How this is built.

On this page

1. What is true of every source
2. Sources and licences
3. Whether it worked — the eight audit sources
4. What government is doing — the sixteen publication sources
5. The two dashboards
6. The weekly briefing
7. Glossary of terms

Open Government Licence v3.0

The licence most of British government publishes under. It permits copying, adapting and commercial use, and asks one thing in return: the sentence “Contains public sector information licensed under the Open Government Licence v3.0.” It does not cover personal data, which is the first item on its own list of exclusions.

• GOV.UK Search and Content API — consultations, news, policy papers, guidance, FOI releases, statistics, RPC opinions, departmental impact assessments, post-implementation reviews, government bodies, and the meetings and regulators sources in part.
• data.gov.uk organograms — Find an official. The licence covers the files; it does not cover the names inside them, which are handled under data protection law rather than under a licence.
• Find a Tender and Contracts Finder — procurement notices. Notice data follows the TED schema, © European Union, 1998–2020.
• Office for National Statistics — the release calendar behind the statistics source.
• planning.data.gov.uk and GOV.UK intervention collections — local government.
• Local Government Association, esd-toolkit programme — council powers and duties, attributed to the esd-toolkit programme as their terms ask.
• Charity Commission — the charity lookup behind meeting briefings. Their API terms add a duty to acknowledge the source and not to misrepresent who owns it.
• legislation.gov.uk — the Act and instrument links on post-implementation reviews. We only link; nothing is copied.

Open Parliament Licence v3.0

Parliamentary copyright is not Crown copyright and has its own licence. It is stricter about attribution than the OGL: where the OGL will accept a link to a page like this one, the Open Parliament Licence asks that the sentence “Contains Parliamentary information licensed under the Open Parliament Licence v3.0.” itself appears. It also excludes personal data and the Crowned Portcullis.

• Hansard, written questions, written answers and written ministerial statements — the Parliament source.
• The committees API — select committee reports and government responses.
• The Members and Interests APIs — the background lookup behind meeting briefings.

Not under a licence at all

One register does not sit under any licence, and it is worth saying plainly rather than leaving to be inferred from the list above. The named contacts on the contacts page — officials, buyers and elected members, with the work email addresses their own organisations published — are personal data. Both the Open Government Licence and the Open Parliament Licence exclude personal data from what they grant, so neither one is what permits us to hold this.

What permits it is UK data protection law: our legitimate interests in helping somebody reach the part of government responsible for a document they have found. That is a weaker and more conditional thing than a licence, and it comes with obligations a licence does not have — a retention period, a stated purpose, and an unconditional right to be removed. All three are in the privacy notice.

The organisations’ own enquiries inboxes on the same page are a different matter: an inbox belongs to a body rather than a person, and those are under the Open Government Licence like the rest of GOV.UK.

Licences of their own

• Find Case Law, The National Archives — the Open Justice Licence v2.0. It permits reading, searching and republishing metadata with attribution, and requires that judgments withdrawn or replaced come out of the source, which is why that updater rebuilds rather than merges. It reserves a longer list of things — text mining, classification, semantic search over the collection — to a separate free licence.
• Scottish Parliament — the Scottish Parliament Copyright Licence. Like the OGL in shape, with two prohibitions the OGL does not have: no use for party political purposes, and none for advertising endorsement.
• Senedd Cymru — unsettled. The Senedd’s open data page offers the Record of Proceedings under the Open Government Licence; its copyright page and its re-use policy point at a non-commercial licence instead. The source carries a plain acknowledgement while that is put to them.
• Audit Scotland — the Open Government Licence, with their own attribution template naming each publication and its date.
• Audit Wales — the Auditor General for Wales permits free re-use in any format, on condition the material is accurate, not used misleadingly, and the copyright acknowledged.
• NHS England — the Open Government Licence, in the form NHS England prescribes for adapted data.
• mySociety — council names and codes under CC BY 4.0, which unlike the OGL also requires us to say that we changed the data. We have.
• Wikidata — CC0. Nothing is required; the credit is given because it is deserved.
• National Audit Office — the exception on this page. NAO reports are not published under an open licence; their terms permit reproduction for non-commercial purposes only. What that means for this source is being settled with the NAO directly.

What a licence does not do

None of these licences covers personal data — every one of them excludes it in terms. Four sources name identifiable people: Find an official, the meetings source, the three parliaments, and case law. They rest on data protection law rather than on any licence, and the privacy notice sets out the basis, what is deliberately left out, and how to ask for something to be removed.

