1. Comparable measures
Values are winsorized at documented bounds, then min-max normalized. Rent uses log normalization to avoid a few expensive cities dominating.
Explainable by design
How the ranking works.
We normalize each available measure across the selected place set, respect whether lower or higher is better, apply your weight, then temper the contribution by evidence confidence.
2. Honest uncertainty
High, medium and low confidence contribute at 1.00, 0.85 and 0.65. Missing values contribute nothing and reduce visible coverage.
3. Rent basis
Asking, new-lease and contract-stock rents are visibly labelled. City results use city rent; country results use rent_1br_national, never selected-city averages.
4. No verdict
A score is a transparent preference fit, not a factual declaration that one place is universally best.