What
The first thing most people do when a loss shows up is blame the fuel level sensor. That instinct is understandable, but it is usually premature. In a stored fuel history, a discrepancy can come from two very different places: the sensor reading the tank inaccurately, or the timing between when refuelling actually happened and when the sensor reported it. Those two causes look almost identical on a report, and they need opposite fixes. One calls for calibration. The other calls for looking at your polling and sampling setup. Separating them before you spend anything is the whole job, and the good news is you can usually do it with data you already have.Start with timing, because it is free to check and it explains more mismatches than hardware ever does. Pull the fuel record for the suspect event and compare three things side by side: the timestamp of the refuelling event, the reporting cycle of the fuel level sensor, and the refuelling window at that site. If refuelling happened near the end of a reporting cycle, or if the vehicle left the bay before the next sample was taken, the volume may be real but recorded late. That gap shows up as a loss on one day and a phantom gain the next, and it has nothing to do with accuracy. A delayed position fix can hide a normal refuel too, especially when the refuelling bay sits inside a covered structure or an underground area where location updates lag.Next, check what the record actually contains. A usable fuel record carries a volume, a location, and a time in one place, and it stays exportable without raising a support ticket. If your suspect event is missing any one of those three parts, you are dealing with an event completeness problem, not a sensor problem. That distinction matters because no amount of calibration fixes a missing timestamp. Ask how bowser and drum refuelling are captured, since those events fall outside sensor-only records and often sit in a separate log or a driver note. When a depot reports a recurring unexplained loss, incomplete manual events are a common reason.Sensor side evidence looks different. Real accuracy trouble tends to creep in gradually rather than appear as one isolated loss. Calibration drift accumulates over time, so you would expect repeated small variances across many vehicles or the same vehicle across weeks, not a single clean event that vanishes. If your history shows a slow widening gap between fill volume and fuel level readings, calibration is worth investigating. If it shows one dramatic loss with complete data around it, timing deserves another look. Without a verifiable source, avoid quoting any specific accuracy figure or treating one sampling interval as universally correct, since the right interval depends on your fleet operations, tank size, and how often vehicles refuel.Before anyone proposes adding a second fuel level sensor, fix calibration and thresholds first. An uncalibrated extra sensor usually doubles ambiguity rather than removing it, and the added hardware plus installation raises your total monitoring cost while the underlying question stays open. That is also where monitoring cost control quietly turns into service risk. Trimming fine fuel level sampling or cutting data plans moves spend from a visible monthly line into maintenance work, driver disputes, and reconciliation hours that nobody tracks as a cost. A cheaper plan cannot close a record that was never complete.Run the check in order. Timing window first, comparing the refuelling event against the sensor reporting cycle. Then record completeness, confirming volume, location, and time land together. Then calibration and thresholds. Then pull a raw event export to confirm the finding end to end. Teams running GPS Controller devices alongside third-party monitoring often find the handoff, not either device, sets the limit on accuracy, so test that exchange during evaluation rather than after rollout. Because the record has to travel from device to report before any dashboard can show it honestly, a clean export matters more than another report screen.Related FAQs: How can you tell if your fleet is being spoofed versus just jammed?