Changelog

This is the record of substantive change: new sections, data updates, and changes to how the site works or what it promises. Factual errors that reached you are recorded separately, in the corrections log.

Each living page also carries its own change history, so you can see how a particular answer has moved over time without reading this whole list.

2026

9 October, evening — later

Solar for business and land opens with two pages, both of which exist because the usual advice is wrong in a specific, checkable way.

Capital allowances: solar does not qualify for full expensing. Commercial solar marketing says it does, more or less universally. Full expensing is a 100% first-year allowance for main rate plant; HMRC’s rates-and-pools guidance lists solar panels among the items on which only 6% may be claimed, which puts them in the special rate pool. The route to 100% in year one is the Annual Investment Allowance — £1 million, available to sole traders and partnerships as well as companies, and claimable on special rate expenditure. HMRC’s own manual settles it by working an example on solar panels.

Most businesses end up in the same place, which is presumably why nobody corrects it. It stops being a footnote the moment you spend over £1m, are not a company, or are buying second-hand equipment.

Business rates: the exemption expires in 2035. English onsite solar and its storage have been outside the rating valuation since 1 April 2022. The instrument’s own end date is 31 March 2035 — partway through the life of anything installed now, and absent from every commercial model we have seen. Scotland uses a relief instead, with a main-use condition that excludes most rooftop arrays and a community benefit test for the 100% band.

What is deliberately not there. Eleven more pages in the plan, including the one most asked for: what a landowner is paid to lease land for a solar farm. No published series exists that we can trace — only figures quoted by parties to the transaction — so the hub names the gap and asks anyone who knows a source to send it.

9 October, evening

The export table now carries checked rates, not just Ofgem’s baseline. Four suppliers read from their own published terms — Scottish Power, E.ON Next, EDF and OVO — in a new hand-maintained overlay, data/seg-verified.yaml, merged into the generated table at build time. A supplier is either wholly checked or wholly baseline; Octopus and British Gas stay on the baseline because their public pages do not give a complete line-up, and the file records why.

Two findings, and the second is the bigger one.

The best rate open to anyone has halved, from 12p to 6p. Scottish Power’s SmartGen was the best open rate in Ofgem’s report at 12p; their own page now shows 6p, with 12p behind being an SP electricity customer. A correction is logged — several pages were using the 12p figure as a rate a reader could switch to today, including the report we published this morning.

There is a third category of tariff that Ofgem’s reporting does not capture. Not just tied and untied: there are now tariffs that require the supplier to have installed your system — Scottish Power at 15p, E.ON Next at 17.5p, OVO at up to 20p. That condition cannot be met retrospectively, so the highest advertised rates in the market are permanently unavailable to anyone whose panels are already on the roof and were fitted by somebody else. The dataset now has an access column with four values instead of a tied/untied boolean.

The consequence for advice. Switching export tariff used to take a typical household’s payback from about twenty years to about twelve. On today’s open rates it takes it from about twenty to about eighteen. It is still free and still worth doing; it is no longer the lever it was, and we had been recommending it as though it were.

9 October, later still

The quote checker — paste the numbers off a solar quote and have them tested against sources you can go and read. The plan has called this the site’s strongest single differentiator since Sprint 0; it needed the cost dataset to exist first.

What it checks, and against what:

• Cost per kW against the DESNZ median and mean for your size band
• The capacity against the panel count times the panel wattage, because those three numbers disagree more often than you would expect
• The generation estimate against an unshaded, due-south, 35° array at your location — a ceiling no real roof beats, so an estimate above it is a question
• The saving figure worked backwards into the self-consumption it assumes, against the 14–23% MCS’s MGD 003 gives. This is where an optimistic quote usually hides, because nobody checks it
• The battery against nothing, and it says so — no UK price series exists

Nothing is sent anywhere. No submission, no storage, no logging. A quote checker that harvests quotes is a lead generator in disguise.

A flaw found in testing, before publication. A quote that bundles a battery into one total was having its whole price divided by the panel capacity and compared with a solar-only benchmark — which inflates the figure and would have flagged perfectly normal quotes as expensive. It now subtracts an itemised battery price first, and where the battery is not itemised it refuses to make the comparison at all and tells you to ask for it separately. Twelve input scenarios were run through the finished tool to check each branch fires correctly.

9 October, later

The State of UK Solar 2026 — the first annual report, written entirely from the six datasets we publish. Every figure names the file it came from, and 31 of them were reconciled against those files by script before publication.

