Pilferage doesn't show up as a missing entry — it shows up as a slow-moving km/L drift. Six signals your telematics already has to catch it before it's two quarters old.

Detecting Diesel Pilferage in Bus Fleets: 6 Signals Your Telematics Already Knows

It's the first week of the month. The fleet owner is staring at the fuel reconciliation sheet at 10pm in a cramped office above the depot in Swargate. The litres bought match the slips stapled to the file. The kilometres run match the GPS report the operations head pulled this morning. On paper, nothing is wrong. And yet the fleet-wide km/L has slipped from 4.4 to 4.2 over six months. The numbers add up everywhere except in the bank account, where roughly six to eight lakh rupees a month is no longer arriving.

This is the diesel pilferage problem in Indian intercity bus operations in one paragraph. It does not show up as a missing entry. It does not show up as a smoking-gun stolen tanker. It shows up as a slow-moving km/L variance that nobody can pin to one driver, one route, or one pump. By the time it is visible in the P&L, it has been bleeding for two quarters.

The good news is that the data to catch it is already in your telematics. AIS-140 GPS, fuel slips, and a per-vehicle baseline km/L are enough to build six signals that surface pilferage as a pattern — not as an accusation. This article walks through those six signals, the math behind why they matter, and the supervisor protocol that turns flagged data into a conversation instead of a confrontation.

The math: why even small pilferage destroys margins

Fuel is widely accepted to be 50–55% of operating cost for an intercity sleeper coach in India. That single line is the reason this conversation is worth having. Almost every other cost — driver wages, insurance, AMC, road tax, permits — is fixed or near-fixed at the start of the year. Fuel is the one variable line item large enough to either deliver your margin or wash it away.

Some illustrative arithmetic, clearly labelled as illustrative:

• A 50-bus fleet, each bus averaging 200 km/day, running 28 days a month.
• Total km/month: 50 × 200 × 28 = 2,80,000 km.
• At a baseline 4.2 km/L: 2,80,000 ÷ 4.2 ≈ 66,667 litres/month.
• At ₹95/litre: ≈ ₹63.3 lakh/month, or roughly ₹7.6 crore/year, on diesel alone.
• 3% pilferage on that base ≈ ₹22.8 lakh/year. Five percent is closer to ₹38 lakh.

That money is already in your budget. It has been sanctioned, paid out, and reconciled. It is just walking out at the pump nozzle.

And here is the compounding cruelty: pilferage is the one cost that does not get cheaper as you scale. Tyre procurement gets cheaper with volume. Insurance premiums negotiate down. EMI rates improve with balance-sheet strength. Pilferage scales linearly with litres bought. Doubling the fleet doubles the leakage if the controls do not change.

Why pilferage is hard to catch (and harder to confront)

If it were easy, it would already be solved. It is not, for three structural reasons.

It is distributed. The pump attendant, the driver, the workshop mechanic, the route supervisor — collusion can sit at any one of those nodes, or at any pair of them. A single-point check (just the slip, or just the GPS) misses the multi-node version.

It is small per event. Five hundred rupees here, twelve hundred there. There is no single transaction big enough to trigger an audit. The fleet that loses ₹25 lakh a year loses it in roughly two thousand individual events of ₹1,000–₹1,500 each. No one of them is worth a confrontation.

It is politically loaded. Indian fleets run on relationships. Drivers stay with operators for ten and fifteen years. Mechanics know which bus has which quirk. Accusing the wrong person, or even the right person without enough data, poisons a depot for months. The cost of a false accusation is a senior driver walking out with three others in solidarity — and you can't run the Pune–Bangalore line without them on Friday night.

The right system, therefore, surfaces patterns and never makes accusations. The supervisor still has the human conversation. The data simply does the homework before the conversation starts.