Our data platform

1. Departmental impact assessments
2. Regulatory Policy Committee opinions
3. Post-implementation reviews
4. National Audit Office reports
5. Audit Scotland
6. Audit Wales
7. Parliamentary committee reports
8. Case law

Six things to know before relying on this

1. This is the NAO's reports, not everything the NAO does. Its site files insights, briefings, overviews, work in progress and press releases as separate things, and none of them is here. A subject with no report may well have had a briefing — and a report listed as in progress today will only appear once it is published.
2. The date shown is the publication date on nao.org.uk. A report's fieldwork can run for a year or more before that, so the date says when the findings became public, not when the period examined ended. For the years covered by any given study, read the report.
3. The three kinds of report are not comparable. A value-for-money study is a piece of analysis that took months. A financial audit report is the statutory opinion on a set of accounts, published annually for every account the Comptroller and Auditor General signs, and its existence says nothing about whether anything was wrong. Counting them together tells you about workload, not about scrutiny — which is why the chart shades them differently and every filter keeps them apart.
4. The bodies named are the NAO's own tags. A report on a cross-government problem may be tagged to one department, to several, or to “Cross-government”. Departments also get renamed and merged, and the NAO keeps historical names in its index, so the same organisation can appear under two entries.
5. A gap is not an all-clear. The list of bodies with nothing recent is built by subtracting this source from the NAO's own index of bodies it has reported on. It shows where recent independent scrutiny is absent. It cannot show why — a body may be small, may be audited by someone else, may have been abolished, or may simply not have come up. The figure beside each name is the NAO's own count of everything filed under that body across its whole archive, reports and other material together, so read it as a rough measure of how much attention it once had rather than an exact report count.
6. The match box works from a hand-written list of sectors, and it hides things. It is the same lexicon as the consultation finder, the news feed, the policy paper source, the guidance source and the transparency source, so a description that works there works here. Your description is placed into one of around thirty sectors, using the phrases each one uses and the bodies whose documents usually belong to it. Directly relevant is a deliberately high bar: the report has to sit squarely in your sector and its own title has to name what you do, in your sector's words. Everything else in your area, or sharing your wording, drops to might also be worth a look. Expect fewer matches here than on the other pages — the NAO audits government, not industry, so a great deal of the source is about how a department runs rather than about any sector. Anything outside your sector and phrased in words you did not use is dropped without being named, so press Clear to see the source in full.

Four things to know before relying on this

1. The listing carries no date, so each publication has to be opened. Audit Wales shows a title, a summary, a report type and a topic on its listing — and no date. The date exists only on the publication’s own page, as a bare line of text that is not marked up as a date and not labelled “published”, which is why four sensible patterns missed it the first time. So the rebuild opens every publication in turn. That is one request each, it paces itself out of courtesy, and it is much slower than the Scottish one.
2. Its own year filter cannot be used. The dropdown on audit.wales stops at 2024 and its values are opaque numbers rather than years, so it cannot be asked for the current year at all. The alternative considered — reading a year out of the address of an uploaded image — is not the publication date and would be wrong the first time they reused a picture.
3. A gap at the end of a year is usually the walk stopping, not silence. Because there is no date on the listing, the rebuild works backwards and stops once it has seen several consecutive pages with nothing from the year it wants. That is the honest way to bound an undated list, but it means the very oldest items in a year can be missed where a page mixes years.
4. This is Wales’s auditor, not the UK’s. Audit Wales reports to the Auditor General for Wales on devolved Welsh public bodies. UK-wide departments sit with the National Audit Office and Scottish bodies with Audit Scotland.

