Should you build or buy Maps, Location & Geofencing APIs?
Maps, Location, and Geofencing APIs give applications the location primitives they need: map tiles for display, geocoding to convert addresses into coordinates, routing and directions, and geofences that trigger an action when a device enters or leaves an area. Google Maps Platform, HERE, Mapbox, and Radar are the major providers, each priced per call.
Buy the APIs, and shop them, because Google is no longer the only price. Open tiles with Protomaps or OpenMapTiles and PostGIS geofences are proven and cheap to own, but consumer POIs, traffic and global address handling are where the vendors earn their fee, and running your own geocoder does not make it accurate. Use the hybrid where a workload is heavy and simple; otherwise pay per call.
Build it, buy it, or bridge?
When building makes sense
The open-source location stack is production-proven, not experimental. Protomaps serves map tiles for Felt, OpenInfraMap, and The Washington Post without a per-call fee to a commercial provider. Valhalla, OSRM, Nominatim, and Pelias cover routing, geocoding, and search, and PostGIS handles point-in-polygon geofences at any scale a database can hold. Google's own March 2025 pricing change replaced its flat $200 monthly credit with per-SKU free tiers and expanded automatic volume discounts scaling to 5,000,000-plus monthly billable events, a sign of how commoditized the base API pricing has become even from the largest vendor. What self-hosting does not reproduce is rich points of interest, live traffic conditions, and global address quality tuned by a vendor's own data operations. Build the tile and geofence layer when your workload is heavy and simple; that is where owning the infrastructure pays for itself fastest.
When buying makes sense
Buying the API is still the right call for the parts of the stack a self-hosted stack cannot cheaply reproduce: rich points of interest, live traffic, and address quality across every country you operate in. Google's own pricing has gotten more favorable since March 2025, when it replaced a flat monthly credit with per-SKU free tiers and expanded volume discounts, and competitors like Radar and HERE now undercut Google on plain geocoding. Location APIs are infrastructure regardless of provider; no company wins on which geocoder it calls. Pay per call for geocoding, places, and traffic, and reserve the self-hosted path for the basemap and geofence workload where the volume is high enough and the data need is simple enough that owning it actually saves money once engineering time is counted.
The desk read
Build-versus-buy analysis for Maps, Location & Geofencing APIs is being written. In the meantime, the framework that drives every B4 call is on the B4 Index page.
Frequently asked
Did Google Maps pricing actually get cheaper recently?
Yes, for many workloads. On March 1, 2025, Google replaced its flat $200 monthly credit with per-SKU free usage tiers and expanded automatic volume discounts scaling to 5,000,000-plus monthly billable events, up from 100,000-plus previously, which lowers the bill for high-volume users.
When does buying location APIs make more sense than self-hosting?
When you need accurate global addresses, live traffic, or rich points of interest. Those depend on data operations a self-hosted open tile server cannot replicate, and Radar and HERE now compete hard on price for exactly that data.
When does self-hosting the basemap make sense?
When your workload is heavy and mostly static: rendering tiles or checking whether a point falls inside a polygon. Protomaps and PostGIS run that workload without a per-call fee, which is where self-hosting saves the most money relative to the engineering cost of running it.
Does open-source routing and geocoding actually work at production scale?
Yes. Valhalla, OSRM, Nominatim, and Pelias power real applications, and Protomaps serves tiles for sites like The Washington Post. The gap with commercial providers is in places data, live traffic, and address coverage, not in the core routing or tile-serving technology.