The 6 detection signals your telematics already has

Each of these six signals is computable from data your fleet is already generating: GPS traces from your AIS-140 box, slips photographed by drivers on WhatsApp, and a per-vehicle baseline km/L. Fleetain runs all six automatically on every fuel event, but the logic is the logic — any operator with the data can run a version of this manually.

1. km/L below vehicle baseline

What it looks like: A specific bus, on a specific route, drops 5% or more below its baseline km/L for three consecutive trips.

What it means: Not necessarily theft. The honest first explanation list is mechanical: a clogged air filter, underinflated rear tyres, an AC compressor clutch dragging, a sticky brake calliper, a degrading injector. An aggressive new driver on a familiar route can cause the same dip. Investigate mechanical first — this is both fair and, statistically, more often correct.

What to do: Pull a fuel-vs-distance chart for that bus for the last 30 days. A sharp recent dip points to mechanical. A slow drift over weeks points to fills. Treat the two differently.

What NOT to do: Confront the driver on the basis of this signal alone. Half the time, the driver will be the one who tells you the air filter is overdue.

2. Fill volume exceeding tank-capacity headroom

What it looks like: A driver submits a slip for 280 litres on a 250-litre tank. Or the slip claims 220 litres on a tank that the GPS-correlated fuel-level reading suggests was 60% full eighteen hours ago.

What it means: The arithmetic is impossible, so one of two things is happening. Either the slip is inflated (the pump charged for litres that physically could not have gone into the tank), or the tank reading is wrong (sensor fault, second auxiliary tank, or someone has fiddled with the float).

What to do: Photograph the dip-stick reading before the next fill on that specific bus. Cross-check against the slip. This is one of the few signals where a single event is worth investigating, because the physics is unambiguous.

3. GPS position vs pump claimed on the slip

What it looks like: The AIS-140 trace shows the bus parked at Vita at 14:30. The slip says the fill happened at a Karad pump at 14:30. Karad and Vita are 50 km apart. The bus cannot have been at both.

What it means: Either a slip from a different vehicle has been submitted against this bus, or the pump issued a backdated slip for a fill that did not happen.

What to do: This is the cleanest single signal on the list. Pull the GPS trace for the slip's claimed timestamp. If the bus was demonstrably somewhere else, the slip needs an explanation from somebody. Unlike signals 1 and 2, this one does not have a mechanical alternative explanation.

4. Time-of-fill outliers

What it looks like: Most refills on a given route happen at one of three or four predictable stops. A 2:40 AM fill at an isolated pump 8 km off the documented route is an outlier.

What it means: Occasionally legitimate — emergency, detour, an unscheduled passenger drop. Often, however, it is a marker for a problem pump that a driver is making a habit of visiting.

What to do: Look at the pattern over a month, not the single event. One outlier is life. Twelve outliers in thirty days at the same pump is a sourcing and route-discipline conversation.

5. Per-driver pattern across multiple buses

What it looks like: When you sort flagged fills by bus, the cluster looks random. When you re-sort the same flagged fills by driver, the same one or two names keep appearing across three or four different vehicles.

What it means: The cluster sits on the driver, not on the vehicle. That is a different conversation entirely — and a different remediation.

What to do: Rotate the driver to a different route and a different bus for a fortnight. Watch what happens to the signal. If the flags follow the driver, the data has answered a question. If the flags stay with the original buses, the cluster was vehicular all along and you have just exonerated a senior driver fairly.

6. Receipt-vs-actual delta from AI-extracted slips

What it looks like: The slip — extracted by AI from the WhatsApp photograph, with litres, rate, total, GST and pump name all parsed — claims 240 litres. The AIS-140 fuel-level sensor recorded a 215-litre jump at the same timestamp. Delta = 25 litres.

What it means: Pump short-fill (the pump physically dispensed less than the slip recorded), a meter calibration drift, or a colluding short-fill where the difference is shared between attendant and driver.

