What we can actually show
Fuel-saving products have a credibility problem, and it is largely self-inflicted. So here is everything we can document — operator-verified fleet trials, a third-party engine teardown, accredited emissions measurement — each with its method and its limits next to the number, including the trial where NanoEFX produced only a small, inconclusive result. The underlying documents are listed one by one in our reports & studies library.
Four rules we hold ourselves to
These are not marketing principles. They are the reason the numbers on this site are lower than the numbers you will find in older material about this product.
The method comes before the result
Every trial page states how the measurement was made, over what distance or duration, against which baseline, and who owned the data — before it states what came out.
The limits sit next to the number, not in a footnote
If a trial has no control group, a seasonal baseline shift or a sample of one vehicle, that appears on the same page as the result, in full sentences.
We publish the whole measurement series
Where a trial has several measurement points, all of them are shown and the aggregate across the trial is the figure we quote — not the best interval.
We publish weak and mixed results too
A product that improves everything is not believable. The IHS Towers generator trial produced a small, inconclusive result and has its own page, with the same detail as the favourable trials.
Twelve trials, four countries, one honest table
Sorted by nothing in particular — not by how good the result looks. The method column is the one worth reading first.
| Trial | Where | Sample | Method | Measure | Result |
|---|---|---|---|---|---|
| Waste Services | Germany · refuse trucks | 5 vehicles · 21,830 km | Own baseline · blinded · routes verified · load fell 10.5% | Consumption, l/100 km | −18.2% |
| Municipality of Frederikshavn | Denmark · municipal truck | 1 vehicle · 1,484 km | Mode-corrected model, with Scania | Consumption, l/100 km | ≈5% (modelled) |
| Alfresa (A・L Plus) | Japan · pharma logistics | 2 vehicles · 4 measurements | Year-on-year, same calendar months | Efficiency, km/L | +8.4% … +20.0% |
| Sanden Kotsu | Japan · service buses | 2 buses · 3 months | Against untreated control group | Efficiency, km/L | +7.8% |
| Toyama Transport | Japan · logistics | 1 truck · 20,422 km | Own baseline · 3 applications | Efficiency, km/L | +13.1% |
| Chuo Taxi | Japan · taxi | 1 hybrid · 13,257 km | Own baseline · 2 applications | Efficiency, km/L | +28.3% |
| Hot-spring hotel boilers | Japan · heavy-oil boilers | 2 boilers · 6 months | Year-on-year, same calendar months | Consumption, litres/month | −10.7% … −16.0% |
| IHS Towers | Nigeria · tower generators | 11 sites · 5 months | Own baseline | Saving vs own baseline | +1.7% median — no clear effect established |
| Maruun Transport | Japan · mixed haulage | 9 vehicles · 2022 → 2023 | Year-on-year, same calendar month | Efficiency, km/L | +2.9% median · 3 of 9 negative |
| Toto Kanko Bus | Japan · coaches | 4 coaches · 12 measurements | Own baseline · 2 buses new | Efficiency, km/L | −3.4% … +44.0% · no common trend |
| Nakamoto Transport | Japan · distribution | 3 trucks · 3 months | Own baseline · single application | Efficiency, km/L | +7.2% median |
| Nishitetsu | Japan · route buses | 3 buses · Feb–Jun 2020 | 12-month rolling baseline — pandemic window | Efficiency, km/L | +7.8% median — not claimed |
Behind these twelve sit 8 further named operator records from our Japanese master dataset — 43 vehicles with recalculable before-and-after figures but no trial protocol. They are a weaker class of evidence and are labelled as one, including the operator whose record contains the only negative measurement we hold.
Open the full trial overview →
All 386 individual measurements, including the 300 we cannot verify →
Evidence that isn’t a fuel figure
Fuel savings are the hardest thing to measure well in the field. These four strands are measured differently — and two of them come from parties with no commercial interest in the outcome.
Teardown verification
A Komatsu marine main engine opened after 1,875 operating hours and examined by the engine manufacturer’s own service agent: no treatment-related damage and no corrosion. On carbon deposits the inspector and the research group disagree, and the page says so.
See the teardown →Measurement methodology
How a NanoEFX trial is run, what an accredited laboratory measurement adds, and what separates a credible fuel trial from a persuasive one.
Read the method →Certifications
Quality management, United Nations technology listing, patents and safety documentation — with the status of each stated plainly.
See certifications →Emissions
One accredited measurement — CO down 30.4%, NOx down 15.1%, CO₂ essentially unchanged — plus a statutory test passed after treatment, and the readings we refuse to quote.
See the emissions evidence →Testimonials
Recorded statements from operators and partners, including video, with the context of who is speaking and about what.
Hear the operators →Seven kinds of evidence we hold but do not present as results
We have more favourable-looking material than what appears on this site. It is not quoted because it would not survive scrutiny — and evidence that collapses when examined is worse than no evidence at all. Most of it is withheld altogether; one dataset we publish in full and quote nothing from.
Excluded, and why
- A single measurement presented as a trial result. Several of our own older case studies lead with the best interval from a multi-application trial. On this site each trial shows every measurement point and the aggregate across the whole trial, and the aggregate is the figure we quote.
- Figures from datasets without a named operator or a documented method. Three quarters of the individual measurements we hold come from a single 2013 campaign with no operator names, no test type and no method recorded, and it shows the largest improvements in our records. What is excluded is not the data but the figures: we publish every one of those measurements, marked as undocumented and removable from the view in one click — and no number from that campaign is used as a result anywhere on this site.
- Emission readings taken in a stationary test immediately after application. Readings of “zero” or “−100%” taken in the workshop minutes after treatment appear in our records. They are not evidence of a durable effect, and we do not present them as results.
- Reports whose summary is not supported by their own data. One 2025 emissions report states reductions that its own measurement tables do not show — including gases where no usable data was recorded at all. It stays out of our published evidence until it is redone properly.
- Percentages taken from a chart with no underlying values. A historical US and Japanese truck dataset exists only as a line graph. Without absolute before-and-after values there is nothing to verify, so we quote nothing from it.
- Two of our own sources that disagree about the same three trucks. For a trial at Sagawa Express we hold an internal summary and a published case study that give different before-and-after values for the same three vehicle numbers — one of them showing an improvement where the other shows a decline. Until we know which record is right, we publish neither, and no figure from that trial appears anywhere on this site.
- Figures we cannot trace to a source we hold. Where a published figure cannot be reconciled with any dataset in our records, it comes off the site until the source is produced — regardless of how good the number looks.
If you are technically minded and want to interrogate any of this, ask. We would rather have the argument before a pilot than after one.
Measure it on your own fleet
Every trial here ran on someone else’s vehicles, duty cycle and baseline. A NanoEFX pilot measures yours — and we will help you design it with a control group, so the result stands up to the same scrutiny we apply to our own data.