The findings we did not expect:

• Every one of the ten highest-paying export tariffs is tied — available only if you move your import account to the same supplier. The best rate open to anyone is 12p against a best tied rate of 27p, and a median untied rate of 4.1p. The advertised top of the export market is not addressable by most people who have panels.
• The middle half of UK places sit 8% apart, between 864 and 936 kWh per kWp. The 54% figure everyone quotes, us included, is Eastbourne against the Western Isles — not a choice anybody is making. Turning the same roof from south to north costs 46%.
• A record 265,682 certified installations in 2025/26, more than ten times the 2017/18 low, with no subsidy equivalent to the Feed-in Tariff in place. Energy prices did that.
• Five of our fourteen shading scenarios lose more than three times as much in winter as they do across the year. The annual figure is the flattering half.

Section 9 is the list of things the report cannot tell you — battery prices, degradation, installer quality, any check of modelled generation against metered output. We would rather publish the gap than fill it.

One error caught in verification, before anyone saw it. A draft sentence called 29,332 “the 2019/20 trough” in installations. 2019/20 is the trough in cost; the low in deployment is 25,565 in 2017/18, which also changes “nine times” to “more than ten”. The reconciliation script found it because the auto-generated chart subtitle and the hand-written prose disagreed. That is the argument for generating chart captions from data.

9 October

The price cap moved, and VAT came off electricity. The cap’s electricity unit rate went from 26.11p to 26.32p on 1 October. More interestingly, there is now no VAT on domestic electricity in Great Britain until 31 March 2027.

New page: VAT on electricity bills. Three things on it are not being said elsewhere:

• The headline rose 0.8%; the price before tax rose 5.8%. Removing VAT absorbed most of an underlying increase rather than handing anyone a cut. Ofgem says as much — the period cannot be compared directly with earlier ones.
• Two VAT reliefs now expire on the same day. The zero rate on installing solar and batteries, and the zero rate on the electricity it displaces. They pull in opposite directions for a buyer, and we have not seen them set beside each other anywhere.
• Holding today’s electricity price flat is no longer a neutral assumption. Every payback figure here holds the import price flat for twenty-five years, deliberately. Today’s price has a tax holiday in it that expires in March, so that assumption now leans against solar. We have not adjusted for it — building an expected tax change into a payback figure is forecasting, and we said we would not — but you should know which way the thumb is on the scale.

The rate now has one home. It was hard-coded in thirteen places: nine pages, two shortcodes and two calculator layouts. It is now data/electricity-price.yaml, rendered by an import-rate shortcode. A figure in thirteen places is a figure that goes stale in thirteen places, which is exactly what happened.

A correction, logged: five pages called the superseded rate “current” or “today’s” without a date for eight days. The figure was 0.8% low, which changed nothing anyone calculated — the assertion of currency is what was wrong.

Why it took eight days is the uncomfortable part, and it is in the correction. We built a register of dated claims on 23 August so that scheduled changes would not slip. The cap was the first row. Nobody opened it.

23 August, later

Every dataset we hold is now published, and the data section has charts.

Four new downloads, taking the total to six datasets and 2,454 rows: installed costs (336 rows, thirteen years), every SEG export tariff (50), Warm Homes: Local Grant allocations (73), and shading loss factors (14). The 625-value generation dataset now has a page of its own rather than living inside a guidance page.

Ten charts, drawn as inline SVG with no JavaScript and no chart library, so they work with scripting off, in the printed page and in either theme. The geometry is precomputed by a script rather than calculated in a template — a chart that is subtly wrong is worse than no chart, and template arithmetic is where subtle wrongness lives.

A basis error caught before publication. The first version of the cost CSV carried a single header saying every figure was on the total installed capacity basis. It is not: the department publishes its long back-catalogue on declared net capacity and its recent tables on TIC. Plotting the two on one axis invents a price fall that never happened. Every row now carries a basis column, the header says so twice, and the history chart labels it on its face. The pages that already used these figures had it right; the file we were about to hand out did not.

Two labels were being cut mid-word in the shading chart, so it read “all building due south” instead of “Tall building due south”. Found by the spell check, which is not what that gate is for and is the second time this month it has earned its keep.

A new gate. verify-downloads.js refuses a build where a published dataset has no licence in its header, no attribution, no source, a ragged row, or no page linking to it — and where a dataset page carries no schema.org markup. The full spelling gate also passes for the first time: it had been failing on network operator slugs nobody had declared.

