Skip to content
Garkonix

IoT · 14 weeks to production

From 3,200 vehicles of raw telemetry to 90-second responses

A real-time telematics platform ingesting data from 3,200 vehicles, turning raw pings into routes, alerts, and maintenance schedules that operators act on in seconds instead of the next morning.

vehicles streaming in real time
3,200
vehicles streaming in real time
alert-to-action time, down from 45 min
90 sec
alert-to-action time, down from 45 min
reduction in fleet fuel spend
12%
reduction in fleet fuel spend

The challenge

Client
National equipment rental company
Services
Custom Software, Cloud Solutions, DevOps & Platform
Timeline
14 weeks to production

The telematics vendor's own portal technically had the data, buried five clicks deep and 20 minutes stale. Dispatchers ran the fleet from a spreadsheet exported every morning, which meant a vehicle idling all day, or driving off-route, got noticed tomorrow.

Meanwhile the devices were streaming thousands of events a minute into an inbox nobody could drink from. The data existed. The operating picture didn't.

What we did

The approach, decision by decision

  1. 01

    Built an ingestion pipeline that respects reality

    Devices go dark in tunnels and duplicate on reconnect. Idempotent event processing with out-of-order handling turned messy pings into a clean vehicle timeline.

  2. 02

    Alerts tuned to what costs money

    Idle time, geofence exits, harsh usage, and missed maintenance windows page the right dispatcher immediately. Everything else aggregates into the daily digest instead of the pager.

  3. 03

    One board per depot, live

    Each depot sees its own fleet on a live board with sub-second updates, built on the same event stream, so dispatch and maintenance finally share one picture.

The results

  • 3,200 vehicles streaming with the pipeline holding p99 ingestion under 2 seconds.
  • Alert-to-action time fell from 45 minutes to 90 seconds for high-cost events.
  • Fuel spend down 12% in six months, driven by idle-time alerts alone.
  • Maintenance moved from calendar guesses to engine-hour triggers, and breakdowns followed.

Built with

  • Go
  • Kafka
  • TimescaleDB
  • Grafana
  • OpenTelemetry
  • AWS

Your project could be the next case study

Tell us the metric you need to move. We'll scope the build that moves it: fixed price, weekly demos, results you can publish.

We reply within 1 business day.