How we verify places
We would rather publish a small number of places we can stand behind than thousands we cannot. Here is exactly how a record becomes a page.
Where the data comes from
Every place begins as a record from a named source: OpenStreetMap, published Bengaluru civic datasets of parks, playgrounds and open spaces, and manual checks we make ourselves. Each source carries its own licence, attribution and reliability, and we keep the original record exactly as we received it.
How sources are combined
Raw records never become pages. They pass through staging, normalisation, candidate matching, entity resolution and attribute resolution before a single canonical place exists. When two sources disagree about a fact, we choose by source reliability, how recent the record is, and the type of field — and a manual check always wins.
How duplicates are handled
The same park often appears in several datasets under different names. We look for candidates within a short radius and then require real evidence before merging: near-identical position with a matching name, overlapping shapes, one record sitting inside another's boundary, a shared official identifier, or a person confirming it. Proximity alone is never enough. Anything uncertain waits for review instead of being published.
What verified means
Verified means the place has a usable real identity, a valid location, an accepted type, no unresolved duplicate, no unresolved conflict between sources, at least one trustworthy source, and at least a couple of facts worth knowing. Places we can see on the map but cannot yet stand behind stay visible in search and on the map, and are kept out of our indexed pages.
How freshness works
We track three separate dates: when the source itself was updated, when we imported it, and when the place was last checked. A checked date means we reviewed the record, not that every fact was inspected on the ground that day.
How corrections work
If something here is wrong or a place has closed, tell us and we will review the record against its sources. Corrections are stored as manual overrides, which means they survive future imports instead of being overwritten.