All case studies
HR tech · Geospatial
Geolocation Attendance at Scale
A geocoding cache that absorbs the morning check-in spike instead of paying a third-party API for it.
60%lower geocoding cost
The challenge
Employee attendance reverse-geocoded every check-in through Google. At the morning peak of 20,000–30,000 requests per second, third-party costs became expensive.
How it works
- 01
Geocode microservice
The attendance service calls a dedicated Go geocode service, which runs a PostGIS ST_DWithin query for a cached address within a set radius, such as 100 m.
- 02
Cache on miss
On a hit, the cached address is returned immediately. On a miss, the service calls Google or HERE, stores the result as a geography point and returns it.
Engineering decisions
Proximity, not exact match
- An address within 100 m is accurate enough for attendance, so nearby check-ins share a single paid lookup.
Built for the spike
- Both services read from database replicas. Load-tested with k6 to 50,000 requests per second.
Outcomes
- 60% reduction in third-party geolocation costs.
- Handles 20–30k requests per second at peak, validated to 50k.
- Provider fallback across Google and HERE.
Technology
- Go · Chi
- Geocode service
- PostgreSQL · PostGIS
- Geospatial queries
- sqlx
- Database access
- Docker
- Containers
- k6
- Load testing
Contact
Have something you want to build?
Tell us what you're building, or what you'd like to automate. Let's figure out the next step.