Tested, not asserted
We bring you evidence that tells you what is wrong, why we think it is wrong, by how much, and how far our evidence can be trusted. We show outputs and how sure we are, never the inner workings.
Forecast against what happened
No 2025 value enters the calibration, and an automated test enforces that rule.
| Market | Measure | Actual 2025 | Signals 2025 | Error | Tolerance |
|---|---|---|---|---|---|
| Jet fuel | Airline fuel, Mt (IATA) | 315 (est.) | 311 | −1.4% | <4% |
| Jet fuel | Jet/kerosene, mb/d (IEA) | 7.79 | 7.73 | −0.8% | <1% |
| Aircraft | Active jets (IATA) | 27,435 | 27,531 | +0.3% | <1% |
| Aircraft | Narrowbody | 18,495 | 18,343 | −0.8% | <5% |
| Aircraft | Widebody | 5,869 | 6,063 | +3.3% | <5% |
| Aircraft | Regional jet | 3,071 | 3,125 | +1.8% | <5% |
| Fuel mix | SAF produced, Mt (IATA) | 1.90 | 1.84 | −3.1% | <10% |
| Fuel mix | SAF share of jet fuel, % | 0.60 | 0.59 | −1.3% | <10% |
| Fuel mix | EU SAF price, $/t (EASA) | 2,175 | 2,001 | −8.0% | <10% |
| Capacity | SAF plant capacity, Mt/yr | 5.18 | 4.92 | −4.9% | <10% |
Largest gap: the EU SAF price, 8% low. Supply of the competing renewable diesel tightened in 2025 in ways the 2024 data could not show. We publish it because a buyer should see where a model is weakest.
How we grade every statement
Established
Peer-reviewed and corroborated, the reference assessment for the field, or a law or specification in force.
Published
A single peer-reviewed study or an authoritative dataset, or our result tested against independent data.
Derived
Our own calculation from graded inputs, with the method shown.
Hypothesis
Plausible but untested. Always worded as "we are testing whether", never stated as a result.
An argument takes the grade of its weakest statement. A correct calculation on a hypothesis is still a hypothesis.
Ten ways to check us
- A published test record. Each year we forecast a year using only earlier data and compare it with what happened, misses included.
- Graded sources. Every figure carries a grade from A to D.
- Stated assumptions. Every Signal lists the assumptions that drive it, in plain words.
- Our central case and your sensitivities. The range is shown, not one number.
- Reasons, not results only. Why the verdict is what it is, in economic, physical and policy terms.
- Signposts. What would change the verdict, and when, so you can check us against events.
- Independent review. Reviewers see our method under NDA; you see their statement.
- Error bands and validation status on every fuel result.
- Consistency and corrections. One market view for all clients; Signals dated and versioned; a public corrections log.
- Accountable people. A named team behind every Signal, an independence statement, and quality review before release.
Evidence, not advocacy
We give evidence-backed assessments tailored to the questions clients ask us. We do not promote client positions or fit evidence to a view. Our interests are declared on the page, and fuel results carry their validation status.
- Bite-size outputs, verdicts, ranges and assumptions in plain words
- The test-record headline and reviewer statements
- The evidence pack and full Signal reports
- A structured walk-through for your technical adviser
- Code, model files, parameters, databases, recipe lists or tool access
The hard questions
What model do you use?
Our own analysis, built and run by our team. What matters is how it performs: we update our data sources daily, test our outputs against outcomes we have not seen, and every Signal shows its assumptions and range.
Can we see the model?
We do not share the details of our proprietary systems, but we can walk your technical adviser through the inputs, assumptions, controls and test results under NDA.
How do we know it is right?
Our test record against actual outcomes; graded sources and stated assumptions on every Signal; an audit trail of checks; and the reasons the Signal is what it is and not something else. We also tell you what would change the verdict, so you can check us against events.
Is it AI?
We use our own proprietary systems. Every Signal is built on in-house industry and subject-matter expertise, engineering, physics and chemistry. Like most firms today, we use AI to support our research and internal workflows.
What data do you use?
Public, licensed and proprietary sources, each graded for quality. The public sources behind your Signal are listed in your report.
How is this different from a data vendor?
We answer the questions behind your decision; we are not a data query service. You get a tested verdict and recommendations, with a range and the reasons, from fuel chemistry to market and capital, from one team of experts.
Get the validation note
The 2025 test, the full scorecard with its misses, and the grading scheme. Shared on request.