How to measure fleet fuel consumption — and tell a real saving from noise
Record the fuel actually dispensed and the distance at every refuelling, for each vehicle, and convert it to litres per 100 km. Then compare like with like: the same vehicles in the same calendar months, or vehicles that receive the measure against comparable vehicles that do not. In our own records, the results we cannot rely on fail on the comparison, not on the arithmetic.
This applies to any fuel measure a fleet tries — tyres, driver training, aerodynamics, route planning or a product like ours. How we run a trial of NanoEFX specifically is set out in our measurement methodology; this page is about measuring well in the first place.
The full-tank method, step by step
It needs nothing but the pump and the odometer, and it measures what you actually pay for: fuel that went into the tank.
1 · Fill to full
Fill the tank to full and note the odometer reading. This is the starting point; the fuel in this fill is not counted.
2 · Drive normally
Run the vehicle as usual. Keep the driver, route type and typical speed as consistent as operations allow, and note when they change.
3 · Record every refuelling
At every refuelling, fill to full again and record the litres dispensed and the odometer reading. If a fill is not to full, carry its litres forward to the next full fill.
4 · Calculate
Litres per 100 km = litres dispensed ÷ kilometres driven since the previous full fill × 100. For ships and generators, use litres per operating hour instead.
Kilometres per litre is the same measurement on the inverse scale. It is fine for describing a vehicle, but a gain in km/L is not the same percentage of fuel saved — why 20% better is not 20% less fuel.
Which fuel data to trust
The figure that counts is the fuel dispensed, recorded at every refuelling, per vehicle.
| Source | What it records | Use it for |
|---|---|---|
| Pump or fuel-card record | Litres actually dispensed, with date and vehicle | Consumption and cost — the basis of any saving |
| Odometer at each refuelling | Distance between fills | Turning litres into litres per 100 km |
| Fleet management system | Usually fuel and distance per vehicle | Usually sufficient, if it holds dispensed fuel — our German refuse-truck trial ran entirely on the operator’s own system data |
| Telematics consumption estimate | Fuel calculated from driving behaviour, not fuel dispensed | Driving style; not sufficient on its own to show a saving |
| Run-hour records (ships, generators, plant) | Operating hours and fuel used | Litres per operating hour instead of per 100 km |
A monthly fleet fuel report you can compare year on year
One row per vehicle per month, compared with the same vehicle in the same month a year earlier rather than with last month.
| Column | Why it is there |
|---|---|
| Month, vehicle, type | Each vehicle is compared with itself |
| Kilometres and litres | Summed over complete full-to-full stretches that end in the month, so the tank level at the start and end of the month does not distort the result; dispensed fuel only |
| Litres per 100 km | Calculated, so every vehicle is on the same scale |
| Same month last year | Removes most of the seasonal effect; the template takes it from the same log |
| Change in % | Calculated on litres per 100 km, so it reads as a change in fuel used per kilometre, not in km/L — and it says nothing about the cause until the notes are checked |
| Notes | Driver, route, load, maintenance, new tyres — anything else that changed |
The template has three sheets: a short guide, a refuelling log that calculates litres per 100 km between full fills and flags missing or out-of-order entries, and the monthly report built from complete full-to-full stretches. It contains no data and uses only basic formulas.
Five things that fake — or hide — a fuel saving
Each of these is in our own published data. We show them because a fleet manager will meet every one of them, whatever measure is being tested.
The season
A spring baseline against a summer test period carries a seasonal difference before anything else changes. Three route buses measured from February to June 2020 coincided with the collapse in passenger numbers; their figures are uninterpretable however good the baseline was. Nishitetsu
The load
Five refuse trucks used 18.2% less fuel than in their own March–May baseline — and their utilisation fell 10.5% between the two periods. A lighter truck burns less fuel per 100 km whatever else is done to it. German refuse-truck trial
The driving
One truck was driven far harder in the test period than in the baseline, so the two periods could not be compared directly. Only a model-based correction for drive mode gave a figure, roughly 5% less fuel, and it is a model, not a measurement. Frederikshavn
The comparison
On the same fleet, eleven vehicles compared year on year give a median of +8.8% in km/L; three compared directly before and after give up to +50.3%. Same product, same operator: the gap tracks the comparison method, and it shows why direct before-and-after figures need caution. ALSOK
The unit
An improvement in km/L is a smaller saving in fuel: +20% km/L is 16.7% less fuel. Budgets and CO₂ follow litres, so convert before you compare or report. The arithmetic
And one more that hides in plain sight: drivers who know a fuel measure is being tested tend to drive differently. Keep operating and driving instructions unchanged during the test; whether drivers are told is a matter for your own workplace, safety and privacy rules.
How long to measure, and against what
Long enough
Record a baseline of two weeks to a month before anything changes, and plan for at least three months afterwards. A signal can appear within weeks, but separating it from normal variation needs distance, refuelling cycles and operating time.
Against the right thing
This month against last month is hard to defend: consumption changes with temperature, daylight, load and traffic. Compare the same calendar months a year apart, or run vehicles with the measure alongside comparable vehicles without it over the same weeks.
A control group, where you can
You do not need one, but a result with one is worth considerably more. Our only dataset with a genuine control group compared treated buses with untreated buses in the same period: Sanden Kotsu. The full trial protocol is in our measurement methodology.