23 August

The batteries section opens. Eleven pages, and the hub is no longer a placeholder: do I actually need one, sizing, chemistry, AC or DC coupled, warranties, safety and siting, a battery without solar, grid services, vehicle-to-grid, and one that was not in the plan at all — will a battery affect my FIT or SEG payments?

That last one exists because reading Ofgem’s co-location guidance in full turned up something we had not seen explained anywhere in plain English: a battery that can charge from the grid raises the question of whether exported units are eligible for payment, the guidance contains declarations that both you and your installer must sign, and on a metered-export FIT installation the wrong metering arrangement stops the export payments outright. We publish it because a household can walk into that without being told.

MIS 3012 clause 4.2.1 is the other find. MCS’s battery installation standard requires a list of things to be given to you in writing before the contract is awarded — warranty degradation, cycle and throughput terms, round-trip efficiency, the current cost of replacement cells, whether it works in a power cut, and the exit terms and early-termination penalties for any third-party control arrangement. Almost nobody knows that requirement exists. It now sits on three of our pages as the thing to hold a quote to.

A dating trap, caught. MCS publishes MIS 3012 at a 2025 address; the document’s own internal date is 22 November 2021. We cite the internal date, per rule 7 — which we wrote after being caught by exactly this in August.

VAT on solar panels and batteries is published alongside, because the battery pages kept needing it. Zero-rated when installed, standard-rated when bought over the counter, and reverting to 5% on 1 April 2027 on HMRC’s published position.

What we did not publish. Battery prices, cost by size, a brand comparison, and real degradation data. There is no UK series for any of them that we can stand behind — our solar costs come from official statistics built on MCS installation records, and no equivalent exists for storage. The gaps are named on the pages rather than filled with plausible numbers.

21 August

The place dataset is published as open data. Modelled solar yield for 1,356 UK places — annual and monthly output per kWp for 995 towns and all 361 UK local authority areas, as CSV and JSON under CC BY 4.0. One PVGIS run per place at its own postcode centroid, no interpolation, and reproducible from three scripts in the repository.

We can license it because we generated it, and because the geography underneath is ONS data under the Open Government Licence. That is the same test that keeps the product database unbuilt.

The number in it worth quoting: the national spread is 696 to 1,074 kWh per kWp — about 54% from worst to best. Turning a roof from south to north costs more than moving the length of the country, and the dataset’s own documentation says so, because a figure that travels without its meaning gets misused.

A correction, logged. Building the dataset exposed a bug: every council page published on 20 August said no regional cost figure existed for its area, which was untrue for English, Welsh and Scottish councils. Fixed, and recorded in full — including that the regional summary pages had silently been built from towns only, and now cover councils too.

20 August, later

1,356 place pages, and the finding that undermines them. Every place page carries a PVGIS run for that place’s own postcode centroid, its own monthly profile, its rank against every other place we model, and the cost published for its own region — 995 towns and all 361 UK council areas, with regional summaries and a league table.

The honest headline is on every one of them: the spread from the sunniest place we model to the least sunny is about 54%, while turning a roof from south to north costs far more. Your postcode is not the question; your roof is.

Find your network operator — the 14 GB licence areas plus Northern Ireland, with Distributor IDs so you can confirm yours from the MPAN on your bill. Operator contact details are deliberately absent: the document that lists them is four years old and 105 works.

The Warm Homes: Local Grant, and which councils are in it. This is the one scheme that pays for solar outright — Solar PV is on its list of eligible measures, a battery is permitted alongside, and delivery closes 31 March 2028. The award list is published as structured data rather than a table; we parse it, and because consortium members are named, 277 councils are covered. Every council page now says whether that authority appears.

Panels, inverters and batteries — and an explanation of the database we are not building. MCS holds the certification records for over 330,000 products and its terms prohibit copying or redistributing them, so instead of a comparison table built from someone else’s data we published how to read a panel datasheet, an inverter specification and a battery specification — the documents MCS’s own installation standard says you must be given before you sign.

A fix worth recording. The section sidebar listed every sibling page, which on a thousand-page section meant 107 kB of links on every page. It is now capped, which took the built site from 191 MB to 76 MB.

20 August

New builds cluster opened. The Future Homes Standard, read from Approved Document L Volume 1, 2026 edition and SI 2026/335 rather than from coverage of them: the 40% ground floor area rule, the timeline and key dates, and what to ask when buying a new build.

Two things the coverage generally gets wrong, and this cluster does not. The 40% is not a coverage requirement — it describes a notional array on a south-facing 45-degree unshaded roof, and the rule is to match that array’s annual output, so a worse roof needs more panels rather than fewer. And the transitional arrangements matter more than the headline date: a home whose building control application went in before 24 March 2027 and which commenced before 24 March 2028 may lawfully have no solar at all.

