Corrections

When we get something wrong we do three things: fix the page, put a dated note at the top of it, and record it here. We do not quietly edit.

Where an error affected a figure someone might have acted on, the entry says what the figure was, not only what it should have been.

Log

9 October 2026 — we said the best export rate open to anyone was 12p. It is 6p.

Affected: every SEG tariff, payback, ROI and IRR, the break-even price, are solar panels worth it, self-consumption, the rooftop calculator, the SEG dataset page, all 1,356 place pages, and The State of UK Solar 2026, which we published this morning.

What was wrong. Our export table is built from Ofgem’s annual report, which is published in arrears. Every row was accurate for the scheme year it covers, and the pages said so. But several pages went on to use the best figure in it — 12p, Scottish Power’s SmartGen — as the best rate a reader could switch to today, and the calculators defaulted to it.

Scottish Power’s own page now shows SmartGen at 6p, with 12p moved behind being one of their electricity customers and 15p behind having them install the system. We checked the four largest suppliers against their own published terms: the best rate open to anyone, whoever supplies their import, is 6p.

What it affected, specifically. The rooftop calculator defaulted to a 12p export rate. The payback and system-size tables used it as the optimistic column. The ROI page’s headline range of “below zero to 9.4%” was built on it, and is now below zero to 7.9%. Anyone who ran those numbers was shown a better case than the market offers.

What we changed. The best open rate is now computed at build time from tariffs that are open, domestic, and not wholesale trackers — so it follows the data rather than a number typed into a template. Rows checked against a supplier’s own terms are marked verified-page with the date and the URL; the rest stay labelled as Ofgem baseline.

What we found while fixing it, which is worse than the error. There is now a third category of export tariff that Ofgem’s reporting does not capture: tariffs that require that supplier to have installed your system. Scottish Power at 15p, E.ON Next at 17.5p, OVO at up to 20p. It cannot be satisfied retrospectively. We had been describing the market as tied against untied, which understated how closed the top of it is.

How we avoid it next time. The monitor we built watches Ofgem’s page, and Ofgem had not changed — the suppliers had. A monitor pointed at the publisher of an annual report cannot detect a market moving between reports. The fix is the file this correction produced: data/seg-verified.yaml, a dated record of what each supplier’s own page says, with a rule that a supplier is either wholly checked or wholly baseline.

9 October 2026 — five pages called a superseded price “current”

What was wrong. The energy price cap changed on 1 October 2026. From that date until this correction, five live pages asserted the old rate as the present one without a date attached: the rooftop calculator and self-consumption each said “currently 26.11p under the price cap” in one place and “26.11p under the current price cap” in another, and the break-even page said “Today’s price cap unit rate is 26.11p”.

The correct figure from 1 October is 26.32p.

Why we are logging it rather than calling it a routine update. Most of this site’s other mentions of the rate named the period — “the price cap unit rate for 1 July to 30 September 2026” — and those were never untrue; they were simply about a period that had ended. These five were different. “Currently” and “today’s” are claims about the present, and from 1 October they were false.

What it affected. The figure was 0.8% low, so nothing anyone calculated from it was materially wrong. What was wrong was the assertion that it was current.

Why it took eight days. We built a register of dated claims on 23 August precisely so that scheduled changes would not slip, and the cap was the first row in it. Nobody opened it. The register has been annotated to say so.

Fixed. 9 October. The rate now lives in a single data file and is rendered by a shortcode, so the next cap change is one edit rather than thirteen. The pages that name the period get the period from the same file.

How we avoid it next time. Two changes. The figure has one home instead of thirteen. And the register now lists the figures derived from it by hand — the battery arbitrage tables and the like — because those do not update when the data file does, and a half-updated page is worse than a wholly stale one.

21 August 2026 — council pages said no regional cost figure existed

What was wrong. Every one of the 361 local authority pages published on 20 August showed the UK median installed cost with a note reading “no regional figure is published for this area, so the UK median is shown”. For English, Welsh and Scottish councils that was false: the DESNZ cost series does publish a figure for their region, and our own dataset holds it.

Why. A mapping bug in the page generator. Town pages took their region from one field and council pages carried theirs under a different one, which the generator did not read — so councils came through with no region at all, fell to the fallback branch, and printed the fallback’s explanation.

What it affected. The cost figure shown was the national median rather than the regional mean, and the accompanying sentence was untrue. No page showed an invented number; it showed a real number with a wrong explanation attached.

Fixed. 21 August. Council pages now carry their region, show the regional figure where one exists, and the regional summary pages — which had silently been built from towns only — now include councils, taking them from 11 to 13.

How we avoid it next time. The fix carries a comment in the generator explaining what the two fields are, because the bug was invisible: nothing failed, and every page rendered a plausible sentence.

