ARXAX PAY: OUR APPROACH TO REFUELING SECURITY

How we protect fleets from fuel loss through data, telematics, and statistical intelligence

Fuel is one of the most significant cost items for transport companies and one of the most exposed to inefficiencies, errors, and fraud.

Anomalous refueling events, fuel diversion to external tanks, duplicate transactions: these are real scenarios that erode fleet operating margins and often remain invisible in traditional management systems.

At Arxax, we have designed a native anti-fraud system, built into our fuel payment and management platform, that operates across multiple layers and in real time.

A three-source data approach

Our system automatically cross-references three independent data sources for every refueling event:

• Vehicle data: technical specifications from the fleet registry (tank capacity, expected fuel type).

• Telematics data: fuel levels and vehicle GPS position, via integration with telematics providers.

• Transaction data: quantity dispensed, station, time, and driver, from our Arxax Pay platform.

Cross-referencing these sources allows us to calculate expected consumption, verify consistency, and identify significant deviations.

Real-time validation

At every refueling event, the system performs a series of automated checks:

• Location verification: comparison between the driver's phone GPS, vehicle GPS, and station coordinates. Any discrepancies trigger immediate alerts.

• Fuel type consistency: the fuel dispensed is compared against the expected fuel type for the vehicle, preventing errors or manipulation.

• Quantity validation: the volume dispensed is verified against tank capacity. Quantities exceeding capacity or otherwise anomalous are flagged automatically.

• Duplicate detection: multiple refueling events on the same vehicle within a short time window are identified and flagged.

Post-refueling validation with telematics

After every refueling event, the system uses telematics data to verify that the fuel level in the tank actually increased by an amount consistent with the quantity dispensed. This check is essential for detecting fuel diversion scenarios that would not be visible from transaction data alone.

To ensure the reliability of these analyses, we continuously monitor the quality and freshness of telematics data, distinguishing between genuine anomalies and data gaps.

Statistical sensor calibration

One of the less obvious aspects is that fuel level sensors and declared tank data are not always reliable. We have developed statistical calibration algorithms that:

• Correct non-linear sensor behavior (such as readings that plateau at minimum levels)

• Estimate the actual tank capacity of each vehicle based on historical refueling data, separating calibration errors from genuinely suspicious anomalies

This drastically reduces false positives and increases the accuracy of alerts.

Long-term analysis: vehicles and drivers

Beyond event-level checks, the system monitors consumption patterns over time:

• Per vehicle — weekly consumption (liters/100 km), refueling frequency, tank utilization. Vehicles that deviate statistically from the fleet average for their model and type are flagged.

• Per driver — refueling frequency, average quantities, anomaly rate. 

Thresholds are based on standard deviations from the fleet average, not rigid rules, making the system adaptive and resistant to false alarms.

Alerting and operational management

Critical anomalies trigger immediate notifications to the operations team, with all the details needed for a rapid investigation. A dedicated dashboard allows filtering, tracking, and resolving each alert, monitoring trends over time, and analyzing the behavior of the fleet and individual drivers.

Why it matters

For transport companies, an effective anti-fraud system is not just a matter of security — it is a competitive advantage. Reducing fuel loss means protecting margins, increasing transparency toward clients, and building trust across the entire logistics chain.

Our goal is not to generate alarms, but to provide reliable and actionable signals — distinguishing with precision between a sensor error, a legitimate operational pattern, and genuinely suspicious behavior.