Two competent analysts can model the same fleet and reach opposite conclusions, and the gap is almost never the vehicle price. Argonne National Laboratory's comprehensive work on total cost of ownership puts the problem plainly: analyses of alternative fuel vehicles "have often focused on the purchase cost and the fuel cost," and examining only those two "does not fully capture the differences in total costs between powertrain types." For commercial vehicles there are eight components rather than six, and the two that most often decide the answer — payload capacity and driver labour — are precisely the two almost no calculator thinks to ask you about. Pressure-test your own model with us.
Why Two Honest TCO Models of the Same Fleet Disagree
For fleet directors, finance leads and anyone being shown a payback chart: the eight cost components a rigorous model includes, the assumptions that quietly decide the outcome, and the eight questions worth asking before you accept any number — including ours.
Eight Components, and the Two That Go Missing
Argonne's framework includes the six a passenger-car model would use, plus two that exist only because these are working vehicles. The second pair is where commercial answers are won and lost.
Argonne's stated component set covers purchase and depreciation, financing, fuel, insurance, maintenance and repair, taxes and fees, and then — for commercial vehicles specifically — payload capacity and labour. Their framing is worth keeping: for commercial vehicles they included the same components as for passenger vehicles "as well as direct operational costs, including both labor expenses and the marginal payload expenses." A model that stops at six is not a simplified commercial model; it is a passenger-car model applied to a truck.
There is a structural reason these two get dropped. Payload is a revenue effect rather than a cost line, and labour usually sits in a different budget from the vehicle. Both are real money and neither fits the shape of a cost spreadsheet, so they quietly fall out — which is how a model can be internally consistent and still wrong about the decision.
The Assumption That Can Invert the Answer
Of everything in Argonne's sensitivity work, one finding deserves to be read twice by anyone evaluating heavy-duty electrification.
"If vehicle fueling qualifies as working, the driver could spend more time charging than driving, causing the TCO for BEVs to increase dramatically."
— Argonne National Laboratory, comprehensive TCO quantification
This is not an argument against electrification. It is an argument for knowing which assumption your model made, because the two cases are enormously far apart and the difference is a question about your own labour agreements rather than about the technology. A depot-charged refuse fleet that fills its dwell window overnight is barely touched by this. A long-haul operation charging mid-shift on paid driver time is a completely different calculation, and a model that silently assumes charging is free time has answered the question for you.
The same logic applies to payload. Argonne notes that battery weight reduces available payload on many vehicles and that the cost "can be substantial." On a weight-limited operation that is a direct revenue loss per trip; on a volume-limited one it may be nothing at all. Which of those you are is a fact about your freight, and no generic model knows it.
Bring the model somebody showed you
We will go through it against these eight questions — not to talk you out of it, but so you know which two or three assumptions are carrying the conclusion. In most models it is fewer than you would expect, and they are rarely the ones being discussed.
Eight Questions to Ask Any TCO Model
Including one built by a vendor, and including ours. A model that answers all eight clearly is worth trusting; one that cannot answer three of them is a marketing asset.
- What electricity price did you assume, and does it include demand charges?An energy-only rate understates the bill on a depot that charges everything at once. If the model has one cents-per-kWh figure and no peak, it is missing a cost that behaves nothing like fuel.
- Is charging time paid time?Argonne's sharpest finding. If fuelling counts as working, a driver “could spend more time charging than driving, causing the TCO for BEVs to increase dramatically.” Ask the question before you accept the answer.
- What happened to payload?Battery weight displaces cargo. Argonne notes many vehicles lose available payload and that the cost “can be substantial” — a revenue effect, not a cost line, which is why it slips out of cost models.
- What residual value, and on what evidence?Argonne modelled depreciation by regression on real used listings and found BEVs and PHEVs holding value better than conventional counterparts in recent years. A model using a flat guess is choosing its own conclusion.
- Does it assume a battery replacement?Assuming one adds a large cost that may never occur; assuming none removes a risk that might. Either is defensible if stated. Neither is defensible if hidden.
- How is infrastructure amortised, and over how many vehicles?Depot charging is largely a fixed cost. Spread over three trucks it looks ruinous; over thirty it disappears. The vehicle count in the denominator is doing more work than any per-mile figure.
- What discount rate?Argonne uses 3% for businesses and 1.2% for households, with real loan terms of 4% over 5.25 years. A model that discounts differently is comparing a different future.
- What utilisation?Every per-mile advantage scales with miles. A truck doing 30,000 miles a year and one doing 90,000 reach opposite conclusions from identical inputs.
The One Component With a Defensible Published Number
Maintenance is the area where independent modelling has produced figures specific enough to use as a starting point.
