Mine Fuel Reconciliation: From Delivery to Tank, Pump and Equipment
Mine fuel reconciliation is a volume balance run for every tank and every period: opening stock plus deliveries received, minus fuel dispensed, should equal the measured closing stock. What remains after correcting for timing, temperature basis and the known accuracy of meters and gauges is the unexplained variance. It must be traced to a place, a tank, a pump, a fuel truck, a machine or a shift, before anyone calls it a loss, and a loss is not yet evidence of theft. This guide sets out the balance, a worked week with its figures, the error sources, a variance table by machine and a pilot plan.
- Published
- Sources checked
- Reading time
- 10 min
- Written by
- Mirage Metrics team
Scope: diesel on a mine site with bulk tanks, fixed pumps, mobile fuel trucks and equipment. All figures in the worked example are invented. The Australian tax point is an example of why records matter, not tax advice.
Decision snapshot
- Best suited to
- Mine managers, fleet and maintenance managers, fuel supervisors and management accountants responsible for diesel on one or several sites.
- Inputs needed
- Delivery dockets, tank gauge readings or dips with calibration tables, pump totalisers, fuel transactions with machine IDs, fuel truck meters, engine hours or telematics, machine assignments.
- What can be automated
- Collecting readings and transactions, aligning them on the same cut-off, computing book and measured stock per tank, attributing fills to machines, computing expected use from each machine's own history, flagging variances and data gaps.
- What requires human approval
- Setting tolerance bands, checking meters and gauges, explaining a variance, deciding when a variance becomes an investigation, and any conclusion about people.
- When this approach is insufficient
- When tanks or pumps have no measurement at all, when machine IDs are not captured at the pump, or when meters have not been calibrated: the balance then shows where to instrument first, not where fuel went.
The perimeter of the balance
Draw the fuel path before building any report. Every node where fuel is stored has its own balance, and every transfer between nodes is a measurement that can be wrong. A site-wide total hides which node is off; a balance per node shows it.
| Node | In and out | Measured by | What the balance checks |
|---|---|---|---|
| Supplier delivery | Supplier's meter or tanker compartment volume | The delivery docket and invoice | Received volume, and on what temperature basis |
| Bulk tank | Deliveries in, pump transactions out | Tank gauge (probe) or manual dip with the tank's calibration table | Book stock against measured stock |
| Fixed pump or dispenser | From the bulk tank | Meter totaliser, transaction records | Totaliser movement against the sum of transactions |
| Fuel truck (mobile tank) | Loaded at a fixed pump | Its own meter, its dip or gauge | Loaded against dispensed, plus its own stock |
| Equipment | Fills from pumps and fuel trucks | Transactions tagged to the machine; engine hours or fuel burn from telematics | Fuel received against expected use |
The balance for each storage node is the same line: opening stock plus received minus dispensed gives the book closing stock, and the difference with the measured closing stock is the variance. Run it per tank and per period, with the same cut-off time for every source.
Data sources and what each one measures
| Source | Measures | Usual failure modes |
|---|---|---|
| Delivery dockets and invoices | Volume delivered, date, tank | Temperature basis (often a standard temperature), docket entered late or against the wrong tank |
| Automatic tank gauges | Level, volume, temperature, water | Probe drift, wrong calibration table, readings taken while product is moving |
| Manual dips | Level, converted with the tank's calibration table | Reading error, sloping ground, a calibration table that no longer matches the tank |
| Pump totalisers | Cumulative volume through the meter | Meter out of calibration, resets, totaliser not recorded at cut-off |
| Fuel transactions | Volume per fill, machine ID, operator, time | Tag misread or swapped, manual entries, duplicates, gaps during outages |
| Telematics | Engine hours, sometimes fuel burn | Units offline, hours not reset after a swap, fuel burn estimates that differ from dispensed volume |
| Maintenance and operations records | Machine assignments, downtime, idle time | Machine moved between pits or contractors without the fuel system knowing |
Put volumes on the same basis first
Suppliers often invoice fuel corrected to a standard temperature, while a gauge may read the volume at the tank's actual temperature. Comparing the two directly creates a variance that is only physics. Correct them to one basis using the standard procedures (ASTM D1250, API MPMS Chapter 11.1), or confirm that your gauge already reports standard volume.
