This is a story about data that was there all along — sitting in fuel card systems, telematics platforms, maintenance logs, and fleet databases — but nobody had ever connected it together.
Once we did, the findings were staggering. Not because the data was complex, but because the problems it revealed had been invisible for years.
The situation
The client was a main contractor running a fleet of several hundred vehicles — vans, trucks, plant transport, and company cars — supporting projects across multiple regions.
Like most contractors with large fleets, they had data everywhere. Fuel card transactions in one system. Telematics in another. Maintenance records in a third. MOT history, service logs, mileage reports — all sitting in separate databases and spreadsheets, managed by different teams, with no unified view.
Nobody was doing anything wrong. Each system worked fine on its own. But nobody could see across all of them at once. And that's where the money was hiding.
✕ Before
- Fuel data in one system, never cross-referenced
- Maintenance logs scattered across spreadsheets
- No way to flag anomalies or trends
- Fleet risk assessed manually — if at all
- Reporting was retrospective and slow
✓ After
- All transport data connected into one BI layer
- Automated dashboards refreshed daily
- Anomaly detection flagging issues in real time
- ML-powered risk scoring for every vehicle
- Savings identified within weeks of going live
What we built
The engagement had two phases: an automated fuel and transport reporting dashboard, and an ML-powered vehicle risk scoring system.
Phase 1: Fuel & transport intelligence
We connected the client's fuel card data, telematics system, vehicle registry, and HR records into a single reporting layer. Then we built automated dashboards that surfaced insights nobody had ever seen before.
Within the first few weeks of the dashboards going live, three major findings emerged:
Premium fuel abuse
Dozens of drivers were regularly filling up with premium fuel instead of standard diesel. On a fuel card, the cost difference doesn't jump out — but across hundreds of vehicles over a full year, the markup was enormous. The company's negotiated discount only applied to standard fuel, so every premium fill-up was costing significantly more per litre than it should have.
Fuel card fraud — personal vehicle fills
By cross-referencing fuel card transactions with vehicle telematics and registered tank capacities, we identified cases where fuel volumes exceeded what the assigned vehicle could physically hold. The conclusion was clear: some employees were using company fuel cards to fill personal vehicles. The dashboard flagged every anomalous transaction automatically.
Vehicles with poor fuel efficiency
By calculating actual MPG for every vehicle in the fleet — using fuel volume data against mileage — we identified a cohort of vehicles running significantly below expected efficiency. Some had mechanical issues. Some were being driven inefficiently. Some were simply old and due for replacement. Each one was costing the business more per mile than it should have been.
None of these problems were visible in any single system. It was only by connecting fuel card data with telematics, vehicle records, and tank capacities that the patterns emerged. The data existed all along — it just needed joining up.
Phase 2: ML-powered vehicle risk scoring
With the transport data now connected, we built a second layer: a machine learning model that evaluated the risk profile of every vehicle in the fleet.
The model pulled data from maintenance records, MOT history, service logs, telematics, and vehicle age to assign each vehicle a risk score on a points-based system. The more risk factors present, the higher the score — and the more urgently the vehicle needed attention.
The risk dashboard gave the fleet management team something they'd never had: a prioritised action list. Instead of reactively dealing with breakdowns and MOT failures, they could proactively target the highest-risk vehicles first — reducing downtime, avoiding compliance issues, and extending fleet lifespan.
The model also improved over time. As maintenance actions were logged and outcomes recorded, the scoring became more accurate — learning which combinations of factors were most predictive of breakdowns and failures.
The results
Premium fuel elimination
Enforced standard fuel policy across fleet
Fuel fraud detection
Stopped personal vehicle fills on company cards
Efficiency improvements
Addressed poor-MPG vehicles and driving behaviour
Proactive maintenance
Reduced breakdowns, compliance risk, and downtime
Total annual impact
The bigger lesson
This project wasn't about buying new technology or overhauling systems. It was about connecting data that already existed and presenting it in a way that made problems visible.
The premium fuel issue had been costing the business for years. The fuel fraud had been happening undetected. The at-risk vehicles had been quietly accumulating maintenance debt. None of it was malicious. None of it was obvious. It was simply invisible — because nobody had joined the dots.
The most expensive data problem isn't missing data — it's data that exists but nobody can see.
That's the pattern we see in every engagement. The data is there. The insights are there. The savings are there. They're just trapped in disconnected systems, waiting for someone to connect them.
This particular example was transport and fleet. But we see the same thing with project cost data, subcontractor management, labour tracking, and procurement. Different systems, same problem: the answers are in your data, if you know how to look.
How we did it — the timeline
Data audit & connection
Mapped all transport data sources, established connections to fuel card system, telematics, maintenance logs, and vehicle registry. Identified data quality issues and cleaned up inconsistencies.
Dashboard build
Built the fuel intelligence dashboard with anomaly detection, MPG analysis, and automated cost breakdowns. Configured automated daily data refreshes.
Risk model development
Developed and trained the ML-powered vehicle risk scoring system. Calibrated the points-based model against historical breakdown and maintenance data.
Rollout & embedding
Trained the fleet and finance teams. Built dashboards into weekly review meetings. Set up automated alerts for high-risk vehicles and anomalous fuel transactions.
Eight weeks from kickoff to live dashboards generating real savings. No system replacements. No multi-year transformation programme. Just clarity, where before there was chaos.
What's hiding in your data?
Every contractor we work with has savings and insights trapped in disconnected systems. Let us show you what connecting your data could reveal.
Book a Free Data Audit →