Fuel Monitoring & Theft Prevention

Almost every argument about fuel monitoring accuracy is really an argument about which measurement method is in use.

A capacitive or ultrasonic sensor in the tank measures the level directly and continuously. It is accurate enough to see a siphoning event as it happens, and it requires fitting into the tank, which means downtime and a calibration step.

Fuel level read from the vehicle's own gauge over the diagnostic port needs no tank work, and is coarse. Most factory senders were designed to move a needle, not to detect fifty litres leaving over twenty minutes.

So the choice is real: the port is cheap and tells you roughly how much fuel there is, the sensor costs a fitment and tells you when it left. If the question is theft, only one of them answers it. See fuel sensor versus OBD monitoring.

It does not look like a gradual decline. It looks like a step.

The signature is a sharp drop in level while the vehicle is stationary and the engine is off, usually at night, usually at a predictable halt. Consumption while driving produces a slope; a siphon produces a cliff.

The second signature is the mirror image at a filling station: a refuel event whose volume does not match the invoice. Short-filling is quieter than siphoning and often larger over a year.

Both are only visible if level is sampled often enough to see the event between two readings. A system reporting fuel every fifteen minutes can miss a drain entirely, which is why reporting interval matters as much as sensor accuracy.

Most fleets start by watching kilometres per litre and conclude that theft is not happening, because the number looks normal.

Mileage is an average over a tank, and it absorbs almost anything. Load weight, terrain, traffic, idling, tyre pressure and driving style all move it by more than a routine theft does. A driver taking twenty litres from a three-hundred-litre tank moves mileage by a rounding error.

It is also computed after the fact, so even when it does shift there is no way to attribute the change to an event, a place or a person.

Level data attributes it. A drop at a specific time and place, with the ignition off, is an event you can point at. See how to detect fuel siphoning.

This is the step that gets rushed and the one that determines whether the readings mean anything.

Tanks are not simple shapes. The relationship between fuel volume and sensor reading is not linear across the range, so a sensor has to be calibrated against known volumes at multiple points to build a curve for that tank.

Calibrate badly and the errors are worst where they matter most, usually near empty and near full, which is exactly where refuel and drain events sit.

It also has to be redone if the tank is changed or the sensor is refitted. A fleet that calibrated once at installation and never again will drift, and the drift will be blamed on the software.

Fuel monitoring dies the same way camera alerts die: too many alarms, no triage, and within a month nobody opens them.

The common false positives are worth designing out rather than explaining away. Fuel sloshing on rough roads produces short-lived swings that are not drains. Parking on a slope shifts the level in a tank permanently until the vehicle moves. Temperature changes volume without changing mass. And a refuel recorded slowly can look like an anomaly.

The fix is thresholds tied to conditions rather than to a single number: require the ignition off, require the level to stay down after movement resumes, and require a volume large enough to matter. Fewer, truer alerts get acted on.

An alert is not evidence, and a driver accused on the basis of one will contest it, reasonably.

What holds up is the combination: the level trace showing the drop, the time and duration, the location with the vehicle stationary and the engine off, and the pattern showing whether this is a first occurrence or the fourth at the same dhaba.

Pair it with the refuel record and the invoice, because short-filling and siphoning look different and are addressed differently, one with the pump and one with the driver.

The point of the record is not to win an argument after the fact. It is that halts stop being anonymous, which is what changes the behaviour. You can put your own numbers through the fuel theft calculator.