Five things to know before relying on this

1. “A public body is a party” is not a field anybody publishes. Find Case Law records a court, a date, a citation and a case name. Nothing in it says “one side of this is the state”. So this source gets at it a different way: it takes whole benches whose jurisdiction means a public body is a party. The Administrative Court hears judicial review. The Tax Chamber hears appeals against HMRC. The General Regulatory Chamber hears appeals against the Information Commissioner, Ofcom and the Charity Commission. SIAC hears appeals against the Home Secretary. On those benches a purely private case is close to impossible. The source page lists the courts and tribunals included, and splits the judgments between them.
2. That means some obvious cases are missing, and the gap is deliberate. A Supreme Court judicial review is the most consequential public-law case of its year, and it is not here. Neither is a council in the Family Court, an NHS trust sued in the King’s Bench, or a government department in the Court of Appeal. Reaching those means reading each case name and working out whether a party is public — which is computational analysis under the Open Justice Licence and needs a separate permission from the National Archives. That permission has been applied for. The code that does the reading is written and tested; it is switched off until the licence arrives, and it refuses to run if you try.
3. Fewer of these are the government losing than you might think. A tribunal appeal is the ordinary machinery of the state being challenged — a tax assessment, a visa refusal, an information request — and most are dismissed. The source shows that the challenge was made and where it was heard. What it says about who won is nothing at all: that is in the judgment, and reading the judgment is the point.
4. The description box works less well here than on the other sources. Everywhere else it has a title and a summary paragraph to work with. A judgment arrives as a case name of six or seven words and no summary, because the feed carries none. So it matches on the name, the court and the citation, and it will miss cases that are squarely yours but named after two people. On this page the free-text search and the court filter will usually serve you better; press Clear and use those.
5. Judgments can be withdrawn, and this source has to let them go. The licence obliges anybody republishing to remove judgments that stop being published or are replaced by a revised version. This file is rebuilt from the source every day rather than added to, so a withdrawn judgment falls out on its own rather than sitting here after it has gone. Dates are the date the court first published, not the day we saw it.

Our data platform

1. Government consultations
2. Press releases and government news
3. Government policy papers
4. Guidance and regulation
5. Transparency and FOI releases
6. Parliament
7. Scottish Parliament proceedings
8. Senedd proceedings
9. Regulators
10. Procurement: tenders and frameworks
11. Local government
12. Who ministers and officials are meeting
13. Find an official
14. Government bodies
15. Contacts, and how addresses are shaped
16. NHS England publications
17. NHS bodies
18. Statistics and releases
19. Written evidence
20. Register of interests
21. Research funding

Six things to know before relying on this

1. It starts on 1 August 2026 and grows. The source is cumulative from that date, not a rolling window, so nothing already listed ever drops off. It is a snapshot rebuilt each morning rather than a live feed, so something published this afternoon appears tomorrow.
2. Bulk tribunal decisions are left out. GOV.UK files employment tribunal, employment appeal, tax tribunal and Upper Tribunal (AAC) judgments in the same feed as press releases — and there are far more of them than of everything else combined. They are case law published in bulk rather than government news, so they are excluded. Everything else GOV.UK lists is here.
3. The four categories are ours, the document types are GOV.UK's. "Press releases", "News and speeches", "Decisions and rulings" and "Letters and other" are a plain-English layer we put over GOV.UK's dozen or so document types, because most people arrive wanting an announcement or a ruling and not both. The exact type is always shown on the card, and the Document type filter works on GOV.UK's own labels. A type GOV.UK adds later lands in "Letters and other" until we place it.
4. Topics come from GOV.UK's own taxonomy. An item can carry several, or none. An untagged item is not off-topic; it simply has not been filed against a top-level topic.
5. A few items have no publishing body. GOV.UK attributes almost everything to a department or agency, but not quite everything. Those are shown as such rather than guessed at.
6. The match box works from a hand-written list of sectors, and it hides things. It is the same lexicon as the consultation finder, so a description that works there works here. Your description is placed into one of around thirty sectors, using the phrases each one uses and the bodies whose announcements usually belong to it. Directly relevant is a deliberately high bar: the item has to sit squarely in your sector and its own headline has to name what you do, in your sector's words. Over a short window most descriptions will clear it for nothing at all, and the page says so rather than promoting near misses. Everything else in your area, or sharing your wording, drops to might also be worth a look. The sector list is ours and it is incomplete: an organisation it cannot place is told so, and falls back to shared words alone. Anything outside your sector and phrased in words you did not use is dropped without being named, so press Clear to see the feed in full, and treat the body and topic filters as a second way in.

