Vehicle Fuel Sensor Monitoring You Can Actually Trust
Every fuel report is only as good as the probe underneath it. KO Fleetz handles the unglamorous part: calibrating each tank to its real shape, reading capacitive, ultrasonic and CAN sources side by side, and telling you the moment a sensor starts lying.
The sensor nobody calibrated
The probes went in two years ago. The installer did a fill-and-record on a Tuesday afternoon, the numbers looked reasonable, and everyone moved on. Since then a tank has been replaced, three vehicles have had bodywork done, one probe was refitted at a slightly different angle, and nobody re-ran the calibration because nobody owned it.
Now the fuel screen shows a truck holding 412 litres in a 400-litre tank. The workshop says the probe is fine. The vendor says it is a calibration issue. The vendor's engineer who did the original install has left the company, and the calibration table lives in a proprietary tool on a laptop that also left.
The predictable outcome is that the data gets quietly ignored. Once a supervisor has been burned twice by a reading that turned out to be nonsense, they stop opening the fuel report altogether, and the fleet goes back to managing fuel by receipt. The hardware is still bolted to the trucks, still reporting, and worth nothing.
Own the signal, not just the dashboard
Calibration is a multi-point table, not a scale factor. Tanks are baffled, saddle-shaped and mounted at an angle, so the litres between 10 and 20 percent are rarely the same as the litres between 60 and 70. KO Fleetz stores a per-tank curve built from measured fills, keeps it versioned, and keeps it in the platform rather than on somebody's laptop.
Sources vary and that is fine. A capacitive probe, an ultrasonic head, a tank percentage broadcast over CAN and an OEM telematics feed all normalise to the same litre-denominated series, with the source recorded against each reading. Two-tank tractors get two curves and a combined total, because averaging them hides exactly the case you care about.
The system will not paper over a bad sensor. It does not smooth a suspicious channel into something presentable, and it does not fill missing minutes with interpolated values. A probe that reports a frozen constant, drifts against known fills, or goes quiet raises a health alert of its own, and the fuel data that depends on it is marked as unreliable until somebody fixes the hardware.
Capabilities
What KO Fleetz fuel sensor integration gives your team
Multi-point calibration tables
Each tank gets a measured curve across its full range, so a reading near empty is not extrapolated from behaviour near full.
Capacitive and ultrasonic probes
Both sensing types are supported with their own quirks accounted for, including the settling behaviour ultrasonic heads show after a fill.
CAN and OEM tank feeds
Where a vehicle already broadcasts tank percentage, that channel is used and labelled as coarse rather than dressed up as probe-grade data.
Multi-tank vehicles
Twin saddle tanks are calibrated and displayed individually as well as combined, so a cross-flow between them is not mistaken for a loss.
Sensor health monitoring
KO Fleetz raises engineering alerts on flat-lined channels, implausible jumps and silent probes, separately from anything the fuel data itself says.
Raw signal inspection
Engineers can open the underlying reading trace behind any chart, which is the difference between diagnosing a probe and arguing about one.
Mixed device estates
Probes and trackers from different vendors sit in one platform, so a hardware change on twenty vehicles does not fork your fuel history.
Calibration history
Every table is versioned with who changed it and when, so a step change in reported consumption can be traced to a recalibration.
How it works
How KO Fleetz does it
Step 1: Fit and wire the probe
The probe is cut to tank depth, sealed and wired to the telematics unit. Fitting angle and mounting point are recorded, because both change the curve.
Step 2: Run a measured fill
The tank is drained low and filled in known increments, with the raw reading captured at each step. Those pairs become the calibration table for that specific tank.
Step 3: Validate against real fills
The first weeks of refuelling are checked against card volumes. A probe that consistently under-reports a 200-litre fill by eight litres is corrected, not tolerated.
Step 4: Monitor the hardware
Health checks run continuously from then on. Drift, dropouts and dead channels surface as maintenance work rather than as mysterious fuel anomalies.
Outcomes
What changes
- Calibration curves, not one global scale factor
- Per tank
- Every table change is attributable
- Versioned
- Probe, ultrasonic, CAN and OEM in one series
- Any source
- Unreliable data is labelled, never smoothed
- Marked
Frequently asked questions
Capacitive probes and ultrasonic heads are the two common aftermarket types and both are supported, as are tank percentage channels published over CAN and tank data exposed through OEM telematics APIs. The platform is deliberately vendor-agnostic at the probe layer. What matters is that the device can deliver a raw reading at a usable rate; the calibration curve and the interpretation live in KO Fleetz rather than inside the hardware.
Because tanks are not rectangular boxes. They are saddle-shaped, baffled, welded around a chassis rail and mounted at whatever angle the body builder chose. A centimetre of level change near the bottom of a tank is a different number of litres than a centimetre near the top. A single factor produces readings that look fine in the middle of the range and are wrong exactly where you need them, which is near empty and immediately after a fill.
Not on a schedule — on an event. Recalibrate after a tank replacement, a probe refit, bodywork that disturbs the mounting, or any repair that drains the tank. Beyond that, let validation drive it: the platform compares reported fill volumes against card volumes over time, and a probe that starts drifting reveals itself in that comparison well before anyone notices something odd on a report.
Usually not. If the existing probes are healthy and the telematics unit can forward their readings, they can be brought in as a source and calibrated in the platform. Replacement is worth discussing in two cases: probes that are physically failing, and probes locked behind a vendor protocol that will not release the raw value. A closed device that only emits a pre-cooked number is the one genuine blocker.
Often, but not always, and it says which. A probe failure has signatures — a reading that jumps to zero and stays there, a value that never moves while the engine runs, a channel that returns implausible litres for the tank size. Those raise a hardware alert. The harder case is slow drift, which looks like gradually worsening consumption. That is why fill validation runs continuously: it catches drift that no single reading would reveal.
For accounting and consumption reporting, a reading every minute is comfortable. For catching a fast drop while it is happening, you want a reading every few seconds, because a siphon can empty a substantial volume inside the gap between two sparse samples. Devices reporting only at trip start and trip end can support the fuel ledger, but they cannot support event-level detection, and the platform will not pretend otherwise.
Get a fuel signal your team believes
Send us a week of raw readings from any probe you already have, and KO Fleetz will tell you what the calibration is doing wrong.