Reconstructing realised first-of-a-kind plant performance from public payment records: a seven-year output trajectory from the NER300 programme

Anyone insuring or lending against a first-of-a-kind (FOAK) industrial plant must price a question with no public statistical base behind it: how much of the promised output will the plant deliver, and when? Nemet, Kraus and Zipperer coded 511 energy demonstration projects and reported that they could not obtain performance data for anything close to a representative sample, noting that it is striking how little is available. This note sets out a mechanism that gets round that obstacle and publishes the first time series it produces. Article 11(2) of Commission Decision 2010/670/EU pays NER300 renewables demonstration projects for energy produced, at a rate derived from awarded funding and committed output, which yields an identity rather than a proxy. Separately, the programme's knowledge-sharing regime generates Commission publications containing output figures in chart form. The JRC annual report for 2020 contains a biomethane output chart with no values on its vertical axis; the 2019 report states that output in 2018 was about 59,000 MWh. Anchoring on that single figure allows the full series to be reconstructed. The result for Verbiostraw (Germany), biomethane from straw, is a seven-year trajectory: from under 2,000 MWh in 2014 to 107,300 MWh in 2020, that is from roughly 1 per cent to 107 per cent of the output the plant was financed to deliver. Two textual statements independent of the anchor validate the reconstruction: the JRC describes 2019 output as approximately 20 per cent below the original plan, against 22 per cent in the reconstructed series, and states that the plant came to exceed its 6,250 MWh monthly target, which the series places in 2019 and 2020 only. Pixel measurements and scale are published so the reconstruction can be refuted. The conclusion is not a shortfall level but a shape. A single-year measurement would have described this plant as running at 26, 59 or 107 per cent depending only on the year chosen, and the classical design of one observation per plant in months 7 to 12 after start-up would have recorded a failure. The risk was not that the plant would not work; it was that it took six years to work, which is a distinct and separately priceable risk. This version also corrects the second plant's denominator: BEST's often-cited 40,000 t/year is its derated capacity, not the original design capacity of over 59,000 t, and the claim in version 1 that two independent routes agreed is withdrawn.