4 recorded corrections
Corrections
We are wrong sometimes. When we are, the correction is published with the same prominence as the original claim, recorded here permanently, and accompanied by an account of why the error occurred.
How corrections work
Published work is never removed and never silently revised. When a correction is made, the publication carries a dated note, the change log records what changed, and this page records why.
We classify corrections by what actually changed rather than by how embarrassing they are. A substantive correction is one where a finding or an interval moved. A figure correction is one where a published number changed but the finding did not. A clarification is one where only wording changed.
If you believe a publication contains an error, please tell us. Corrections are answered before all other correspondence.
Change log
- 30 July 2026FigureA published number changed; the finding stands.
Two interconnection queue figures in Chapter V were revised after a system operator republished its queue register with corrected milestone dates. The median request-to-energisation duration moved from 49 to 51 months.
WhyWe had used a register snapshot taken before the operator's own correction. Our source register now records snapshot dates for all queue data so that superseded snapshots are detectable.
- 30 July 2026ClarificationWording changed; no number or finding changed.
The 2026 partial-year capital composition estimate in Chapter VI was revised following additional disclosure, and the table now states explicitly that 2026 figures cover January to May only.
WhyThe original table gave the period in a footnote rather than in the column header, which several readers reasonably read as a full-year figure.
- 8 April 2026SubstantiveA finding or interval changed.
The permissive-filtering stock estimate was widened by roughly half an order of magnitude, and the accompanying claim about repeated-epoch degradation was softened. The published interval had been too narrow.
WhyNewer repeated-epoch results indicated that degradation is more gradual than the literature we had relied on suggested. Our original interval reflected a single line of evidence where two were available.
- 26 February 2026ClarificationWording changed; no number or finding changed.
The control cost model index was relabelled to make clear that it measures loss reduction per unit of expenditure rather than absolute effectiveness.
WhyThe original label invited the reading that hardware isolation is a weak control, which is the opposite of the finding. It is highly effective and expensive.
The correction process, the evidence hierarchy it relies on and our transparency commitments are described in the methodology.