Data should come with a traceable source.
Discovery is useful when provenance, uncertainty and conditions are visible. A source record, a technical check and a human review are distinct.
Searching and ranking
Search runs in the local PostgreSQL catalog across titles, descriptions, keywords, publishers and topic labels. Titles receive the highest weight. Accents are ignored. A maintained Dutch–English–French synonym file connects common terms such as bevolking/population and gezondheid/health/santé. A single unknown word of four or more characters can use word similarity for limited typo tolerance. This is not general translation or semantic search.
Filters and dates
Explicit filters are hard constraints. Country searches include worldwide datasets and exclude unknown coverage. Period filters require overlap with known source time boundaries. Format, access, cost and license must match the same active distribution. Only confirmed free distributions appear by default; limited free access is opt-in. Paid and unknown prices are excluded. Relevance has no age bonus or penalty. Added means first catalog import; data updated means a publisher data change. Unknown update dates sort last, with stable IDs resolving ties.
Imports and editorial changes
Only approved connectors can automatically publish validated records. Stable source IDs and official identifiers prevent repeat imports. Similar titles only create review candidates. A merge preserves both original records and can be reversed. Editorial text is stored separately and survives synchronization. Absence from a limited import is not a removal signal; only explicit source deletion can archive automatically.
Separate dates
Data coverage, publisher data update, source metadata update, catalog import, technical link check and human review have separate fields. A reachable link does not prove content accuracy. Missing information remains missing.
Samples and maps
A worker fetches approved CSV and JSON extracts where caching is permitted, normally limited to 512 KiB, ten seconds, 100 rows and 24 columns. Unsupported sources show an explanation. World Bank samples show a first API page; Eurostat samples show an explicit filtered extract. No aggregates or conclusions are calculated. Maps depict catalog country or region categories on Natural Earth boundaries, not observations or spatial analyses.