01Claims are true somewhere
A data-sheet efficiency, a pilot-plant yield or a projected cost is usually accurate for the conditions under which it was produced. The diligence question is whether those conditions match the intended duty: the same scale, the same operating profile, the same feed, the same integration and the same point in the equipment’s life. Most material gaps sit in the difference between the conditions of the evidence and the conditions of the project.
02Five ways claims drift from reality
| Gap | Typical form | What to ask for |
|---|---|---|
| Boundary | Stack efficiency quoted as system efficiency; LHV versus HHV; auxiliaries excluded | An energy balance with the boundary drawn and every parasitic load listed |
| Scale | Laboratory or pilot results extrapolated linearly | The scale-up basis: which mechanisms change with size and how they were tested |
| Duty cycle | Steady-state test data for a variable-load application | Performance and degradation under the intended cycling profile |
| Life | Beginning-of-life performance used for lifetime economics | Degradation rates, replacement intervals and end-of-life performance |
| Integration | A component proven alone, assumed to perform inside a system | Interface definitions, integration test results, dependencies on other suppliers |
Each of these can move a project’s economics more than the headline figure itself. An electrolyser quoted at stack level can look markedly more efficient than the system that has to be bought. A battery modelled at beginning-of-life capacity can overstate lifetime throughput materially. A process proven at pilot scale can meet new heat-transfer, mixing or distribution limits at full scale.
- Claim or assumption
- Evidence provided
Each finding names its evidence
- Supported
- Supported with qualifications
- Open — evidence not available
- Not supported by the evidence provided
03A worked example: one efficiency, three meanings
Consider an illustrative electrolyser offered with a single efficiency figure. The same machine can be described on a higher-heating-value or a lower-heating-value basis for hydrogen, and the two give noticeably different percentages for identical performance. Measured at the stack, it excludes rectifiers, compression, water treatment, cooling and controls, all of which add to the electricity the system actually draws per kilogram.
None of these numbers is wrong. They describe different boundaries. A cost model that silently mixes them can understate electricity demand, and therefore cost, by a margin larger than most of the other uncertainties in the project. A diligence finding on this claim would record the basis, restate it at the project boundary, and mark any unquantified parasitic loads as open.
04Maturity is plural
Technology readiness levels are useful but incomplete. A component can be at high readiness while its integration with the rest of the system is unproven, and both can be ready while the supply chain to manufacture it at volume is not. Assessing technology, integration and manufacturing readiness separately prevents a single impressive number from standing in for the whole proposition.
05Evidence positions, not verdicts
For every material claim, the review assigns an evidence position: supported; supported with qualifications; open, where the evidence needed is not available; or not supported by the evidence provided. The position describes the evidence, not the technology. An open claim may well be true; it has simply not been shown.
Evidence also has a hierarchy. Independent test data under representative conditions outranks vendor testing, which outranks models calibrated to tests, which outrank uncalibrated models and stated expectations. Inference drawn from evidence can be reasonable and should still be labelled as inference.
06What would have to remain true
The most useful output is often not the list of findings but the short list of conditions the opportunity depends on: the efficiency that must hold at system level, the degradation rate that must not be exceeded, the feed specification that must be met, the integration that must work first time. Each condition is tied to the evidence that supports it and to the verification that would close the gap, ranked by its effect on the decision.