Three more new-build pages, on the questions the regulations leave open: who owns the panels, whether the developer’s system is any good, and upgrading one. The useful number in there is 3.68 kW — the largest registered capacity inside the simpler G98 network process, and below the target capacity the 40% formula produces for most houses. What inverter a developer fits therefore decides how much you can add later.

Five glossary terms added with the sources read and dated: Future Homes Standard, requirement L3, BREL report, ground floor area, higher-risk building.

A gap closed rather than left open. Solar on new build homes said in August that we could not describe the building regulations requirement because we had not read the documents. We have now read them, and the page says so and links to the cluster.

20 August

The costs section opens, with real prices. For months this site has declined to print a solar price because we had no source we trusted. We now do: Solar photovoltaic (PV) cost data, an official statistics series that DESNZ derives from the MCS Installation Database — prices actually paid for certified installations, not prices advertised, published under the Open Government Licence.

Four pages built on it: the hub, cost per kW, thirteen years of price history, what is actually in a quote, and the costs that are not in the headline. Every figure on them is generated from the dataset by a shortcode rather than typed, so they cannot drift out of step with it.

Three things the data shows that the folklore does not. Solar is barely cheaper than it was in 2013 in cash terms — £1,897/kW then, £1,780/kW now, having peaked 53% higher in 2023/24. Bigger is cheaper per kW, because much of the price is fixed rather than per-panel. And new build installations cost about 17% less per kW than retrofits, which is the evidence for a claim we had previously made without it.

Cost by system size, eleven pages, from 2 kW to 20 kW. Generated by a script so the prose is per-size and the numbers come from the datasets at render time — no page hard-codes a price.

Payback, with the assumptions visible. The payback page shows a grid rather than a number: nine years to twenty-six on the same 4 kWp system, depending only on self-consumption and export tariff. Are solar panels worth it takes the same arithmetic and says plainly which households it does not work for.

Four new calculators, all running in your browser with nothing sent to us: how many panels do I need, roof area to system size, export earnings and battery sizing. The export one ranks every open untied tariff against your own surplus — on a typical 4 kWp system the gap between the best and the median rate is over £200 a year.

The battery calculator deliberately stops short of payback, and says why: our cost pages rest on an official statistics series for solar, and no equivalent exists for storage. It does the sizing arithmetic — bounded by daily surplus and by evening demand — and leaves the number we cannot source unprinted.

The economics set, modelled rather than asserted. ROI and IRR puts a median-priced 4 kWp system between −0.8% and 9.4% a year over 25 years — the same panels, the same roof, the same price, with only self-consumption and export tariff changing. Solar against a savings account works the comparison through, including the point most versions miss: avoided spending is not taxed the way interest is. What price makes solar stop making sense asks the question the industry never asks, and finds that a low self-consumption household on the median export rate is already below break-even at today’s cap. Does solar reduce your bill or your consumption explains why the bill falls by less than the generation suggests.

The new builds cluster is finished — self-builders and the heat pump package complete it, apart from a developer-by-developer comparison that stays unbuilt until specifications can be sourced.

How to pay for it, and the traps. How to pay for solar compares cash, card, regulated credit, buy-now-pay-later and mortgage borrowing on protections rather than rates — section 75 joint liability, the section 66A 14-day withdrawal right, and what you give up when a debt is secured on your home. Solar finance: the traps sets out the mechanisms, and green mortgages covers borrowing against the house.

One dated change worth knowing: buy-now-pay-later becomes regulated on 15 July 2026 under the FCA’s PS26/1. Anything written on that before this year is out of date.

VAT has a deadline. HMRC’s guidance is that installation reverts from zero-rated to 5% on 1 April 2027. On a median 4 kW system that is about £320.

The devolved standards are not the same standard. Wales, Scotland and Northern Ireland sets out four different positions: Wales requires on-site renewable generation from 4 March 2027, three weeks before England; Scotland requires nothing of the kind; Northern Ireland proposed something in 2023 and has not enacted it. This page carries more “we could not establish” than any other on the site, because gov.wales blocked us and we would rather show the hole.

18 August

Plug-in solar cluster finished. The post-legalisation pages, written from the statutory instrument, the specification and the government response rather than from coverage of them: SI 2026/848 explained, the regulatory timeline, the buying guide, mounting, why batteries stay excluded, and Germany’s experience.

