Home·Rating model
Rating model

Six inputs. One rating. Every input is visible.

The ParentProof Score is a Bayesian Wilson lower bound on a 0–100 range. We pick that statistical method on purpose; the score is conservative under low review volume and stable under brigading.

What the score is

The headline rating combines six inputs into a single number on a 0–100 scale. The combination is weighted, normalized, and audited; the inputs are visible on every listing page so a parent (or a journalist, or the listed entity) can inspect any of them.

Why Bayesian Wilson lower bound

A naive average misleads when review volume is low: a listing with 3 five-star reviews looks identical to a listing with 3,000. The Wilson interval gives a confidence-aware lower bound that grows with sample size; the Bayesian variant lets us encode a sensible prior so brand-new listings are not penalized to zero. The score you see is the lower bound of that interval, which is the conservative reading.

What the score is not

  • Not a ranking. Listings are not sorted into a leaderboard; the score is a measurement, not a competition.
  • Not a ban list. A low score is not a recommendation to remove the entity from a child's life; it is information.
  • Not a universal good/bad label. Suitability depends on the child, the household, and the use case. The age band, drift indicator, and verdict line carry context the score alone does not.

What you see on every listing

The six inputs

Every input is logged on every listing with timestamp and source. Click any input below for the full methodology, weighting policy, and moderation procedures.

How disputes work

Any entity can dispute their rating via the community council process. The council can override AEGIS classification, attach context to community reviews, or trigger a full re-rating. Disputes are public; the decision and reasoning are published in the decisions log.