it is. Every fact on a scorecard is labeled with one of six states: Prototype seed, Label captured, Partially verified, Verified, Conflict, or Stale. Lower confidence does not lower a score — it labels the gap.

    Why we wrote it. Other rating sites collapse evidence quality into a single number. We separate the score from the evidence behind it so a partially-verified 7.5 reads honestly different from a fully-verified 7.5.

    Identifier: Source confidence v1

  3. April 28, 2026

    Verification workflow · April 2026

    In progress

    What it is. Ten dog food products define the verification workflow before broader page scaling. Verification packets cover label, nutrition panel, recall, warning letter, and pricing source confidence.

    Why we wrote it. Methodology is only as good as the data feeding it. The gold set forces us to prove the workflow on a small, defensible set before opening the floodgates.

    Identifier: Gold set v1

What changes a version

The bar for a version bump

  • A new rule event that can affect a composite score (e.g., adding a feeding-trial credit).
  • A weight change to any sub-score (e.g., moving Nutrition Fit from 30% to 32%).
  • A threshold change inside an existing rule (e.g., tightening the dry-matter protein band).
  • A new sub-score, a retired sub-score, or a renamed dimension.

Wording fixes, source-confidence relabeling, and adding a new product do not bump the rubric version. They are tracked on the affected scorecard’s own page.