How a Fuel Retailer Found Out Its Pumps Were Giving Fuel Away
Use case Fuel management & reconciliationFuel and convenience retailer, 64 sites
How a fuel retailer found out its pumps were giving fuel away.
Tank variances were reviewed once a month and blamed on temperature. Cortex reconciled every tank every day, and found fuel leaving through the meters, the deliveries and one faulty probe.
Monthly. One network total, always “within tolerance.”
Daily. Every tank, every delivery, every meter.
At a glance
What changed, stage by stage.
01 · The challenge
A monthly report that always said “within tolerance.”
Every month, the team compared book inventory with tank gauge readings across 230 tanks. The total variance was always small enough to sign off, so nobody looked tank by tank.
Losses at a few sites were averaging out against gains at others. Temperature took the blame for everything.
Before Cortex
What the team saw first
- Network varianceWithin tolerance, every month
- Tank-level reviewNot done
- Delivery checksBOL accepted as delivered
- InstinctIt’s temperature. Sign it off
The difference
Same question. Different results.
The question“Reconcile every tank every day, and tell me which variances aren’t temperature.”
Claude / GPT / Public LLMs
Uploaded file
Answer
Network variance is 0.08% this month, within tolerance. No action needed.
Just an answer.
Cortex AI
Connected intelligence
Cortex AI- Tank gauge readings
- Fuel temperature
- Delivery BOLs
- Pump sales
- Dispenser meters
- Carrier history
- Probe diagnostics
- Weather
- Calibration records
- Delivery times
- 3 sites with meter drift
- One carrier drops 0.6% short
- Faulty probe at Site 41
Answer
Three problems hide inside a healthy total. Meters at 3 sites over-deliver, one carrier drops 0.6% short and Tank 3 at Site 41 has a faulty probe. Calibrate, claim, replace.
A decision.
02 · The finding
Small, steady losses, hiding inside a healthy average.
Cortex matched gauge readings, deliveries, BOLs and pump sales for every tank, every day, corrected for temperature. What was left followed patterns: by dispenser, by carrier and by probe.
Losses that rise with gallons sold point to meters. Losses that appear on delivery days point to the drop. Jumps right after a delivery point to the gauge itself.
What Cortex found
Thirty days, three tanks
- Site 12 · meter driftLoss grows with every gallon sold.
- Site 41 · faulty probeReads high right after each delivery.
- Site 27 · healthyTemperature swings, back to zero.
Cumulative daily variance in gallons, after temperature correction.
Calibrate, claim and replace, in that order.
- Recalibrated 11 dispensers at the three sites with meter drift.
- Filed short-delivery claims with the carrier, with gauge evidence attached to every drop.
- Replaced the faulty probe at Site 41 and set daily variance alerts for every tank.
03 · The impact
What changed.
An average can hide a leak.
Every month the total said we were fine. Tank by tank, three sites were giving fuel away.
Fuel loss is rarely one big event. It’s small, steady and spread out, which is exactly what a monthly total hides.
