Every model,in the open
Each score blends every model that measured the same thing. Add a better source and every report improves; remove one and confidence drops.
How scores are made
Weight. Each model's weight is its authority times its resolution, its age and how specific it is to the place. A regional source outranks a global one where both exist.
Confidence. How much of a criterion's evidence is present, lowered when the models that measured the same thing disagree.
Rollups. A criterion with low confidence is shown but never moves its category, and a category with too little evidence stays ungraded.