What to do: One delta is noise. A repeat pattern at the same pump across multiple buses is documentation. BS-VI buses, with their fuel rail pressure sensors and AdBlue consumption logs, give a third cross-check on actual diesel burnt. File documented patterns with the franchise HQ — BPCL, IOC and HPCL franchise audit teams take this seriously when the operator brings dated, GPS-stamped evidence rather than just a complaint.

Building a per-vehicle CPK / km-per-litre baseline

None of the six signals work without a credible baseline. A generic "all buses should do 4.2 km/L" threshold catches nothing useful and flags everything in the monsoon.

The baseline that actually works has three properties:

• Per vehicle. Two identical buses ex-showroom drift apart within 18 months. The baseline has to be the specific bus's own history.
• Per route, with load context. Pune–Goa loaded outbound and lightly-loaded return-leg differ by 8–12% routinely. The system needs to know which leg it is comparing.
• Seasonally corrected. Monsoon ghat sections consume more diesel. Peak summer expressway runs at night consume less. AC load through April–June is materially different from December.

Each bus needs roughly 30 days of clean operating data before its baseline is statistically reliable. Fleetain holds these as separate baselines per (vehicle, route, season), which is what makes a 5% deviation a meaningful signal instead of normal seasonal noise. For the operational side of acting on these signals once they are surfaced, see our Fuel Efficiency Improvement playbook.

The supervisor's "talk to the driver" protocol

The data does the homework. The supervisor does the conversation. Here is the protocol that works.

Never lead with accusation. Lead with the data, openly and visibly. "Your km/L on the Kolhapur–Pune run has dropped from 4.5 to 4.0 over the last two weeks. Let's figure out together what has changed."

Most of the time, the driver knows. The AC clutch has been dragging. The rear tyres need rotation. There was a diversion through bad road for ten days. The honest mechanical cause and the conversation about leakage surface together — and the driver becomes part of the diagnostic, not the target of it. This is also where a clean Work Order Management trail pays off, because the mechanical alternatives can be ruled in or out from records, not from memory.

When the driver genuinely has no explanation, and the data still holds up, the conversation can continue. But the data did the accusing, not the manager. The driver who has nothing to hide is cleared faster by the data than by anyone's hunch — which is itself a recruitment and retention argument, not just a fraud-prevention one.

Pump and route patterns Maharashtra operators should know

Without naming specific pumps — because the variable is the franchisee and the staff, not the brand — certain stops near Solapur, parts of the Karad–Satara stretch, and some outlets in the Bhusawal corridor come up repeatedly in conversations with other operators. The Konduskar-style intercity routes that run these corridors nightly are the most exposed simply because of repeat exposure.

The fix is not avoidance. You cannot reroute the Pune–Hyderabad run away from its natural fuel stops. The fix is documentation. Every flagged fill, with its GPS trace, its photographed slip, and its delta from the fuel-level sensor, becomes one row in a case file. When that file has thirty rows on the same brand of pump along the same corridor, the brand's regional office has something concrete to act on — and franchisees have, in our experience, been audited and replaced on the back of operator documentation. It does not happen on the back of phone complaints.

What this does for your accounts and ops team

The cultural shift is as valuable as the rupee shift.

The accounts team stops being the "month-end reconciliation panic" team and becomes the "daily anomaly review" team. Five minutes of queue review every morning replaces a week of January reconciliation hell. The flagged fills surface as a list of, typically, four to eight items per day for a 50-bus fleet — most of which the supervisor dismisses with one click as legitimate, with a handful escalated for a driver conversation or a pump-side query.

The drivers who are clean — which is the majority — are cleared faster and more fairly than any hunch-based system could clear them. The drivers with a problem find the data conversation harder to dodge than the manager conversation. And the depot, on balance, becomes a more honest place to work.

None of this happens overnight, and none of it is a substitute for the basics of a well-run fleet. Pair this signal layer with disciplined preventive maintenance (see our note on Zero Breakdown Strategy) and the fuel line of your P&L starts behaving like a controlled cost again instead of a leak you cannot find.

FAQ