Two findings worth flagging from that reading. The law names version 2 of the specification, published 16 July 2026 — not “the specification” as a living document, so a later version would not automatically become the one the regulations point to. And regulation 3 does not extend to Northern Ireland, because the instrument it amends does not apply there.

Buyer protection cluster published. Eighteen pages on the part of solar where people actually lose money: cancellation rights, section 75, what to do when an installer goes bust, checking an installer, what a good quote looks like and the rest.

Property cluster published. Seventeen pages covering the transactions and building types general guidance skips — buying and selling a house with panels, rent-a-roof leases, flats, leasehold, and the roof types.

Planning cluster published. Thirteen pages, including the 200 mm projection rule, Article 4 directions and right to light.

Two templates published as free PDFs. A conservation area application template with a worked heritage statement, and a quote checker whose location and orientation tables are generated from our own dataset rather than typed. No sign-up for either.

A build gate added after making the same mistake three times. A stray closing shortcode inside another shortcode rendered without complaint in Hugo, so the build stayed green while the markup was wrong. There is now a check that every paired shortcode opens and closes the same number of times, and it runs with the tests.

17 August

Grid connection cluster published. Five pages covering the part of UK solar that is most technical and worst served: G98 vs G99, how to apply for a G99 connection, export limiting, what happens if your DNO refuses, and what MCS is. Written from the Energy Networks Association’s own engineering recommendations, read in full.

Generation dataset published as an open download. All 625 modelled values — 25 locations, five pitches, five orientations — as a CC BY 4.0 CSV. No sign-up, no email address. The /data/ section opened to hold it.

Fundamentals cluster published. Nine pages on the things that actually decide whether solar works on a particular house: kWp vs kWh, irradiance, self-consumption, shading, roof orientation and pitch, east–west versus south, north-facing roofs, and generation by region.

Rule 7 added. Cite the issue number and the document’s own internal date, never just the URL. Added after finding we had cited a superseded standard across nine pages. See methodology and the correction.

Both calculators’ self-consumption defaults replaced with sourced figures. This materially reduced the savings they report. Logged as a correction because the old figures were live.

Citation practice changed for restricted sources. Where a document is free but behind a registration wall, or where we read a freely available copy rather than the publisher’s gated one, the citation now says so. You should know what a source will ask of you before you click.

16 August

SEG tariff table published. Every export tariff in Ofgem’s latest annual report, untied tariffs first, open before closed, closed ones struck through rather than hidden. Built from Open Government Licence material, so it is republished in full rather than summarised.

SEG change monitor started. A daily job that detects when the published tariff landscape moves. It stores a hash rather than a copy, because detecting that a page has changed engages neither copyright nor database right.

Grants hub published. What actually pays for solar, grouped by whether a scheme funds solar PV at all — including the schemes that appear on every roundup and never covered it.

Rooftop calculator published. Generation and value split into what you use and what you export, because those are worth very different amounts and that difference is what most calculators hide.

15 August

Plug-in solar cluster completed. Twelve pages covering the regulatory position, the 800 W question, G98 notification, insurance, renting, leasehold, fire safety, conservation areas, and an honest comparison with rooftop solar.

Downloadable letter templates published. Free, ungated, traceable clause by clause: a tenant’s request for consent, a leaseholder’s request for freeholder consent, and a pre-installation checklist. This became rule 6 — where a page tells someone to write a letter, we write the letter.

Generation dataset built. 625 PVGIS model runs at the loss factor MCS specifies, so that every figure on the site traces to a model we ran rather than a table we copied.

13 August

Site launched. Foundation, page templates, glossary mechanism and the trust furniture — verified dates, source lists, corrections notices. Initial policy pages published: about, privacy, terms, cookies, accessibility, methodology and sources.

Errors caught in review

Not corrections, because they never reached you. Recorded because the checks that catch them are part of how we work, and because a site that only publishes the mistakes that escaped gives a misleading impression of how many it makes.

17 August. A dataset minimum quoted as 181 kWh/kWp when the true value was 157 — we had read the last row of a file as the smallest. A year-to-year variation range given as 2.4–3.7% when the data said 2.4–4.1%, with a tidy explanation attached about northern locations being steadier that the data flatly contradicted. A comparison table that rendered its headline UK figure as zero because of a template bug. All three were caught by checking prose claims against the underlying datasets before publishing.

17 August. A page criticising others for citing superseded standards was itself carrying a citation we could not confirm was current. Rather than quietly linking to something newer, the page now states what we read, what we could not check, and why.

What goes where