Worked example: one week, one tank, one fuel truck
Illustrative example, invented figures: bulk tank T1 and fuel truck FT-2, one week, litres
Assumptions, all invented. Volumes are already on the same temperature basis. Pump P1 fills equipment and fuel truck FT-2. The tolerance bands are set by the site from its meter and gauge specifications and history; they are not standards.
| Line | Tank T1 | Fuel truck FT-2 |
|---|---|---|
| Opening stock (gauge, Mon 06:00) | 48,200 | 1,200 (dip) |
| Received | 102,000 (3 dockets) | 18,400 (from pump P1, tagged FT-2) |
| Dispensed | 96,850 (pump P1 transactions) | 18,050 (FT-2 meter) |
| Book closing stock | 53,350 | 1,550 |
| Measured closing stock (next Mon 06:00) | 52,100 (gauge) | 1,320 (dip) |
| Raw variance | -1,250 (1.29 % of dispensed) | -230 (1.27 % of dispensed) |
| Known correction | +1,080: 12 paper transactions during a system outage, not yet entered | None found |
| Variance after correction | -170 (0.17 %) | -230 (1.27 %) |
| Site tolerance band (assumed) | ± 600 | ± 200 |
| Result | Within band: watch the trend | Outside band: investigate |
Tank T1 first shows 1,250 litres missing, 1.29 percent of what it dispensed. Before any investigation, the fuel supervisor finds twelve fills written on paper during a system outage on Thursday and not yet keyed in: 1,080 litres. With them, the variance falls to 170 litres, inside the band. Nothing is investigated; the figure goes on the trend line.
Fuel truck FT-2 is different: 230 litres short, outside its band, with no known correction. The first checks are about measurement, not people: when FT-2's meter was last calibrated, whether the dip was taken on level ground, and whether any fill from FT-2 went unrecorded. Those checks come before any other hypothesis.
Timing, calibration and sensor errors
Timing
- Readings and transactions cut at different times: a gauge read at 06:00 and a transaction report closed at midnight.
- Manual fills entered days later, as in the example.
- Deliveries docketed on one day and pumped into the tank the next.
- Shift and time zone boundaries in systems that store time differently.
Calibration
- Dispenser meters: under OIML R 117-1, fuel dispensers for motor vehicles are accuracy class 0.5, a maximum permissible error of 0.5 percent for the measuring system and 0.3 percent for the meter. A site pump that is not verified against that class may do worse.
- Fuel truck meters, often verified less often than fixed pumps.
- Tank calibration tables that no longer match the tank after repairs or settlement.
Sensors and data
- Probe drift, water at the bottom of the tank, readings taken while fuel is moving.
- Machine ID tags misread, swapped between machines, or bypassed with manual entry.
- Duplicated transactions after a network retry.
- Telematics units offline, or engine hours not carried over when a unit is replaced.
Measured loss is not presumed fraud
Use four words, and keep them apart in reports and meetings:
- Variance: the gap between book and measured stock, before any explanation.
- Explained variance: the part accounted for by timing, temperature basis, known meter and gauge accuracy, and data errors that were found and corrected.
- Unexplained loss: what remains, traced to a node and a period.
- Suspected misuse: a hypothesis that an investigation examines, with evidence such as transaction times, machine positions and camera records. A conclusion about a person belongs to your investigation and disciplinary process, not to a report.
The distinction protects the investigation as much as the people. An alert raised as "theft" that turns out to be an uncalibrated meter costs credibility for the next alert that is real.
The variance table by equipment
Once the tanks balance, attribute fuel to machines. Compare what each machine received with what it would be expected to use, computed from its own engine hours and its own recent consumption rate, not a catalogue figure: two identical trucks on different haul roads burn different amounts.
| Machine | Dispensed (L) | Engine hours | Own median (L/h) | Expected (L) | Difference | Status | Next check |
|---|---|---|---|---|---|---|---|
| HT-07 haul truck | 4,900 | 62.0 | 71 | 4,402 | +498 (+11.3 %) | Above band | Check tag reads and fills on FT-2 |
| HT-11 haul truck | 3,950 | 58.0 | 70 | 4,060 | -110 (-2.7 %) | Within band | None |
| EX-02 excavator | 6,300 | 88.5 | 72 | 6,372 | -72 (-1.1 %) | Within band | None |
| HT-14 haul truck | 0 | 55.0 | 70 | 3,850 | -3,850 | No fuel recorded | Find the missing fills |
| D-03 dozer | 2,100 | not available | 65 | not available | not computable | Data gap | Telematics offline Tue to Fri |
Read the table as a set, not line by line. HT-07 received about 500 litres more than expected and HT-14, which worked 55 hours, received none. The first hypothesis to test is that HT-14 was fuelled from FT-2 under HT-07's tag, or on the paper fills from Thursday. Checking FT-2's transaction times against both trucks' positions settles it. D-03 cannot be judged at all this week, and says so.
Alerts and human validation
- Each alert names the node, the period, the size of the variance, the band it breached and the data behind it
- Each alert has an owner and a time to first review, set by the site
- The reviewer records a cause from a fixed list: timing, temperature basis, meter or gauge, tag or data error, unexplained
- Only 'unexplained' alerts move to an investigation, and only a named manager opens one
- Measurement causes create a maintenance or calibration task, so the same alert does not return
- Bands are reviewed monthly against the share of alerts that turned out to be measurement issues
Useful indicators and false positives
| Indicator | What it tells you | Usual false positive |
|---|---|---|
| Variance after correction, % of throughput, per tank per period | The headline number, compared with the band you set | One bad reading; look at the trend |
| Transactions with a valid machine ID | Whether fuel can be attributed at all | Tag hardware faults look like missing fuel |
| Machines with a data gap | Where the balance is blind | Offline telematics is not a fuel problem |
| Fuel per engine hour against the machine's own baseline | Unusual use by machine | Duty changes: new haul road, new pit, heavy grades, long idle |
| Alerts closed with a cause, and time to close | Whether alerts are being investigated | Alerts closed without a cause hide real problems |
| Share of alerts that were measurement or data issues | Health of the measurement chain | A high share is useful: it tells you which meter to fix |
In Australia, the ATO lists fuel issue records, meter readings, engine hours and telematics among the records that support fuel tax credit claims, to be kept for five years (ATO). A reconciliation built on those same records serves both purposes.
