How to detect fuel theft in your fleet by volume, time, and location

Why fuel theft goes unnoticed

In most fleets, fuel is the first or second largest operating cost after payroll — yet it is often managed with the weakest controls: station invoices collected at month end, a spreadsheet dividing liters by kilometers, and the intuition of an experienced supervisor. These tools catch only flagrant anomalies, such as a vehicle consuming twice as much as its peers. A disciplined drain of 20 or 30 liters every few days from a 400-liter tank sits comfortably inside the variance created by loads, routes, and driving styles.

The second problem is discovery lag. Even when the monthly reconciliation does flag a deviation, weeks have passed. Nobody knows which day the theft occurred, at which location, or who was with the vehicle at the time. The confrontation turns into mutual guesswork, usually ends with no action taken, and the scene repeats the following month. The comparison below shows the gap between approaches:

The four common theft patterns and their data signatures

Before discussing tools, it helps to know exactly what you are looking for. Four patterns recur across fleets, and each leaves a different trace in the data:

• Direct siphoning: a hose and a container during an overnight stop or in a yard away from cameras. Signature: a sharp drop in tank level while the vehicle is stationary with the engine off.
• Partial refills through collusion: the company pays an invoice for 400 liters while only 300 enter the tank, and the difference is split in cash. Signature: the measured increase in the tank is consistently smaller than the invoiced volume, by a suspiciously regular margin.
• Invoices with no refill at all: a receipt from a station the vehicle never visited. Signature: no refill event in the tank history at the receipt's timestamp, and the vehicle's position far from that station.
• Off-hours use: not liquid theft but operational theft — private trips burning company fuel and vehicle life. Signature: movement and consumption inside time windows where no work is scheduled.

Notice that a system recording one monthly "consumption rate" cannot distinguish between any of these, while a continuous level curve tied to time and location separates each pattern clearly and dictates how to respond to it.

The volume signal: from raw readings to trustworthy liters

The starting point is a fuel-level sensor mounted in the tank — not the dashboard float, which was designed to reassure drivers, not to measure liters. The sensor's raw reading is just a number; converting it to real liters requires a calibration table built for that specific tank. Tanks are irregular in shape, and in most designs 10% of tank height does not equal 10% of tank volume.

Then comes noise smoothing. Fuel sloshes with every turn, braking event, and slope, and an instantaneous reading can jump up and down dozens of times in a single trip. Without smoothing, the log fills with false events that bury the team in meaningless alarms until they ignore all of them — including the real one. A good algorithm separates three curve shapes: a sharp rise followed by a plateau is a refill; a sharp drop that neither distance travelled nor engine runtime can explain is a drain; a gradual slope correlated with movement is normal consumption.

On the Pixa platform, every reading passes through this full chain — raw value, calibration table, smoothing, then classification as a refill or drain event with its volume — and the result appears on the fuel screen as a single clean curve with event markers. The platform supports 18 tested device protocol families today, from Teltonika and Queclink to fuel-metering specialist Omnicomm; compatibility details are on the devices page.

Time and location turn a signal into evidence

The drained volume alone proves that fuel went missing; time and location prove how. A 60-liter drain event at 2:40 a.m., on a vehicle that has been parked since evening in a yard that is not its approved overnight location — that is a complete narrative, not just an outlier in a spreadsheet.

In practice, that context is built with two instruments from the tracking toolbox itself:

• Geofences: circular or polygonal zones around approved fuel stations and overnight parking locations. A fuel event inside them is routine; one outside them deserves a look — and a measured refill with no matching invoice from an approved station deserves a harder look.
• Time windows: rules that watch fuel events during quiet hours — say midnight to dawn — with far higher sensitivity than during normal operating hours.

Combined with live tracking, the supervisor can open the vehicle card the moment an alert arrives and see the current position, engine state, and recent trips — before picking up the phone.

From instant alert to action

The real value is not in detecting the event but in getting it to the right person within minutes. Instant alerts are delivered over 5 channels — WhatsApp, SMS, email, app push, and browser notification — plus an in-platform notification centre, so the response never depends on a single channel being available.

Delivery, however, is only half the story; the other half is the quality of the rule that fires the alert. The rules engine accepts composite conditions instead of one naive trigger: a drop exceeding a defined threshold, and outside approved refuelling locations, and inside a sensitive time window. Three operational safeguards come with it:

• Escalation: if the supervisor does not acknowledge the alert within a set time, it escalates automatically to their manager.
• Acknowledgement tracking: a record of who saw the alert and when they acted, so no event dies in a crowded inbox.
• Flood control: grouping of near-simultaneous events, so one turbulent night does not become 40 consecutive messages.

When the alert lands, the response is straightforward: verify the position, call the driver or the site guard, and document the event with its volume, timestamp, and coordinates — evidence solid enough for internal accountability, and for whatever may need to follow it.

A practical rollout plan

You do not need to instrument the whole fleet at once. These steps deliver early results:

• Start with the heaviest consumers: that is where the spend concentrates and where control shows results fastest.
• Calibrate each tank individually: one calibration table per vehicle. A conical tank behaves nothing like a cylindrical or a twin tank, and a generic table corrupts every measurement downstream.
• Define your approved geography: geofences around the stations your fleet is supposed to use and around official overnight parking.
• Tune rules with realistic thresholds: begin with a conservative detection threshold that avoids false alarms, then tighten it as calibration cleans up the readings.
• Close the loop with reconciliation: compare measured refill events against station invoices every month; the gap between them is your true leakage figure. Fuel reports are part of 29 ready-made reports in Arabic and English, exported as PDF or Excel and schedulable by email — see the reports screen. On the roadmap (coming soon): PetroApp and SASCO fuel-card integration to close the reconciliation step automatically.

Suspect that part of your fuel bill is evaporating outside your tanks? Contact the Pixa team via the contact page for a working session on your own fleet data — with your current sensors, or with a device recommendation matched to your vehicles.