Argonne's analysis found that electric and electrified powertrains have lower maintenance and repair costs than internal combustion across all vehicle sizes relative to vehicle price, and it published scaling multipliers. For medium- and heavy-duty vehicles, relative to a baseline conventional truck at 100%, hybrids come in at 87%, plug-in hybrids at 83%, and battery electric and fuel cell vehicles at 60%.
That 60% figure is more useful than most because it comes from an independent laboratory rather than a manufacturer, and because it is expressed as a ratio rather than an absolute — which means it travels between fleets far better than a dollars-per-mile number would. It is still a modelled figure rather than your figure, and the gap between them is exactly what a year of your own recorded work orders would close.
Note what the multiplier does not include: the body and the chassis. On a refuse packer or a boom truck, a large share of maintenance belongs to equipment the powertrain change does not touch, so a 40% reduction on the propulsion share is a considerably smaller reduction on the whole vehicle. Apply the ratio to the part of your spend it actually describes.
How FleetRabbit Handles This
Four things, and all of them exist to replace an assumption with a measurement.
Argonne's MHDV repair multiplier puts BEV at 60% of ICE. Useful as a starting point — but a year of your own logged work orders beats any published figure for your own model.
Cost per mile by powertrain, from meter readings and fuel records rather than from an assumed tariff, which is the input most models get wrong.
Days out of service per unit per year. It rarely appears in TCO models at all, and on a fixed-route operation it is often the largest real difference.
Diesel, electric and CNG on the same system means the comparison uses your fleet as its own evidence instead of borrowing someone else's assumptions.
Residual Value, and Why It Is Not a Guess Any More
Residual value used to be the weakest link in any electric TCO model, because there was no used market to observe. That is changing, and the method matters as much as the number.
Argonne derived depreciation for medium- and heavy-duty vehicles by regression on actual used vehicle listings, modelled as a function of vehicle type, age and mileage — and found that both plug-in hybrids and battery electric vehicles have held their value better than conventional counterparts in recent years. That is an observation from a market rather than a projection from a spreadsheet, which puts it in a different class of evidence from most residual assumptions in circulation.
It also cuts against the intuition many fleets still carry, which is that an electric truck will be worthless at trade-in because of battery uncertainty. Worth checking what your own model assumed here, because residual sits at the end of the calculation where nobody looks and carries more weight than its prominence suggests.
Once you have a defensible view of the assumptions, running your own numbers is the next step — our EV fleet TCO calculator takes your fleet data and produces a comparison and payback period, and it is considerably more useful once you know which inputs are doing the work.
Questions Finance Teams Ask
Why do published EV TCO figures vary so much?
Mostly because they include different things. Argonne's review found that very few studies make original detailed estimates of the main non-energy operating costs — insurance, maintenance and repair — and that most focus on purchase and fuel, which does not capture the real difference. Two models with different component sets are not disagreeing, they are answering different questions. Build one from your own data.
Is electric actually cheaper to maintain?
Argonne's modelling says yes on the powertrain: for medium and heavy duty, battery electric sits at 60% of a conventional truck's repair cost scaling, against 87% for hybrids. But apply that to the share of your spend it describes — on vocational vehicles the body and hydraulics are untouched by the powertrain change. Work out your own split.
Should we include battery replacement?
State your assumption explicitly either way. Assuming a replacement adds a large cost that may never materialise; assuming none removes a risk that might. What is not defensible is leaving it unstated, because a reader cannot tell which model they are looking at. Model both cases.
How much does charging time really matter?
Potentially enormously, and it depends entirely on your labour arrangements rather than on the vehicles. Argonne found that if fuelling counts as working time, a driver could spend more time charging than driving and the TCO increases dramatically. Depot-charged, fixed-route operations are largely insulated; mid-shift charging on paid time is not. Check which case you are.
What about payload loss?
It is a real cost and it is missing from most models because it reduces revenue rather than increasing expense. Argonne flags it as potentially substantial. Whether it affects you depends on whether your operation is weight-limited or volume-limited — a fact about your freight that no generic calculator can know. Record your actual loads.
How long before we know if we were right?
About a year of real operation, provided you captured the data from the first vehicle. Fleets that start tracking maintenance, energy and downtime by powertrain on day one can answer the question from evidence; fleets that did not are still arguing about assumptions three years later with a yard full of trucks. Set the tracking up now.
Can we compare across diesel, electric and CNG at once?
Yes, and you should, because the honest comparison is rarely two-way. Running all three on one record means the cost per mile, the downtime and the maintenance spend come from the same definitions applied the same way — which is the single biggest source of error when comparisons are assembled from separate systems. Start capturing costs free.
Replace the Assumptions With Your Own Numbers
FleetRabbit captures maintenance, energy and downtime per unit across diesel, electric and CNG on one record — so a year from now the TCO question is settled by your own history rather than by whose model you found most convincing.
Free for 3 vehicles · no credit card required
