Fuel Monitoring That Survives an Argument.

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.

Fuel Monitoring in Practice

📉

What gets detected

  • Sudden drops while stationary
  • Drops during movement, which are harder
  • Refuels smaller than the receipt
  • Consumption drift against a baseline
  • Tank readings that disagree with CAN
🔧

How it is measured

  • Capacitive probe in the tank (RS232/RS485)
  • CAN consumption from the engine
  • Analogue gauge input as a last resort
  • Both probe and CAN on fleets that matter
  • Divergence between the two is itself a signal
📐

Calibration

  • Tanks are not cuboids — litres per millimetre varies
  • Multi-point calibration table per vehicle
  • Filled in steps, recorded, not estimated
  • Re-checked after tank or sensor work
  • Uncalibrated data is why fuel projects fail
🧾

Evidence

  • Event linked to time, position and driver
  • Route playback around the event
  • Exportable for an HR or insurance process
  • Retention set by your dispute window
  • Alert and evidence are the same record

The Decisions That Actually Matter

Fuel measurement methods compared
MethodWhat it measuresBest forWeakness
Capacitive probeActual level in the tankSiphoning and refuel shortfallNeeds calibration and tank access to fit
CAN consumptionFuel the engine burnedDriving efficiency, consumption per tripCannot see fuel removed from the tank
Analogue gauge inputThe vehicle’s own float gaugeRough trend onlyCoarse, non-linear, easily disputed
Probe + CAN togetherBoth, and the gap between themFleets where fuel loss is materialTwo sensors to fit and maintain
Fuel card reconciliationWhat was purchasedCatching purchase-side fraudSays nothing about what reached the tank

Why most fleet fuel projects produce numbers nobody trusts

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.

Calibration is the whole job

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.

One sensor is an alert; two are evidence

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.

Detecting a drop during movement

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.

What the driver conversation actually needs

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.

Questions Fleet Managers Ask

How accurate is fuel monitoring in practice?

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.

Can the system tell the difference between fuel used and fuel stolen?

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.

Why does my current system raise so many false fuel alerts?

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.

Do I need to take the vehicle off the road to fit a probe?

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.

How long should fuel event data be retained?

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.

See Fuel Monitoring on Your Own Fleet.

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