Neither is right on its own, and neither has to be. The fuel invoice and the hour meter are two independent records of the same work, kept for different reasons by different people, and a farm that has both already owns most of what is needed to check one against the other.
One piece is usually missing, and it is worth naming before anything else. The hour meter counts one machine. The invoice covers a delivery into the bulk tank that several machines draw from. Somebody has to connect the two, either with a metered dispenser that records which machine took what, or with a clipboard at the pump. On a month when one machine did nearly all the work, that clipboard is the whole system. Without it the comparison still runs against the tank as a whole, and just not against a single machine.
What the comparison produces is a gap, and a gap has ordinary explanations long before it has an alarming one.
What the machine should have burned
Engineering standards will tell you what the machine should have used, within a range. The equations in the ASABE agricultural machinery management standard are reproduced with attribution in Virginia Cooperative Extension publication 442-073, which gives the traditional estimate as 0.044 gallons per horsepower-hour of rated power at the power take-off, or about 0.223 liters per kilowatt-hour, and a fuller version that accounts for partial load.
Field-operation figures are published too. Iowa State lists roughly 1.7 gallons per acre for moldboard plowing, 0.4 for a seed-only planter and 1.45 for combining maize, which in metric is 15.9, 3.7 and 13.6 liters per hectare. The same publication attaches the warning that matters more than the numbers: actual fuel requirements may be as much as 35 percent higher or lower than the values listed.
Why the benchmark cannot settle the argument by itself
A single published figure will not tell you whether your gap is real. Two things make the range wide. The first is that measured consumption for the same operation moves with speed, depth, width, soil moisture and soil density. One 2024 study of chisel plowing recorded specific fuel consumption per draft power of 0.787, 0.629 and 0.525 liters per kilowatt-hour as speed rose from 3.7 to 6.9 kilometers per hour, which is 0.155, 0.124 and 0.104 gallons per horsepower-hour as speed rose from 2.3 to 4.3 miles per hour. The authors report that as a 33.3 percent decrease, from one tractor working one soil at one depth.
A Polish study of 2024 shows the same spread across whole operations rather than one implement, reporting that per hectare farmers used the most fuel for harvesting, 25.2 liters, and plowing, 22.4 liters, which is 2.7 and 2.4 gallons per acre, against 8.5 liters per hectare or 0.9 gallons per acre on grassland. Same operations, different continent, and a range wide enough that no single figure settles anything.
The second reason is that two defensible estimating methods disagree with each other. The Virginia publication works a 350-acre example, 142 hectares, and gets 1,064 gallons or 4,027 liters from the traditional flat method against 904 gallons or 3,422 liters from the load-aware method. That is a 15 percent difference between two respectable ways of predicting the same year’s fuel, before any real machine is involved.
So how large does a gap have to be
Borrow the limit from the people who had to write one down. Fuel storage regulators in two countries have set formal reconciliation standards, and they agree on what gets reconciled even though they set the bar in different quantities and by different methods.
| Requirement | United States, 40 CFR 280.43 | New South Wales, UPSS Guidelines |
|---|---|---|
| What is reconciled | Inputs, withdrawals, and volume remaining | Fuel delivered, fuel sold, fuel remaining |
| How often | Monthly, with daily readings | Monthly, with daily readings |
| Kind of limit | A monthly volume discrepancy | A continuous leak rate |
| Where the bar sits | 1.0 percent of flow-through plus 130 gallons, 492 liters | 0.76 liters per hour, 0.2 gallons, at 95 percent probability of detection |
| Deliveries checked against | Delivery receipts | Delivery records |
The two limits in that table are not two numbers for the same measurement, and neither converts into the other. One is a monthly volume that failed to balance. The other is the smallest steady leak the equipment has to be able to catch. What travels between them is the method above those two rows: what came in, what went out, what is still in the tank.
The American rule requires inventory control to detect a release of at least 1.0 percent of flow-through plus 130 gallons on a monthly basis, with inputs reconciled against delivery receipts. Neither rule was written for farms, and both describe the reconciliation this article is about.
The second term in the American rule is the one that matters at farm volume, and it is the one usually dropped when the rule gets quoted. On a 2,000-gallon month, 7,571 liters, one percent is 20 gallons and the fixed term is 130, so the rule actually tolerates 150 gallons, which is 7.5 percent. Quote the rule as one percent and you will chase a gap the regulator would have ignored.
The four boring explanations, in the order to check them
New South Wales guidance lists what a failed reconciliation can mean, and the list is worth memorizing. Describing statistical inventory reconciliation specifically, it states that a failed or inconclusive result could be caused by mis-calibrated dispensers, inaccurately metered deliveries, human error in recording, or stolen product.
Three of those four are administrative. The pump is reading wrong, the delivery was measured wrong, or someone wrote the wrong figure in the book. Only the fourth is the one people jump to, and it is the one to consider last, after the other three have been ruled out with a dipstick and a calibration check. Accusing a person before checking a meter is how a manager spends credibility that took years to build.
There is a fifth explanation the regulators do not list, because it is specific to machinery. Fuel consumption is a symptom. Where the fuel can be attributed to one machine, consumption drifting upward with hours unchanged is a maintenance signal before it is an accounting one, and the document that says what maintenance was already due is the factory service interval.
A sixth belongs to no list at all, because nothing about it is a failure. Some of the fuel left the gate on purpose: the tractor lent out for a weekend, the grader sent up the road to the school bus stop, the tank filled for somebody else’s job. It never reads as theft and never reads as calibration, and it stays invisible until the community support record gets a line of its own with a date and a name beside it.
What the record buys you when the numbers agree
The reconciliation is worth doing even when nothing is wrong, because a matched pair of records is evidence and an unmatched pair is an argument. A report commissioned by the New South Wales Department of Primary Industries puts the case for measurement directly: without a reading taken on some interval, farmers only have ballpark numbers when it comes to the fuel used and would be unable to detect even major variations in fuel consumption indicative of maintenance issues, component failures or poor operation.
The stake is not trivial. The same report finds that over 80 percent of the energy consumed by agriculture in New South Wales is diesel, and Western Australian grains benchmarks put fuel and oil at 5 percent of farm income, plus or minus 2, with machinery operating costs, contractors included, running larger than fertilizer and chemical costs combined. That is the order of magnitude the rest of Finance works in, and it is large enough that a gap left open the whole cycle is not a rounding error.
Where to start, with what is already in the drawer
About an hour: pulling one machine’s invoices or pump notes and its two hour-meter readings is the quick half, and balancing the whole tank’s deliveries against everything that left it is the half that takes longer, because it depends on every machine’s draw for the month, not just the one you started with.
Take one machine and one month. Add up the fuel that machine drew, read the hour meter at the start and end, and divide. Compare the result to the published figure for the work that machine did, allowing the 35 percent band the extension publication warns about. Then, separately, balance the whole tank for the same month, which is a different comparison with a different limit and answers a different question. That same pair of meter readings, taken at the two ends of a whole cycle instead of a month, is also the denominator of the repair bill for one machine.
What turns this into management rather than arithmetic is what happens to the gap. A gap with no owner and no date is a fact nobody acts on, and it will still be there next cycle. A gap with a calibration check booked for Thursday and a name against it has been closed out, which is the step the four functions of management turn on. Finding the same number every year and never correcting it leaves the cycle open, no matter how carefully the number gets recorded.
There is a second measurement hiding inside the first one, and nobody takes it: the number of days between the month closing and somebody putting the two records side by side. That number has a name, the detection delay, and a tank reconciled in March for January is answering a question that stopped being useful in February.