Turning journey data into fraud detection on AWS

A SaaS B2B business working with Cloud Combinator on AWS. The client is anonymised at their request.

The challenge had three focus areas, all centred on measuring reported journeys against reality at fleet scale.

Journey data from the platform and FieldPower lands in an Amazon S3 bucket. Each arrival triggers an AWS Lambda that assembles a complete journey, then checks it against the misuse criteria. It takes the start and end points, asks a maps API how long the trip should take, allowing for traffic and a 20% tolerance, and compares that to the reported time. The same route logic flags journeys that miss the destination or run well beyond the ideal route.

The proof of concept delivered all five functional checks, verified within the client's own AWS environment. Reported journeys are now measured against live and historical route data, misuse is logged with a full audit trail, and the operations and payroll teams are alerted automatically.

• Automated fraud detection — Overreported hours are flagged by comparing the reported time to a live maps estimate with a 20% leniency tolerance.
• Route and misuse checks — Suspicious routes, excessive idling, private use, and start-time and mileage discrepancies are all detected without manual review.
• An audit trail and driver scoring — Every flag is logged in DynamoDB, laying the foundation for monthly driver compliance scores and performance reviews.
• Alerts where they matter — SNS notifications reach operations and payroll the moment misuse is detected, alongside a daily per-driver summary.
• Handover — The solution was deployed and tested in the client's own AWS environment, ready for the operations team to validate against known cases.

With the detection engine proven, the client has the foundation to turn per-trip flags into full driver compliance scoring, extending the same event-driven pattern across the wider fleet.