Seven things to know before relying on this

1. This is the new guidance, not the whole feed. GOV.UK's finder shows everything it has re-dated in the current year. This source is the part of that which was first published in the year; the source page gives both numbers. The older documents in that feed are real guidance and may well have changed substantially; they are simply not new, and a page that mixed them in would not be able to tell you which was which. If you are checking whether a particular piece of guidance still stands, go to GOV.UK, not here.
2. The document types are GOV.UK's, and they are uneven. Nearly half of a year's new guidance is filed under the single word "guidance", with "detailed guide" and "notice" taking most of the rest. That is the source's own filing, kept unchanged on purpose: tidier categories would be our judgement dressed up as the department's. Only statutory guidance and regulation tell you anything about legal force, and even those are a filing label rather than a legal opinion.
3. The change note is the department's, the reading of it is ours. Departments write a free-text note every time they amend a published document. The exact wording is always shown on the card. The label beside it — document added, text changed, housekeeping — is a keyword reading of that wording, and it will get some of them wrong. Where the note is ambiguous the document is labelled as such rather than guessed at.
4. The month chart measures exposure as much as churn. A document published in January has had the rest of the year to be amended; one published last month has had weeks. The lime share therefore falls as the chart goes down, and that fall is not evidence that guidance is settling. Comparing a month against the same month next year is the only fair reading.
5. An amendment count is not a measure of change. The most-amended document here has been reopened many times over; some of those will be a new annex and some a corrected link. Counting them says how often a page has been reopened, and nothing about how much of it moved.
6. Guidance is not the same thing as the rule. Almost none of what is here is legislation. It is government explaining what it expects, which is usually what people comply with in practice but is not always what a court would enforce. Where the distinction matters to you, the underlying instrument matters more than the guidance page.
7. The match box works from a hand-written list of sectors, and it hides things. It is the same lexicon as the consultation finder, the news feed and the policy paper source, so a description that works there works here. Your description is placed into one of around thirty sectors, using the phrases each one uses and the bodies whose documents usually belong to it. Directly relevant is a deliberately high bar: the document has to sit squarely in your sector and its own title has to name what you do, in your sector's words. Everything else in your area, or sharing your wording, drops to might also be worth a look. The sector list is ours and it is incomplete: an organisation it cannot place is told so, and falls back to shared words alone. Anything outside your sector and phrased in words you did not use is dropped without being named, so press Clear to see the source in full.

Five things to know before relying on this

1. This page loads one month, not one year. Every other source here holds a year and offers the ones before it. This one cannot. Parliament answered 56,923 written questions in 2026, against 4,883 debates and roughly a hundred statements a month — so a single year of it, stored as leanly as it is worth storing, is close to thirty megabytes. That is larger than the rest of this site put together. A month runs from about a thousand items in a recess to eleven thousand in a busy one, so the page loads the current month and the control above fetches the others one at a time. A search only looks at the months you have loaded, and the line above the results always says which those are.
2. Written answers are most of what is here, and they are the smallest unit. A written answer is one minister replying to one member’s question, in writing, usually within a week. It is not a policy announcement and it does not bind anybody — but it is often the only place a department says in plain words how a rule applies to a particular trade. That is why they are in here despite the volume. Where a minister answers only by pointing at an earlier answer, the row keeps the question and shows the subject heading rather than reprinting the pointer. Where several members ask the same thing and get one reply, the question appears once.
3. The debate summary is an extract, not a summary. Hansard publishes no abstract of a debate, so the line under each debate is the opening of the first substantive contribution — skipping the procedural formula (“Motion made, and Question proposed…”) that starts almost all of them. It tells you what the debate opened on. It does not tell you what was concluded, and a long debate can end somewhere its first speech does not point.
4. Hansard’s search API is not a published one. The debates here come from the service behind the Hansard search page. It works, it is stable, and it appears in no specification, on no developer hub page and in no swagger file — so unlike the questions and statements API there is nothing to check our reading of it against. If a field is renamed, the daily refresh is meant to fail rather than publish an empty source: there is a test against recorded responses that runs first, and a guard that refuses to overwrite a good month with a much smaller one. Both are precautions, not guarantees.
5. What is deliberately not here. Oral questions have their own page and the evidence submitted to committees is not in this source; committee reports have their own page under Whether it worked. Statutory instruments are not here either — they are law rather than communication. And the match box works from the same hand-written sector list as the consultation finder and the rest, so a description that works there works here; a debate or an answer has to reach your sector, or use your own words, to stay on screen. Press Clear to see the month in full.

