The documents behind the evidence
Every figure on this site traces back to a document. This page lists all of them — the accredited measurement, the two university studies, the engine teardown, the operator trial reports and the master dataset — with what each one shows, what it does not show, and a link to where we analyse it.
We do not publish the files themselves. Several contain operator names, vehicle registrations or customer data, and two are confidential to the operator. Where a document can be shared on request, it says so. What we do publish is every number we take from them, and the reasoning.
What we hold, oldest first
Sorted by the date of the work rather than by how useful it is to us.
| Document | Date | What it establishes — and what it does not | Analysed at |
|---|---|---|---|
| Stationary exhaust readings, Shimonoseki and Mitaka | 2012 | Workshop measurements on petrol vehicles at idle, before and immediately after application. Indicative only: idle readings respond strongly to engine temperature and running time, and one vehicle shows hydrocarbons rising. | Emissions |
| JATA emissions measurement | 31 Oct 2013 | The only accredited exhaust measurement in our records: CO −30.4%, HC −23.3%, NOx −15.1% on one vehicle. Does not establish durability — one reading before, one after, 88 km apart — and shows CO₂ essentially unchanged. | Emissions |
| LTO / LTFRB emissions report, Philippines | 2013 | A vehicle that failed a statutory emissions test and passed it after treatment at an accredited centre, still passing a month later. A pass/fail outcome applied by a regulator, not a percentage. Separate readings of “−100%” in the same programme are not usable and we exclude them. | Emissions |
| Master test dataset | 22 Mar 2020 | 386 before-and-after fuel measurements, 21 manufacturers, four countries. 86 carry an operator name; the remaining 300 come from a 2013 campaign with no operator, no method and no period recorded, and we mark them undocumented throughout. | Field data |
| Marine engine compatibility and post-trial inspection study, with Hiroshima Merchant Marine Technical College, Ishida Shipbuilding and Kyowa Industries | 2022 – Mar 2024 | A Komatsu 6M108AP-1 main engine on the training ship “Hikari”, opened at Ishida Shipbuilding after 1,875 operating hours and examined by Kyowa Industries, the manufacturer’s service agent: no treatment-related damage and no corrosion. On carbon deposits the inspecting company and the research group recorded different things, and the evidence page gives both. The same report also measured fuel. We exclude every fuel figure from that series, in both directions — including one substantially higher than anything we claim — because the treatment was applied only once across an extended operating period and does not reflect the reapplication protocol we now recommend. | Engine inspection |
| Graduation study, National Institute of Technology (KOSEN), Hiroshima | 2023 – 2024 | Engine test bench at 25% and 50% load. The treated engine reads worse at the first measurement and crosses below the reference after five to ten minutes — best readings −5.9% and −2.7%. A student thesis on one engine, and the series fluctuates. | Methodology |
| Diesel consumption study, Prof. Toshifumi Chaen, Hiroshima National College of Maritime Technology | Apr – Dec 2024 | Named in our own chronology, but we do not hold a copy of this report in the evidence set. The fuel figures that circulate for this vessel programme come from the joint research group’s March 2024 report rather than from this study, and all of them are excluded from what we claim. We quote nothing from this study until the document is in hand. | Engine inspection |
| Field-testing method | undated | The protocol our trials are supposed to follow: at least three applications, three to six months of records, same routes and drivers. Useful as a standard; most of our older datasets do not meet it, which is why they are labelled the way they are. | Methodology |
| IHS Towers proof-of-concept analysis, Nigeria | 14 Nov 2025 | Eleven telecom generator sites over five months: median +1.7%, two sites worse, four recorded by the operator as showing no observable savings. Establishes a boundary — not a saving. Contains no load data, so the idle explanation cannot be verified from it. | IHS Towers |
| Case study, Municipality of Frederikshavn, with Scania | 2026 | One Scania truck over two periods. The raw figures show no saving; a drive-mode correction indicates roughly 5%. The document states its own conclusion plainly — model-based, not final documentation of effect. | Frederikshavn |
| Waste Services proof-of-concept report and weekly dataset (confidential to the operator) | 2026 | Five refuse trucks, drivers blinded, 21,830 km. Our best-documented trial. The accompanying operational records — destination, hours, distance and load per day — are what revealed that the vehicles also ran 10.5% lighter during the trial. | Waste Services |
| Safety data sheets (SDS / MSDS) | current | Eight regional and language versions covering the EU, UK, US and UN formats. Available on request — ask us and we will send the version for your jurisdiction. | Request a copy |
Documents we hold but do not use
These exist in our files. None of them appears anywhere else on this site, and this is the only page that names them.
Two 2025 fuel reports from an education centre
Carried out at EUC Nord in Frederikshavn. The measurement setup is sound; the analysis is not — the summary states reductions that the report’s own tables contradict, including gases where no usable data was recorded and one configuration where a pollutant rises sharply. We would rather commission a clean repeat than quote these.
A historical truck dataset that exists only as a line graph
US and Japanese trucks, presented as a chart with no underlying before-and-after values. There is nothing in it to verify, so we quote nothing from it.
A third-party truck performance table
A summary sheet of results from vehicles we did not measure and cannot trace to an operator or a method. It stays out.
If you are evaluating us, ask
Some of these documents can be shared, some belong to the operator, and some we would only hand over alongside an explanation of their limits. Tell us which dataset matters for your decision and we will tell you exactly what we can produce.