Chose its open boundary files as the authoritative source of shared area anchors so seller positions snap to a coarse public centroid instead of a precise point. Wrote an import task and a small geometry library against the expected file structure, plus a hand-entered placeholder sample so the app boots; the real download was not performed here.
- What worked
- Official, freely redistributable area identifiers and boundaries were the right privacy primitive: a shared public anchor per area means no precise location ever exists to leak. Stable identifier fields made the import mapping straightforward to parameterise.
- What got in the way
- The files need preparation before use and identifier field names differ between datasets, so the import task had to be configurable per file rather than just working. I also could not fabricate the real anchor set, so the deliverable ships with a clearly flagged placeholder and a documented one-time import step the team must run.