Fuel distribution
Fuel delivery software that stops the guessing
Dispatch usually lives in one person's head. When they take a week off, the run-outs start. We build the system that holds what they know.
What fuel distributors should know before buying software
4 things that decide this
- 01A run-out does not just cost a delivery. It stops a farm, a fleet or a generator, and the emergency drop that follows breaks the rest of the day's route.
- 02Most vendors selling AI routing are selling a solver. Route planning here is constraint optimisation over compartment sizes, product compatibility, ullage and driver hours. The honest AI contribution is demand forecasting feeding that solver.
- 03Book stock and measured stock never agree exactly. Variance has at least six causes that look identical in a daily number, and only a statistical window separates them.
- 04Software can triage alarms. It cannot be your release detection method unless it is certified as one, and we will not pretend otherwise.
The plan is fine until a truck breaks down
Morning routing is the part everybody demos. It is also the easy part. The day then rearranges itself: a customer calls in an emergency, a site is blocked, a driver runs short of hours.
Re-planning under live constraints is harder than planning from a clean sheet. A route that is mathematically ideal but pushes a driver past their limit is not clever. It is illegal.
So the useful system is not the one with the prettiest optimiser. It is the one that knows what is really in each tank, and that survives contact with a Tuesday.
Reported by the client
What TankAware changed for Sutherland Excavating Ltd.
30%
less operational overhead, reported after TankAware launched
40%
fewer manual errors, reported by Sutherland Excavating Ltd.
20%
faster maintenance turnaround, same source
30%
higher inspection accuracy, same source
The distribution work we take
Sorted by how much difference it makes, not by how well it demos.
Tank monitoring and run-out forecasting
A level sensor plus a consumption forecast replaces a truck sent to read a number. The forecast is regression on usage and weather. Calling that AI is the hype line, and we do not use it.
Delivery planning that respects the constraints
Compartment sizes, product compatibility, tank ullage, delivery windows and driver hours. A plan that ignores any one of them gets abandoned by dispatch in week two.
Wetstock variance analysis
Separating temperature effects, meter drift, short loads, water ingress and genuine loss. The signal only appears over a window, which is why a daily number tells you nothing.
Alarm triage
Deduplicating and prioritising alerts from certified detection equipment. An alarm that is usually wrong trains people to clear alarms without reading them.
Driver paperwork and BOL matching
Capturing the delivery record at the drop instead of three days later. Loaded and delivered volumes legitimately differ after temperature correction, and the system should expect that.
Offline field capture
Bulk plants and rural cardlocks have no reliable signal. TankAware works fully offline and keeps the time the work happened, not the time it synced.
- Book stockWhat the paperwork says is there.
- MeasuredWhat the gauge reports today.
- VarianceSix possible causes, one number.
- WindowSignal appears over days, not hours.
- TriageMeter, temperature, loss or theft.
- DispatchOnly when the evidence supports it.
Skip the window and you get both failures at once: real losses ignored as noise, and technicians sent to sites where nothing is wrong.
A slightly wrong tank chart makes every reading wrong
A sensor reports a height. Converting that height to a volume needs a strapping table for that specific tank. If the table is slightly off, every reading is off, and the error is not constant as the tank empties.
This is how a monitoring rollout quietly loses credibility. Deliveries arrive when the tank is fuller than predicted, dispatch stops trusting the dashboard, and within a month everyone is phoning the site again.
The fix is to treat calibration as a first-class part of the product. Track which chart each tank uses, when it was last checked, and how far predictions have drifted from actual drops.
- Store the tank chart with the tank, and version it.
- Compare every predicted level against the delivered volume.
- Show dispatch a confidence, not just a number.
Offline field capture and multi-site operations
“TankAware has revolutionized how we manage petroleum sites. The real-time data and automation have exceeded expectations.”
Blake Sutherland · President, Sutherland Excavating Ltd.
The AI pitch against what is actually running
| Criterion | What gets marketed | What we would build |
|---|---|---|
| Run-out prediction | AI demand forecasting. | Regression on consumption and weather. Honest, effective, and testable against real drops. |
| Routing | AI route optimisation. | A constraint solver that knows about ullage, compartments and driver hours. |
| Leak detection | AI leak detection. | Triage on top of a certified method. The regulated method stays the regulated method. |
| Variance | A daily loss dashboard. | A statistical window, with the six causes reported separately. |
| Connectivity | Cloud-first, degrades to read-only. | Fully offline capture, with the original timestamp preserved on sync. |
What TankAware runs on
Application
Data and infrastructure
Field
Analytics
Questions distributors ask
01Will tank monitoring pay for itself in avoided truck rolls?
That is the right way to frame it, and the answer depends on how many tanks you currently check by driving to them. Sutherland Excavating Ltd. reported 30% less operational overhead after TankAware went live. Count your monthly level-check trips and your emergency deliveries first, because those two numbers drive the whole case.
02Our dispatcher does not want software. How do you handle that?
Take them seriously, because they are usually right about why the last system failed. Their knowledge is real and mostly undocumented: which customer never takes a Friday drop, which yard is impossible after rain. We build to capture those rules rather than override them, and a plan a dispatcher can edit gets used while a locked one does not.
03Can software be our release detection method?
Regulated detection methods must meet defined performance standards and carry certification, so the certified method stays the certified method. What software adds on top is real: deduplicating alerts, ranking them by likelihood, and holding the evidence trail. An alarm that is usually wrong trains people to clear alarms without reading them, and fixing that is an engineering problem worth solving.
04How do you reconcile a BOL against what actually arrived?
By expecting the two to differ and modelling why. Volumes change with temperature, so a loaded figure and a delivered figure can both be correct. Capture the delivery record at the drop with the driver present, hold the terminal document beside it, and flag differences outside a tolerance you set rather than every difference.
05We run several yards on different systems. Where do you start?
With the tanks and the records, not the reporting layer. Multi-site operators usually want a dashboard first, and a dashboard over inconsistent site data just makes the inconsistency visible faster. WAIQ took the same route for multi-site operations: standardise how work is captured, then the roll-up means something.
Go deeper
- Energy and fuel compliance →Inspection records, retention periods and what an inspector actually asks.
- Statistical inventory reconciliation →How a variance window separates a leak from a meter.
- Wetstock management →Tracking what is in the tank against what the paperwork says.
- Oil and gas software development →Field capture and proving an inspection happened.
- AI and IoT remote monitoring →An alarm nobody answers is worse than no alarm.