Six things to know before relying on this

1. This is the question, not the answer — and that is the difference from the Parliament source. The Parliament source holds written answers, keyed on the day the answer was published, one rolling month at a time. It cannot show you a question nobody has answered, because until there is an answer there is nothing for it to key on. This one is keyed on the day the question was tabled, so a question appears the day it is asked and the answer is filled in when it arrives. An answered question is therefore on both, said two different ways. If you want what a minister said, use Parliament. If you want what a department has been asked and has not yet replied to, it is only here.
2. It starts in July 2024, and that is a choice. The 2024 general election was on 4 July and the new Parliament first sat on the 9th. Asking the API for anything earlier returns the tail of the previous Parliament — a different government, answering to different members, on the same page with nothing to say so. The floor is set at July 2024 deliberately, and the count on the page is the count since then, not since records began.
3. This page loads one month, not one year. Parliament was asked 65,108 written questions in the first nine months of 2026 alone. A year of them, stored as leanly as it is worth storing, is around thirty megabytes — larger than most of the rest of this site. So the page loads the current month and the control above fetches the others one at a time. A search only looks at the months you have loaded, and the line above the results always says which those are.
4. The archive holds no answers, only the questions and where to find the answers. Keeping the text of every answer since July 2024 would come to well over a hundred megabytes. Recent months carry a short extract of the answer; older ones carry the question, the member, the department, the dates and the link, and nothing else. That is a real limitation and it is the reason every row links back to Parliament, where the answer sits in full. Which months are which is recorded in each month’s file.
5. The question text is Parliament’s, and Parliament’s list cuts it. The service’s list of questions returns about the first 250 characters of each one, sometimes mid-word, with nothing to mark that it has been cut. A row that arrived that way ends in an ellipsis here so that you can see it. The full text is always one click away on Parliament’s own page, and a long question — and they run to five hundred characters and more — frequently asks several things at once, only the first of which is on screen.
6. “Waiting 12 days” is elapsed time, not lateness. It counts from the day the question was tabled to today, across recesses, weekends and dissolutions alike. Most written questions carry no binding deadline at all: the Commons expects an answer within a working week and the Lords within a fortnight, but those are expectations rather than rules. The exception is a named day question, which names the date the member wants an answer by; those are marked as such, and one that is past its date says so. Treat everything else as a description of how long something has taken, not as a finding that anybody is late.

What is deliberately not here

1. Withdrawn questions. A member can take a question back. Parliament keeps the record; this source does not, because the row would carry a question nobody is going to answer and no answer to wait for.
2. The same question asked by several members. When a group of members ask the same thing it is answered once and Parliament links them. An answered group appears here once. An unanswered group does not get folded: two members each waiting is two questions the department owes, and collapsing them would understate that.
3. Answers that only point at another answer. “I refer the hon. Member to the answer I gave on 3 June” is not a description of anything, so the row keeps the question and shows the subject heading instead. The answer itself is on Parliament’s page.
4. Oral questions. Questions asked and answered at the despatch box are a different record and are not here. Written ministerial statements are here, on the same page, because they come from the same service and a reader looking for what a department has put in writing wants both.

Five things to know before relying on this

1. A row is an agenda item, not a speech and not a day. The Senedd’s export is finer than this source: it holds every individual contribution, several hundred on a Plenary afternoon. Those are grouped back up to the agenda item the Senedd itself names, and the row links to the first contribution in it, so you land at the start of the item rather than the top of the day.
2. This is the English record. The Senedd works in both languages and publishes Welsh, English and bilingual versions of every meeting. The English export is read here, which means a contribution made in Welsh appears in translation. If the exact words matter — and in a bilingual legislature they sometimes do — the bilingual version is a click away and shows both columns side by side.
3. Private business is not in the record, so it is not here. A committee that resolves to exclude the public keeps a minute rather than a transcript. You will often see the motion to go into private session as the last item of a meeting and nothing after it. That is the record being accurate, not this source being short.
4. The kind of business is the Senedd’s word, or nothing. The type beside each date — questions, a statement, a procedural motion, an evidence session — is set only where the item’s title matches one of the Senedd’s standing forms. Anything else stays “item of business” rather than being guessed at.
5. The match box works from a hand-written list of sectors, and it hides things. It is the same word list as the National Audit Office source, the consultation finder and the rest, so a description that works there works here. It has no more to work with here than on the Scottish source: both carry the item’s title and not its words. Press Clear to see the year in full.

