The challenge
Orbit Labs sells dispatch software to small freight operators. Its own operations data sat across three systems — orders, GPS tracking and finance — in different formats and refresh rates, so the team assembled a global view by hand each morning.
Those reports described yesterday rather than now, and route delays were often discovered hours after they began. The team also wanted machine-learning alerts but doubted operations staff would trust a black box they could not interpret.
- Three systems with no shared data pipeline
- Manual reporting described yesterday, not now
- Unexplained alerts would not be trusted
- Late detection turned delays into customer complaints
What we built
An automated pipeline now brings order, GPS and finance data into a single data layer in real time, and a React dashboard presents the whole operation on one screen.
Anomaly detection runs against each route's historical baseline. When a route deviates, the dashboard highlights it with a plain-language note stating how far it has drifted and which customers are likely affected.
The AI was scoped as an explainable alert rather than an all-knowing model, and the system was delivered in phases.
- One real-time pipeline across orders, GPS and finance
- Full operation visible on a single dashboard
- Anomaly alerts with plain-language impact notes
- Phased delivery, explainability before sophistication
The result
Operations staff now see the whole network on one screen, and exceptions are raised within minutes instead of hours.
Because each alert states its likely impact, intervention rates rose and the share of complaints caused by delays fell.
The dashboard has also become a key asset in product demonstrations, including conversations with prospective customers and investors.
How the dashboard is maintained
Thresholds are version-controlled with the application code. When a service changes its expected behaviour, the threshold changes with it instead of drifting out of date in a separate configuration file.
Orbit Labs operates the dashboard internally. Handover covered the data pipeline, the alert configuration and the runbook used when an alert fires outside business hours.
Data handling
The dashboard reads operational data only. Customer records stay in the source systems, and the dashboard holds identifiers and aggregate metrics rather than personal information.
Access is role-based: operations staff see the live network view, account managers see the account-level view, and both are read-only.
