Fuel pilferage: what a system can prove and what it cannot

Fuel is usually the largest controllable cost in a fleet and the one most businesses suspect they are losing. The suspicion is often right, and the usual response — buy sensors — addresses the symptom while leaving the measurement problem untouched.

Before hardware, there is a process question worth settling: can you currently calculate consumption per vehicle at all?

Issue against the vehicle, not the driver

In many operations fuel is logged against whoever collected it. That record cannot produce a per-vehicle consumption figure, and no amount of reporting will reconstruct one afterwards.

Change this first. Fuel is issued to a vehicle; the driver is recorded separately as a second attribute. Everything downstream depends on it, and it costs nothing but a process change.

What consumption data actually tells you

With fuel against vehicle and a reliable odometer, you get litres per hundred kilometres per unit. That figure is useful in three ways and misleading in a fourth.

Useful: comparing a vehicle against others in the same category; comparing a vehicle against its own history; and spotting the step change that usually indicates either a mechanical problem or a behavioural one.

Misleading: comparing across categories, or across route types. A truck on hill routes against one on motorway will differ legitimately, and treating that gap as pilferage will cost you a driver who was doing nothing wrong.

Where the loss usually is

In our experience the common patterns are mundane: fuel drawn on a vehicle that was off road; quantities rounded up consistently at a particular pump; a vehicle whose consumption worsened sharply and permanently on a date that matches a driver change; and tank capacity exceeded by a single issue, which is arithmetic rather than inference.

That last one is the most useful automated check you can build, and it needs no hardware. If a vehicle with a 300 litre tank is issued 340 litres in one transaction, something is wrong with either the record or the transaction.

What data cannot do

Consumption analysis establishes that a vehicle is using more fuel than it should. It does not establish why, and treating it as proof of theft is both unfair and operationally counterproductive.

A worsening figure can be a failing injector, an underinflated tyre, a route change nobody recorded, or a load increase. Investigate before accusing. A system that generates accusations rather than questions will stop being trusted, and the data will quietly degrade as people route around it.

Sequence that works

Fix the issuing process so fuel attaches to vehicles. Capture odometer readings at the gate with validation against the previous reading. Run three months and establish what normal looks like per category. Then investigate the outliers individually.

Only after that is there a case for sensors, and by then you will know which vehicles justify the expense rather than fitting them across a whole fleet on suspicion.

Common questions