Logistics telematics, consolidated. Four feeds. One operational view.
A regional logistics carrier ran telematics from four OEMs, three TMSs, and an old historian nobody would touch. The dispatch team saw a fraction of the signal. We consolidated the feeds onto one platform, rebuilt the dispatch tooling, and moved reporting off the legacy stack.
-
40%
Reduction in median latency from signal to assignment.
-
4 OEMs
Plus 3 TMSs and the legacy historian retired.
-
32%
Lower than the legacy stack it replaced.
What was hard.
How we worked.
-
Source and shape audit
Two weeks mapping every feed, every downstream consumer, every report anybody actually used. Fourteen reports retired in that audit alone.
-
Target platform design
Snowflake chosen over Databricks based on the team's existing skills and the query shape. Ingestion through Fivetran and Kafka. Modeling in dbt.
-
Unified telematics model
One canonical model across OEMs, normalized units, late-arriving data handled, fault codes reconciled. Dispatch tooling rebuilt on top.
-
Migration and retirement
Legacy historian read-only, then quiet, then archived. Reporting moved to Looker. Old dashboards rebuilt once against the new model.
-
Day-two handoff
Runbooks, monitoring, and a quarterly managed-services arrangement for the data platform. Cost governance reports in the first monthly cycle.
What we shipped.
- Dispatch team working from one screen instead of four.
- Legacy historian decommissioned six weeks ahead of plan.
- Warehouse costs trending down every month for the last three quarters.
- One on-call rotation, one escalation path, one team.
Who shipped it.
- 14 weeks (project)
- 5 people
- Week 14, runbooks signed
- Quarterly managed retainer