Your GPS data already exists — make it useful
An existing GPS platform collected 50,000 messages a day from the fleet, but its built-in reports didn't work. We built an analytics layer on top — anomalies, reports and scores over data the company was already paying to collect.
- 50,000 messages a day processed
- 18M historical records
- Reports in milliseconds instead of minutes
The client
A transport company with a fleet of heavy trucks, each equipped with a GPS tracker and 9–11 sensors: fuel level, engine temperature, RPM, battery voltage, ignition.
The challenge
The data had been collected for years — around 50,000 messages a day — but produced no value. An audit of the GPS platform's built-in reports showed only 3 of 8 templates returned any data at all, and most modules sat empty and unused. The company was paying to collect data nobody could read.
Management's questions were simple: where is fuel leaking, which machines overheat, where do trucks lose hours waiting, which route is unreliable. The answers existed — buried in the raw data.
The approach
- Build on top, don't replace: the GPS platform stayed — it does its collection job well. Our system reads from it (strictly read-only), normalizes the raw stream into its own database and computes everything itself instead of relying on the broken built-in reports.
- Anomalies with explanations: rules for fuel drops, speeding, long idling with the engine running, overheating, weak batteries, silent devices. Every anomaly has a page with a map, a sensor chart around the event and an explanation of why it fired — verifiable facts, not a black box.
- Reports that open instantly: heavy analyses are precomputed periodically in the background, so a page opens in milliseconds — instead of chewing through millions of records on every view.
- Engineering the costs: enrichment with external data (addresses, weather, traffic) was designed so the monthly bill for external APIs is effectively zero — instead of hundreds of dollars a month with the naive approach.
The result
- 12 months of history — about 18 million records — loaded and available for analysis; previously effectively inaccessible.
- Data-quality scores per vehicle and for the whole fleet, reports on idle time, route reliability, vehicle health and driving style.
- Reports in milliseconds over data that previously took minutes to process — or didn't exist at all.
- All of it on top of already-paid-for infrastructure — no new devices, no provider switch, no interruption.
Most companies with GPS tracking don't suffer from a lack of data — they suffer from a lack of answers. The difference between the two is software.
Technology
.NET 10 · React · MariaDB · telematics API integrations · maps and charts