Eight things to know before relying on this

1. Two services, and they do not cover the same ground. Above the procurement thresholds a buyer must advertise on Find a Tender. Below them, and for most of what English councils, NHS trusts and universities buy, the notice goes on Contracts Finder instead. This source reads both, and the Service filter picks between them. It matters because the two are not the same size of thing: a Find a Tender notice is typically a national framework worth millions, while a Contracts Finder notice can be a £25,000 school taxi route. Every row says which service it came from once both are on screen, and the line above the table gives the split. The two overlap at the top end, because a large buyer may publish in both places.
2. Contracts Finder is fetched only when you ask for it. It is much the larger of the two, so the page loads Find a Tender alone and reaches for the rest the moment you touch the Service filter or search for a supplier. That is why the table pauses for a moment the first time.
3. The live register is a window; the history is separate. Find a Tender publishes about 296 notices a day, so ninety days is roughly 26,000 records and nobody is bidding for a contract that closed last autumn. The source therefore keeps ninety days and drops what falls out of the back. Earlier years are kept apart, under Year, and hold awards and contracts only — what was let and to whom. An expired opportunity is not history. Descriptions and closing dates are not kept in the archive either, which is what makes a whole year small enough to fetch.
4. The winner is named on the notice, where the buyer named one. This source only began reading that field on 5 September 2026, and a notice already on file was not re-read, so the older end of the window still says “not recorded”. Nothing is missing from Find a Tender itself — it is missing from our copy, and it fills in as the window turns over. The line above the results says how many of the notices you are looking at name a supplier.
5. A framework award names everybody appointed to it. When a buyer sets up a framework — a list of pre-approved suppliers to call off from later — the award notice can name forty firms and carry the ceiling value of the whole thing rather than any money actually spent. Ten notices account for over half the total value in the window. So read a value as the size of the arrangement, not as a cheque.
6. The five stages are derived, not given. Find a Tender publishes to an international standard that recognises planning, tender and award; “pipeline” and “contract” are British concepts on top of it, worked out here from the notice type. One error would matter more than the rest — showing a cancelled procurement as a live opportunity — so cancellations, terminations and withdrawals are dropped outright rather than matched by a rule that could catch them by accident.
7. The filters read the notice, not the contract. Sector comes from the buyer’s procurement code and Type from the box they ticked, so a mis-filed notice is filed wrongly here too. Size bands use the value on the notice, which on a framework is the ceiling of the arrangement rather than money spent. And Closing only means something on an opportunity — an award has nothing left to close, so choosing it empties the award stages by design.
8. The match is a suggestion, and it says so. A two-digit classification cannot tell a baker from a horticulturist, so a code corroborates a reading rather than making one. Where a notice reaches you on wording alone the page says so. And a sector government does not buy from will come back empty — that is an answer, not a fault.

Seven things to know before relying on this

1. This is what departments declare, and a meeting is not influence. A row means somebody’s office recorded a meeting and wrote down a purpose, usually in a few words. It does not tell you who asked for it, who else was in the room, what was said or what followed. Categories still sit outside the returns — party business, and anything a department judged not to be departmental business. Read a row as a fact about the diary.
2. The three roles are not published to the same standard. Ministers’ returns are required by the Ministerial Code and every department files them. Senior officials’ and special advisers’ returns are required by separate transparency commitments, and departments interpret them more loosely: some name every director general, some only the permanent secretary; some publish advisers’ meetings and some publish only their gifts. A role that is thin for a department is usually thin because that department published less, not because fewer people were in the room. Use the Whose return and Department filters together before drawing any comparison between two departments.
3. Nil returns are counted, not shown. A large share of what is published is a department recording that somebody met nobody, gave nothing and claimed nothing that quarter — on the officials’ tables, every named person appears every quarter whether or not they did anything. Leaving those in would have made most of this source a list of things that did not happen, so they are dropped when the source is built and counted instead. The count is kept with each year’s data. A person absent from a quarter either filed a nil return or was not named at all, and this source cannot tell you which.
4. Departments are months apart, and the gap is theirs. The returns are quarterly and published in arrears, so the most recent quarter is normally three to six months old before it appears. Some departments are much further behind than that, and one or two have stopped for a while and then caught up in a batch. The Department filter is the honest way to see this: pick one and look at the newest date, and remember that a department’s ministers, officials and advisers are three separate collections that can each be at a different stage. A department missing from a period published nothing for it.
5. Gifts and hospitality often carry no date, so they are dated to the quarter. Many of these tables give a value and an outcome but no day. Rather than invent one, the row takes the last day of the quarter it was published for and says dated to the quarter beside its kind. Sorting by date will therefore bunch those rows at the end of each quarter. Meetings and travel almost always carry a real date.
6. The spreadsheets disagree with each other, and a few cannot be read at all. The same return arrives as CSV from one department, Excel from the next and OpenDocument from a third, with the columns named differently in each. They are matched by pattern rather than by position. Files in the old binary .xls format — mostly pre-2019 — are not read, and the refresh job names every file it could not open rather than passing over it in silence. So the further back you go, the thinner this gets.
7. The match box works from a hand-written list of sectors, and it hides things. It is the same word list as the National Audit Office source, the consultation finder and the rest. Your description is placed into one of around thirty sectors, and a row has to reach that sector, or use your own words, to stay on screen. A row here is a name and a short purpose, which is less to go on than a document title, so expect fewer matches and press Clear to see the year in full.

