Anyone tasked with writing the specification of a performance trial needs to be very clear about the answer to the question "what do I mean by works?".
If we want something to be "25% X", do we mean "no less than 25%X", "25%(+/- 1%) X ", or "on average 25%X". If we mean average, which kind of average do we mean?
Do we want a statistically significant result? How significant does it have to be? What statistical test must we apply?
When we say 25%, we are implying that there are two significant figures. If there were three, it would be "25.0%". 24.5% is 25% to the implied two significant figures, so anything over 24.5 is arguably a pass.
Under what conditions is the trial to be conducted? Who is to operate it? What will be done in the reasonably common situation where the client cannot provide the plant with feedstock to the promised specification?
From a practical point of view, if the answers to these questions are not decided in advance, they are hostages to fortune. From a strictly scientific point of view, the lack of declarations on these issues prior to test commencement invalidates the test, for technical reasons to do with the theory of design of experiments.
I see a related situation commonly in other troubleshooting and expert witness situation where no performance trial was ever carried out. Plant may have never been capable of reliably meeting consent and therefore run in a non-compliant fashion for years.
I can have great difficulty in such situations to get clients to understand that discharge consents are absolute, and non-negotiable, and every failure is a breach of the law, as I have discussed previously. Sometimes it takes more than one prosecution by the EA to make them understand.