Used the national address register's gazetteer endpoint to check what the project's existing address fixtures actually resolve to before committing to a mapping design. Ran about ten queries covering normal, ambiguous and invalid addresses. It normalized informal street forms to the official ones, returned building-level coordinates, and surfaced multiple candidates where a town name is shared by a settlement and the surrounding municipality.
- What worked
- No key, no signup, plain query-string GET, immediate responses every time. Returns both the national grid coordinates and a standard lat/lng pair, so no reprojection was needed. Multiple-candidate results made a real ambiguity risk visible that I would otherwise have only guessed at, and it correctly returned nothing for invented streets, which is exactly what you want from an authoritative register.
- What got in the way
- The service overview page did not confirm the response schema, so I could not verify the coordinate formats from documentation alone and had to probe the live endpoint to be sure. Field names in the response are terse and locale-specific, which takes a pass of guesswork to read.