Your dashboard is guessing. It estimates fuel use from injector pulses and airflow, rounds generously, and never learns what the pump actually charged you in gallons. The tank-to-tank method is the gold standard because it measures the only two things that matter — miles driven and gallons burned — at the two places they cannot lie: the odometer and the pump. This is how to calculate fuel mileage properly, why the result disagrees with the number glowing on your dash, and how to read the figure once you have an honest one.

The tank-to-tank method, step by step

Fill the tank until the pump clicks off, and reset the trip odometer (or write the odometer reading down). Drive normally. At the next fill, fill to the click again and note the gallons. Miles since the last fill ÷ gallons just added is your MPG. That is the whole method — everything else in this article is about why it works and why shortcuts do not.

Two habits make the measurement clean. Fill completely each time: the full tank is the fixed reference point both measurements hang from, and a partial fill leaves the starting level unknown. And record as you go — a receipt, a phone note, a fuel log — because the method’s one true enemy is the unrecorded top-up, whose miles turn up in the next fill carrying none of their fuel. Every fuel mileage calculator on the web, including the one on this site’s mileage page, is this same division with a keyboard attached; none of them can repair a missing gallon.

Setting up the log: what to record

Three numbers per fill are the whole apparatus: the odometer reading, the gallons, and the date. The odometer reading is what turns a pile of receipts into a measurement — gallons without miles is just spending. Record the gallons to the hundredth the pump shows, because on a small fill the second decimal is not pedantry: rounding 6.14 gallons to 6 swings a short-run figure by more than the click-off error you are trying to cancel. Consistency helps where it is cheap: the same pump at the same station fills to the same level more repeatably than random forecourts, and filling a cold car in the morning rather than after a hot highway run keeps the ritual identical. None of this requires an app. A notebook in the door pocket outperforms a sophisticated tracker abandoned in week two — the log that exists beats the log that was planned, every tank of the year.

One modern complication deserves its own warning. In a plug-in hybrid, the tank-to-tank method measures gasoline MPG faithfully — and the figure can look absurd, because half the miles arrived silently from the battery. 900 miles on 9 gallons is a true 100 MPG of petrol and a meaningless description of the car’s energy use. For a plug-in hybrid, keep two logs: gallons for the petrol figure above, and kilowatt-hours from the charger for the electric miles, then read the two side by side. Adding them into a single MPG requires an energy equivalence conversion with assumptions baked in; keeping them separate requires none, and separate is honest.

A worked log over three fills

MPG = miles ÷ gallons at each fill. The combined row — total miles ÷ total gallons — is the car’s real average; averaging the three per-fill figures instead would give 30.5, a number no tank ever produced.

Read the three fills the way a scientist would. Each is a genuine measurement over a different kind of driving — a highway run, a mixed week, a gentle weekend — and each is hostage to how full “full” was at that pump. The combined row is the one to quote: 1,145 miles on 37.9 gallons is 30.2 MPG, the car’s true average gas mileage over the whole period. Note what averaging the averages would have told you instead: 30.5. The naive mean treats a 245-mile fill and a 512-mile fill as equals, and the error it introduces is not rounding — it is a different, wrong statistic.

Why the dashboard number drifts

The trip computer is not broken; it is answering a slightly different question with slightly different data. It counts fuel by injector timing rather than by gallons delivered, and tiny calibration errors compound over a tank. It counts idling fuel against zero miles, correctly — but drivers compare its figure against a memory of “mostly highway” driving and call the difference optimism. It trusts the odometer, which trusts the tyres: worn or non-standard tyres change the distance the car believes a wheel revolution covers. And on short trips it shows you a figure dominated by cold-start running, then gets blamed for it at the pump. Treat the dashboard as a trend instrument — brilliant for spotting a sudden change — and the tank-to-tank figure as the measurement.

This matters most when the number is used for something. A gas mileage comparison between two cars is only honest when both were measured the same way, over comparable driving: to compare gas mileage properly, run both through the fill-up method for two or three tanks each. The laboratory figures behind a fuel economy calculator or an official rating come from a standard cycle no driver reproduces — useful for ranking cars against each other on the day they were tested, misleading as a prediction of your tank.

Turning MPG into money — carefully

MPG becomes money through one reciprocal and one multiplication. The reciprocal answers how much gas per mile the car uses: at 30.2 MPG, 1 ÷ 30.2 = 0.033 gallons — your gas per mile burn. Multiply by your pump price and any mileage cost calculator or gas mileage cost calculator has its answer: at an illustrative $3.50 a gallon, $0.117 of fuel per mile. The same reciprocal is what a fuel consumption calculator displays as L/100km — one measurement, two dialects.

The reciprocal also disciplines the savings question, which is where an mpg savings calculator earns its keep. Over 12,000 miles, improving from 20 to 25 MPG cuts fuel from 600 to 480 gallons — 120 gallons saved, $420 at the illustrative price. Improving from 30 to 35 MPG — the same five-point step — cuts 400 to 343 gallons: 57 saved, $200. MPG gains are worth less the higher you climb, and only the gallon figures, never the MPG difference, tell you how much less. A fuel efficiency calculator that reports the step instead of the gallons is flattering you.

Every step is the same five MPG points. The gallons — the thing actually saved — shrink at every step, because fuel use is the reciprocal of MPG.

Read the table as a shopping tool. A buyer choosing between a 25 MPG car and a 30 MPG car is choosing $280 a year at these prices; choosing between 35 and 40 is choosing $150. The table also explains a quiet scandal of car advertising: the MPG gap between models is marketed hardest exactly where it is worth least — at the efficient end, where the same engineering effort moves the gallons by the smallest amount.

The log as a diagnostic instrument

Keep the log running after the curiosity fades, because its second job is more valuable than its first. A driver with a settled baseline — say a rolling combined figure of 30.2 MPG, recomputed over the last three fills each time — owns an early-warning system. Random wobble moves the rolling figure by tenths. A real change moves it by points and keeps it moved. The diagnostic order is cheapest first: tyre pressures (the one cause that is free to check and embarrassing to miss), then the load and the roof — is the car still carrying last month’s holiday on its roof or in its boot? — then the driving pattern (a new commute is a new car, fuel-wise), and only then the machinery: a dragging brake, a failing sensor, a thermostat stuck open so the engine never quite warms. Each cause announces itself in the log before it announces itself anywhere else, and the log cannot do that work unless the boring, honest fills were recorded when nothing was wrong. Mechanics have a version of this rule about oil; the fuel log is the same discipline with a receipt.

Reading your number honestly

Expect the figure to breathe. Winter fuel, cold starts, a roof rack, a month of short trips — each takes its share, and a single tank is an anecdote about one kind of driving. Keep the log running, read the combined figure over several fills, and investigate steps, not wobbles: a sudden drop that no change in driving explains is the car reporting a fault in the only language it has. Measured this way, MPG stops being a sticker claim and becomes what it should have been all along — your car’s own number, from your own road.