Fuel Theft Investigation Checklist
Most fuel theft investigations fail in the first hour, because they start with a suspect instead of a signal. This checklist runs the other way: prove the loss is real, eliminate every innocent explanation, and only then work out who was holding the hose.
The moment a fuel report shows a gap, the room fills with theories. Someone names a driver. Someone else names a depot. Within a day the investigation has an accused and is looking for evidence to support a conclusion it already reached, which is the fastest way to lose an unfair dismissal case and a good driver in the same week.
The discipline is boring and it is the whole job. A loss on a report can be a probe drifting out of calibration, a tank replaced without recalibration, fuel that was never delivered despite being paid for, a coding error that put one vehicle's fill against another, or genuine siphoning. Four of those five are not theft, and three of them are your own process failing.
Work the list in order. Each section eliminates a class of explanation, and by the time you reach the last one you either have an event that nothing innocent explains, or you have found a problem worth more than the fuel. Both outcomes are wins. Skipping to the last section is how you get neither.
39 checks across 6 sections
Prove the loss before anything else
- Establish the exact volume and the exact window. 'Roughly forty litres sometime Tuesday night' is not an event; it is a feeling with a number attached.
- Pull the raw sensor trace for the period, not the summarised chart. The shape of the curve is the evidence, and summarisation destroys exactly the detail that matters.
- Check the sensor's health record either side of the event: dropouts, flat lines, implausible jumps or a channel that has been quietly reporting nothing.
- Confirm when this tank was last calibrated and whether anything since then would invalidate it — a tank swap, a probe refit, bodywork, or a repair that drained the tank.
- Compare reported fill volumes against card volumes on this vehicle over the past several fills. A probe that under-reports every fill is drifting, not being robbed.
- Check whether the same apparent loss appears on other vehicles on the same device type or the same firmware. A fleet-wide pattern is an engineering fault.
- Verify the vehicle's tank configuration. On a twin-tank tractor, cross-flow between saddle tanks reads as a loss on one and a gain on the other if you only look at one.
Eliminate the physics
- Check the vehicle's attitude at the time. A truck parked nose-down on a slope reads differently from the same truck on the flat, and it reads differently again when it drives away.
- Look for the level returning once the vehicle moved or settled. Fuel that comes back was never gone; it was somewhere else in the tank.
- Account for temperature. Diesel expands as it warms and contracts overnight, and a cold morning after a hot afternoon produces a level drop with nobody present.
- Rule out a leak before ruling in a thief. Check under the parking spot, check the fuel filter housing, check the return line, and ask whether the vehicle smelled of diesel when it came in.
- Confirm whether the engine was running. Fuel consumed by an idling engine, a hydraulic pump or an auxiliary heater is consumption, and a cab night heater will burn a real volume overnight.
- Check whether a power take-off, refrigeration unit or crane was in use, and whether it draws from the same tank.
Eliminate the paperwork
- Match every card transaction in the window to a tank rise on the vehicle named on the receipt, and treat a transaction with no matching rise as an open question rather than a fill.
- Check for fuel bought for one vehicle and put into another. This is common, usually innocent, and produces a perfect phantom loss on the vehicle that was never filled.
- Reconcile depot bowser issues against the logbook and against the tank rise. Own-site fuelling is where records are thinnest and where investigations most often stall.
- Test the bowser meter itself. A pump that dispenses eighteen litres while recording twenty produces a loss on every vehicle it touches and looks like a fleet of thieves.
- Check whether a manual odometer or fuel entry was keyed with a transposed digit. A single fat-fingered reading generates a consumption anomaly that survives three meetings.
- Confirm the vehicle was not filled from a jerrycan or a second source that never appears in any system.
Build the event picture
- Place the event on a map and establish what was at that location: a yard, a layby, a customer site, a residential street, or somewhere with no reason to be visited.
- Establish whether the vehicle was there as planned. An unscheduled stop of the right duration in the right place is the single strongest signal you will get.
- Establish who had the vehicle. Key issue, trip assignment and shift rota all need to agree, and where they disagree, that disagreement is itself worth investigating.
- Check for a second vehicle at the same location at the same time, particularly one from your own fleet.
- Pull any camera footage before the retention window closes. Most systems overwrite within days, and this is the step people remember on day nine.