A pilot on one fuel bay or one site
- Pick one node set: a bulk tank, its pumps and the fuel trucks it feeds.
- Inventory every data source, its owner, its cut-off time and its calibration date. Fix the cut-off first.
- Rebuild the last few weeks from existing records to get a baseline variance and each machine's own consumption rate.
- Run the balance daily and review it weekly with the fuel supervisor and maintenance.
- Classify every alert by cause. Do not open an investigation during the pilot without the normal approval.
- At the end, compare: variance before and after corrections, alerts by cause, data gaps closed, time spent per week.
Where Mirage fits
Mirage's fuel intelligence system (called FuelMetrics until September 2026) is the fuel system Mirage builds and adapts to a site: it connects tanks, pumps and terminal equipment, deliveries, transactions, vehicle and telematics data and purchasing records, reconciles theoretical stock against measured stock and outflows, and flags anomalies for the team to review. That page describes the system, not a named client deployment. The documented deployment is at Transwin, a Moroccan logistics operator that runs a fleet of trucks and machines, including on mining freight, where the system tracks deliveries, on-site tanks, pumps and vehicle transactions and is being extended to more sites and mobile tanks.
Questions readers ask
What is a normal fuel variance on a mine site?
There is no single normal figure. It depends on how each point is measured: tank gauges, dispenser meters, fuel truck meters and dips each have their own accuracy, and deliveries may be measured on a different temperature basis from the tank. Set a tolerance band per tank from your own equipment specifications and history, and investigate what falls outside it and what drifts over time.
How accurate are fuel dispenser meters?
Under OIML R 117-1, the international recommendation for dynamic liquid measuring systems, fuel dispensers for motor vehicles are accuracy class 0.5, with a maximum permissible error of 0.5 percent for the complete measuring system and 0.3 percent for the meter. Whether a given site pump is certified to that class, and when it was last verified, is something to check on its own records.
Why do delivered litres not match tank litres?
Diesel expands and contracts with temperature, and suppliers often invoice volumes corrected to a standard temperature while a tank gauge may read the volume at the actual temperature. Put both on the same basis before comparing, using the standard correction procedures (ASTM D1250, API MPMS Chapter 11.1). Timing is the other usual cause: a delivery or a batch of fills recorded on the wrong side of the cut-off.
When does a fuel variance become a fraud investigation?
Only after measurement, data and timing explanations have been checked and ruled out, and the remaining variance is traced to a place and a time. Even then, the finding is unexplained loss at a named point, and deciding that someone took fuel is a matter for your investigation and disciplinary process, with evidence, not for a dashboard.
Does fuel reconciliation matter for fuel tax credits in Australia?
The Australian Taxation Office asks businesses to keep records showing the fuel acquired and how it was used, and lists examples such as fuel issue records, meter readings, engine hours and telematics data, kept for five years. A reconciled fuel record per tank and per machine is the kind of evidence that supports those claims. Check the ATO's current guidance for your situation.
Sources and scope
Checked on . Scope: diesel reconciliation on mine sites. Measurement standards are cited for their accuracy classes and correction methods; local legal metrology and tax rules vary by country.
- OIML R 117-1:2019, Dynamic measuring systems for liquids other than water: Table 2: fuel dispensers for motor vehicles are class 0.5; Table 3: 0.5 % for the measuring system, 0.3 % for the meter
- ASTM D1250, Guide for the Use of the Joint API and ASTM Adjunct (API MPMS Chapter 11.1): temperature and pressure volume correction factors for refined products, 15 and 20 °C bases
- Australian Taxation Office, Fuel tax credits: records you need to keep (updated February 22, 2023): records showing fuel acquired and used, including fuel issue records, meter readings, engine hours and telematics; five-year retention
This guide describes an operational method. It is not metrology, tax or legal advice, and any investigation involving people should follow your own procedures.
Related on this site
- Mining: operational systems for mine sites
- Fuel intelligence system: the fuel system behind this method
- Transwin case study: the fuel system in production
- Fuel card fraud detection for fleets: the road fleet side, where cards replace site pumps
Rebuild a past month on your own records
Send us a month of delivery dockets, tank readings, pump transactions and engine hours for one fuel bay. We will return the balance per tank and per machine, with every variance traced to its data, so you can judge it against what your team already knows.
Prefer to talk it through? Book a call