17 August 2026 — our calculators overstated self-consumption, and therefore the saving

Affected: the rooftop calculator and the plug-in solar calculator.

What was wrong. Both calculators asked you what share of your generation you would use yourself rather than export, and both defaulted to a range we had chosen: 25–60% on the rooftop calculator and 50–90% on the plug-in one. Both were labelled as our assumption rather than a sourced figure, which was honest as far as it went. It did not go far enough.

When we read MCS’s guidance note MGD 003 properly, we found the authoritative UK method gives substantially lower figures. For the rooftop calculator’s own default scenario — a household using 2,500 kWh a year with a 3,676 kWh array — MGD 003 gives 23% if someone is home all day, 19% if the property is empty half the day, and 14% if it is empty on weekdays. The bottom of our old range sat above the top of theirs.

What we changed. The rooftop calculator now defaults to 14–23% and the plug-in calculator to 40–56%, both taken from MGD 003 for the relevant generation and consumption, and both labelled with where they come from. We also published the page explaining the method and why the figure matters more than anything else on a quote.

Why it matters. Self-consumption is the single biggest lever on what solar is worth, because a unit you use is worth roughly six times a unit you export. Our old defaults made systems look better than the evidence supports. The corrected figures reduce the headline saving our own calculator reports, which is the direction an honest correction usually goes.

17 August 2026 — nine pages cited a superseded version of MIS 3002

Affected: the plug-in solar cluster and both calculators.

What was wrong. We cited MCS MIS 3002 as issue 5.0 throughout. It is now issue 6.0, dated 18 March 2026, published at a different address. The old PDF is still online and still resolves, so nothing broke and nothing announced itself.

What we changed. Every citation now names issue 6.0 with its internal date, and we re-read the standard rather than assuming continuity.

What we did about the cause. This produced rule 7: cite the issue number and the document’s own internal date, and trust the number printed inside the document over the filename or the folder. We had also been citing the 2019 issue of MGD 003, which sits at a newer-looking URL than the current 2022 issue — the same trap, twice, in one afternoon.

17 August 2026 — our link to the G98 standard had quietly stopped working

Affected: do I have to tell my network operator about plug-in solar?

What was wrong. Our citation pointed at an Energy Networks Association URL that had begun redirecting to a generic publications index. The link resolved, so no checker flagged it, but it no longer reached the document.

What we changed. Citations now point at the freely downloadable copies on the GB Distribution Code site, where we read the documents, with ENA’s own database named as the publisher’s index and flagged as requiring registration.

17 August 2026 — an arithmetic slip repeated on two pages

Affected: plug-in solar or a proper roof system? and the rooftop calculator.

What was wrong. We gave annual generation for a 4 kWp array in London as 3,674 kWh. Our own published dataset gives 919 kWh per kWp, so the figure is 3,676 kWh. The error came from multiplying an unrounded intermediate value instead of the number we publish, so our prose disagreed with our own calculator by 2 kWh.

What we changed. Corrected to 3,676, and the range across our 25 locations from “2,770 to 4,169” to 2,768 to 4,168.

Why we are logging something this small. Because the whole claim of this site is that you can check our working. A figure that does not reconcile with our own published dataset undermines that whether the gap is 2 kWh or 200.

15 August 2026 — we framed the 800 W limit as per circuit before that was settled

Affected: the plug-in solar cluster.

What was wrong. We described the 800 W limit as applying per circuit. The government response is more conditional than that: the change is stated to be subject to a corresponding amendment to G98, and until that amendment is made the existing requirement limiting installations to one device per household continues to apply.

What we changed. Every affected page now gives the one-per-household position as current, with the per-circuit change described as pending the G98 amendment, and a dated status box saying so.

Why it matters. Someone acting on the earlier framing could have installed a second device believing it permitted.

Caught before publication

These did not reach you, so they are not corrections. We mention the category because the checks that catch them are the same ones described in our methodology, and because a corrections page that only lists what escaped gives a misleading impression of how often we are wrong.

Recent examples: a dataset minimum quoted as 181 kWh/kWp when the true figure was 157; a year-to-year variation range given as 2.4–3.7% when the data said 2.4–4.1%, attached to a tidy explanation about latitude that the data did not support; and a table that rendered a headline figure as zero because of a template bug. All three were caught by checking prose claims against the underlying data before shipping. They are recorded in the changelog.

Reporting an error

Please include a link to the page and the specific claim you think is wrong. If you have a source, even better.

Report a correction

What counts

We log:

• factual errors — wrong figures, dates, rules or names of schemes
• material omissions that made a page misleading
• calculations that produced the wrong result
• broken or misattributed sources, including citations to superseded versions

We do not log typographical fixes, rewording, or routine updates where a figure changed because the world changed rather than because we got it wrong. Those appear in the changelog.