- Look for repetition rather than a single incident. The same location, the same weekday, the same shift, across several weeks, is a pattern; one night is an anecdote.
- Check the fuel cap, the anti-siphon device and the tank for physical evidence: scratching, a broken seal, a lock that turns too freely.
Handle the people part properly
- Involve HR before the first conversation, not after it. The interview you have already had is the one you cannot undo.
- Ask the driver to explain the stop before you tell them what you think happened. An innocent explanation offered freely is worth more than one extracted after an accusation.
- Show the data and let it be questioned. A driver who is allowed to challenge the sensor and the calibration is a driver who will accept the conclusion when the challenge fails.
- Record what was said, when, and by whom, in a form that would survive being read out at a tribunal.
- Keep the sensor trace, the location record, the card transactions and the calibration history together as one evidence bundle for the event.
- Decide the threshold for police involvement before you need it, and apply it consistently. Reporting one driver and not another for the same conduct will cost you more than the fuel did.
- Do not discuss an open investigation across the depot. Fleets lose cases on the informal conversation, not on the evidence.
Close it out and stop the next one
- Record the outcome against the event: theft, leak, sensor fault, paperwork error, or explained. An investigation with no recorded outcome teaches the system nothing.
- Fix whatever the investigation exposed even when the answer was not theft, because a drifting probe or an uncalibrated bowser will manufacture the next false accusation.
- Tighten the alert threshold using what you learned, so the same signature raises an event sooner next time.
- Fit anti-siphon devices or locking caps where the pattern justifies it, and record which vehicles have them so the next investigation knows.
- Review the parking policy for the location involved. Overnight parking on an unlit street is a decision somebody made, and it can be unmade.
- Tell the fleet what happened in general terms when a case is closed. Silence after an investigation is read as either weakness or injustice, and both are expensive.
How to use this checklist
- Start it the day the anomaly appears, not the week the report is reviewed. Camera footage, driver memory and the physical state of the fuel cap all decay fast, and none of them come back.
- Give it one owner. Fuel investigations that are 'being looked at' by three people are being looked at by nobody, and every one of the three assumes another pulled the footage.
- Work the sections in order even when you are certain. The certainty is the risk: the cases that end badly are the ones where somebody knew who did it on day one and spent a fortnight proving it.
- Set a stopping rule. If sections one to four produce no unexplained event, close it as explained and move on. Investigations without an exit condition become grudges.
Frequently asked questions
Set the threshold before you have a suspect, and set it on repetition rather than on a single volume. A one-off discrepancy inside your sensor's realistic accuracy is not worth a fortnight of anybody's time, and chasing it teaches the depot that management does not understand its own data. A recurring signature — same vehicle, same window, same shape of drop — is worth investigating even when each individual event is small, because the pattern is what the evidence rests on.
By checking, genuinely, and being seen to check. It is a reasonable objection and it is sometimes correct. Show the calibration history and the date of the last measured fill. Show fill volumes matching card volumes on that vehicle over recent weeks. Show that other vehicles on the same device type are not producing the same signature. If the sensor survives that, the objection is dead and everyone in the room knows it. If you refuse to test it, you have handed the driver the entire defence.
It can prove an unexplained loss of volume at a place and a time, with a vehicle and a keyholder attached. It cannot prove intent or identify the person holding the hose, and it should not be presented as though it does. What it does is remove every innocent explanation until only one class remains, and it puts the burden of explanation where it belongs. Treat it as the foundation of a case, not the whole case, and pair it with the physical and paperwork evidence.
Then the investigation is harder and considerably more important, because a depot with an internal fuel problem also has an internal records problem — someone is writing the logbook. Change who runs the investigation so it is not being conducted by anyone in scope, secure the bowser records and meter readings before they can be edited, and involve HR early. Fleets are usually much better at investigating drivers than at investigating the yard, and thieves work that out.
Yes, explicitly and in writing, before the first event. Concealed monitoring buys you one prosecution and costs you the trust of every driver who did nothing. Stated monitoring prevents more theft than it detects, which is the outcome you actually want, and it removes the argument that nobody knew. It also happens to be what most jurisdictions require of an employer monitoring staff. Announce it, explain what is recorded, and then enforce it consistently.
Run this checklist automatically
Fleet Fuel Management Software turns these checks into scheduled, assignable work with a full audit trail.