Five things to know before relying on this

1. Closed bodies are kept, on purpose. A large share of these no longer exist, and most of those name the bodies that took over. That is the most useful thing here: a document, a statutory instrument or an old contract that names the Homes and Communities Agency is talking about Homes England and the Regulator of Social Housing, which both exist today. A directory of only live bodies could not tell you that. Use the Status filter to see live ones alone.
2. “Split” means more than one successor, and that is not a tidy answer. When a body is split its work goes to two or more places and there is no single organisation you can point at. The card names them all rather than picking one. Merged, replaced and renamed each say what they mean; where GOV.UK records a closure with no successor at all, the card says that too rather than leaving a blank.
3. It is government as GOV.UK publishes it, which is not all of government. Local authorities are not here — for those, use the local government. Nor are most bodies of the devolved administrations, the courts, or anything that has never published on GOV.UK. A body missing from this list is missing from that list, which is a different thing from not existing.
4. The contact is a department’s front door, not a person’s desk. Where a body appears in the Cabinet Office’s list, its ministerial correspondence address is shown. That inbox is staffed and read, and it is published so that people write to it. What is deliberately not here is any individual official’s work address or telephone number: those are personal data, which the Open Government Licence does not cover, and guessing at firstname.lastname@ produces addresses a departmental mail server will accept and never deliver. Only the twenty-odd departments on that list have a contact line. An agency or public body shows none, because writing to its parent department about its business is not the same as writing to it — use the body’s own GOV.UK page for that.
5. Types are GOV.UK’s own categories. “Executive non-departmental public body”, “Other” and the rest are the labels GOV.UK applies, not a classification made here. They are uneven — “Other” is large — and a body’s type says less about what it does than its parent does.

Five things to know before relying on this

1. “Active” in the source does not mean the body exists. The source marks 380 NHS trusts active; England has about 215. It marks the Health Protection Agency active, and that was abolished in 2013. The status field describes the record, not the organisation. What describes the organisation is its legal end date, so this directory reads every body’s own record and marks it closed when that date has passed. It is the reason a refresh takes a few minutes rather than a few seconds, and it is the difference between a directory and a list of names.
2. Abolished bodies are kept, on purpose. A large share of these no longer exist. A contract, a board paper or a statutory instrument written five years ago names bodies that have since merged, and a directory of only live ones could not tell you what you were reading. Use the Status filter to see the current ones alone.
3. This is the commissioning and provider tier, not every NHS address. The source also holds roughly 6,300 GP practices, every pharmacy, every dental and optical practice, every hospital site and thousands of prescribing cost centres. None of those are here. A trust with forty hospitals is one body, not forty.
4. An integrated care board’s region is its own; a trust’s is inherited. Boards record which NHS England region they sit in. Trusts do not, so a trust shows the region of the board it names. That is the only thing in this directory inferred rather than recorded, and it is one step.
5. The names are shouted in the source and quietened here. The source stores names in capitals. They are converted for reading, which is the only change made to any value on this page — codes, postcodes, dates and relationships are passed through exactly as published.

Six things to know before relying on this

1. This is not what government published. Every other source here is the state saying something. This one is the opposite: a trade body, a charity, a company or a regulator telling a committee what it thinks, usually because the committee asked. It is the best public record there is of who turns up on a subject — and it is an argument being made, not a finding. Read it that way.
2. A row is one submission, and a submission can have several names on it. Where two organisations sign one piece of evidence the row names both, the way a joint inquiry names both committees. The submission reference beside the date — BBC0094, RD15 — is the parliament’s own: it is how you cite the thing, and how you find it again if a link moves.
3. Individuals are not named here, and they are not quoted either. Most written evidence comes from people rather than organisations, and Parliament publishes their names. Those rows read “Individual” and the link goes to the parliament’s own page, where the name sits. That is partly a licence point — the Open Parliament Licence covers what Parliament publishes but expressly not “personal data in the information” — and partly a plainer one. People write to committees about their own lives: their health, their money, their families. The name is the least of it. So a private submission here is a link and nothing else, and the words stay where their author put them.
4. There is a link, not the text. A row holds who, to whom, about what and when. Where the submitter is an organisation — a trade body, a company, a charity — the row also carries the opening of what it said. Nothing else is stored, for two reasons. The archive runs to about 130,000 submissions and the text would have to be fetched one document at a time. And a published submission can be amended afterwards: Erskine May allows it where publication “interferes with a witness’s reasonable expectations regarding privacy”. A copy would outlive the redaction. A link would not.
5. The back-fill runs from 1 January 2021, and it arrives in that order. The UK Parliament has an API and is collected first; the Senedd and the Scottish Parliament are read from their published pages and follow. The Northern Ireland Assembly is not here at all. Oral evidence — the transcripts of committee sessions — is not here either, and neither are committee reports, which have their own source. The line above the results always says which months are loaded, and a search only looks at those.
6. A joint inquiry files under both committees at once. Where two committees run an inquiry together the committee cell holds both names joined, exactly as the committee reports source does — so filtering on one of them alone will not find the joint ones. Search the inquiry instead. And the match box works from the same hand-written sector list as the consultation finder, so a description that works there works here; press Clear to see the month in full.

