Most fuel dashboards produce numbers nobody can defend when a driver disputes them. This one is built around calibration, evidence and the difference between a dip in the tank and a dip in the sensor.
| Method | What it measures | Best for | Weakness |
|---|---|---|---|
| Capacitive probe | Actual level in the tank | Siphoning and refuel shortfall | Needs calibration and tank access to fit |
| CAN consumption | Fuel the engine burned | Driving efficiency, consumption per trip | Cannot see fuel removed from the tank |
| Analogue gauge input | The vehicle’s own float gauge | Rough trend only | Coarse, non-linear, easily disputed |
| Probe + CAN together | Both, and the gap between them | Fleets where fuel loss is material | Two sensors to fit and maintain |
| Fuel card reconciliation | What was purchased | Catching purchase-side fraud | Says nothing about what reached the tank |
A fuel monitoring deployment usually fails in one of two ways, and neither is a software problem. The first is skipped calibration. The second is treating a single sensor reading as proof.
A fuel probe reports a level, not a volume. Converting one to the other depends on the shape of the tank, and vehicle tanks are irregular — a saddle tank on a truck holds very different litres per millimetre at the bottom than in the middle. Without a multi-point calibration table built by filling the tank in measured steps and recording the reading at each, every litre figure the system produces is an approximation dressed up as a measurement.
This is tedious, it takes a technician a couple of hours per tank shape, and it is the single highest-return step in the entire deployment. Fleets that skip it spend the next year arguing about whether the system is accurate, which is a debate they cannot win because it is not.
A tank probe can see fuel leaving. It cannot tell you whether it left through the engine or through a hose. CAN consumption data can tell you what the engine burned, but sees nothing of what was siphoned. Run both and the arithmetic becomes conclusive: fuel that left the tank and did not pass through the engine went somewhere, and the time and position of the discrepancy tells you where.
This is why serious fleets fit both, and why our Ruptela and Teltonika guides both cover probe and CAN options rather than treating them as alternatives.
Stationary theft is straightforward: the vehicle is parked, the level falls, nobody is driving. Detection during movement is where most systems produce false positives, because fuel sloshes, gradients tilt the tank, and a probe reading taken on a roundabout is not a measurement. The platform filters against speed, acceleration and heading change, and requires a sustained deviation rather than a single sample — which means fewer alerts, and the ones that arrive are worth acting on.
An alert nobody believes gets switched off within a month. Tuning for precision over recall is deliberate: a system that reports three real incidents a quarter changes behaviour, while one that reports thirty possible incidents a week trains everyone to ignore it.
When a fuel event is raised, the useful artefact is not a chart. It is a record that says: this vehicle, this driver on shift, this time, this position, this much fuel, with the route playback either side of it. That is what survives a disciplinary meeting or an insurance claim, and it is why every fuel event on the platform is linked to position, driver identification and playback rather than living in a separate report.
Fuel loss is usually the fastest-paying part of a telematics deployment, and the easiest to get wrong. Book a demo with your own numbers or message us on WhatsApp — tell us your fleet size, tank types and current loss estimate and we will tell you what is realistically recoverable.
With a properly calibrated capacitive probe, accurate enough to detect a theft of a few litres and to defend the number in a dispute. With an uncalibrated probe or an analogue gauge input, accurate enough to show a trend and nothing more. The sensor choice matters less than whether calibration was done properly.
Only if you measure both sides. A tank probe sees fuel leaving; CAN consumption sees fuel burned. Fit both and the gap between them is the answer. With a probe alone you see a drop and infer the cause from context — position, whether the engine was running, time of day — which is usually enough but occasionally arguable.
Almost always because it is sampling rather than filtering. Fuel sloshes under acceleration and tilts on gradients, so a single reading taken mid-corner is noise. Requiring a sustained deviation and filtering against speed and heading change removes most of it. Systems that alert on raw samples get switched off, which is worse than having no system.
For a few hours, yes — a capacitive probe is fitted into the tank and calibration means filling in measured steps. Plan it into a service window. CAN-based consumption data needs no tank work at all, which is why some fleets start there and add probes only on the vehicles where siphoning is suspected.
At minimum, longer than your dispute and insurance window — an event you cannot produce six months later is an event you cannot act on. Tell us the period you need and it is sized into the hosting plan; retention is one of the three numbers that determines server size.
A demo against your vehicles and your numbers is more useful than a sample dataset. Tell us your fleet mix and we will set it up.
Book a Live Demo →