Evidence floors

An evidence floor is the minimum independent evidence a brand must accumulate before any score is published. It is the single rule behind most pages that show no number — and it is about evidence sufficiency, never about quality.

What it means

Two correlated sources are not two pieces of evidence. The floor counts effective independent sources — after collapsing feeds that echo each other — and only publishes a score once enough genuinely independent evidence exists, across the required categories.

Inputs & data fields

FieldTypeMeaning

source_weightfloatTrust weight of a source, 0–1, fixed by tier.

correlationfloatHow much a source echoes ones already counted.

eff_sourcesfloatEffective independent sources after collapse.

categoriessetWhich source categories are present (regulator, review…).

Process & formula

eff_sources = Σ wᵢ · (1 − ρᵢ)wᵢ = source trust weight · ρᵢ = correlation with already-counted sourcespublish ⟺ eff_sources ≥ 2.0 ∧ {regulator, review} categories present

Collapse correlated feeds (syndication, mirrors, operator echoes) into effective independent sources.
Require the regulator category and the review-derived category to each clear their own minimum.
If eff_sources ≥ 2.0 and both categories are present, the score publishes; otherwise it is withheld.

How it affects public pages

below floorrenders Insufficient evidence — score withheld, gap shown.

at / abovethe score and its interval publish on the /evidence page.

Worked example

Lumen Exchange · fake data

Sourcewᵢ1−ρeff

AggregatorX1.01.001.00

Operator blog (echo)0.80.500.40

eff_sources1.40

1.40 < 2.0 → score withheld. A provisional 6.1 exists internally but is never published. Two more independent sources would clear the floor.

Publication states →Source weighting →Country dashboard →