How we taught search to find animals nearby, not “somewhere in the city”
In "Find a Friend," we introduced geo search with PostGIS and GeoHash: search now takes into account proximity, listing density, push matches, and time relevance.
In "Find a Friend", we implemented geo search using PostGIS and GeoHash. The task sounds simple, but in practice it’s classic IT: in central Moscow there are hundreds of listings, while in the suburbs there are a couple and a tumbleweed.
What we did:
— we look for lost animals nearby, not in an abstract district; — we take listing density into account: in overpacked areas we narrow results more precisely, in sparse ones we expand them sensibly; — we send push notifications if a match appears within the specified radius; — we added time dynamics: the older the listing, the lower its relevance.
The result: search became smarter and closer to real life. Because, as with delivery or dating apps in 2025, users expect instant and contextual results — only here the cost of a mistake is much higher.