Six things to know before relying on this

1. These are the letters somebody published, which is not the same as the letters that were sent. Departments answer hundreds of thousands of letters a year — the Cabinet Office publishes the percentages answered on time and nothing else. What reaches this register is the small, deliberate slice that a committee, a library or a department chose to put on the record. It is the best public account there is of who is pressing whom in writing, and it is a sample, not a census. Nobody can tell you how large a sample.
2. Who wrote to whom is read out of a sentence, not out of a field. None of the three sources publishes a sender and a recipient as separate data. Both names are pulled out of the description, and where the sentence does not make them plain the row keeps its own description as its heading, shows no direction, and offers no filter. That is deliberate: an empty answer is recoverable and a confident wrong one is not. If you are counting rows rather than reading them, count the ones with a direction and say so.
3. Some letters can be read here in full, and most cannot. Where a minister, a senior official, a government department or a parliamentary committee wrote the letter, its words are Crown or parliamentary copyright and the Open Government Licence and the Open Parliament Licence allow it to be published here. Regulators are checked one by one, because they differ: those that are part of the Crown or publish under the Open Government Licence (Ofqual, Ofgem, Ofwat, Ofsted, the Office of Rail and Road, the CMA, the Information Commissioner and others) are shown in full like a department; Ofcom is shown in full under its own terms, with its copyright credited at the top of the letter; and the FCA, the Payment Systems Regulator, the Bank of England and the PRA, the Civil Aviation Authority and the National Audit Office, whose terms do not allow it, are links only. A letter written by a company, a charity, any other regulator or a private person belongs to whoever wrote it — a committee publishing it does not license it to anybody else — so those rows carry a link to the original and nothing more. The same rule takes out any deposit whose own title says it attaches or encloses somebody else’s letter, because that is two letters in one document and there is no way to publish half of it.
4. A letter published as a scan has no text to show. Where the text is here it was read out of the PDF the publisher uploaded. Some letters are photographs of paper, and those open to a note saying so and a link to the original rather than to a page of nonsense.
5. The date is the date the letter was written, where anybody said so. Deposited papers put it in the title and that is what the row uses. Committee correspondence and GOV.UK give a publication date, which can be weeks after the letter — so a reply that looks slow may not have been, and a gap between a letter and its answer is a ceiling on the real delay rather than the delay itself. Where a letter was written on one day and made public on another, the register keeps both: it shows you the day it was written, and the weekly briefing uses the day it became public, because that is the day it was news.
6. Three sources, three different slices, and one big absence. Committee correspondence is chairs and ministers; deposited papers are ministers answering parliamentarians; GOV.UK is departments writing outwards, to councils and regulators. What no source covers is the open letter — the trade body, the charity coalition, the hundred signatories in a newspaper. There is no public index of those anywhere, so they are not here, and their absence is the single biggest thing this register does not see.

Six things to know before relying on this

1. This is not an Open Government Licence source, and it is the only one here that is not. UKRI’s terms of use put their site under a Creative Commons Attribution-NonCommercial-ShareAlike licence — not the OGL that covers GOV.UK, and not the Open Parliament Licence that covers Hansard. NonCommercial and ShareAlike are both awkward for a paid product. So this register takes facts and not prose, keeps the window to live calls rather than mirroring the archive, links out on every row, and attributes UKRI in the footer of the page. That is the most careful reading available; it is not the same as permission, and we have asked UKRI for theirs.
2. Open and opening soon. Nothing that has closed. The finder holds a little over 1,600 opportunities once closed ones are counted; 142 of them were open or upcoming on 15 September 2026, and those are what is here. A closed call is not something you can apply to, and an archive of them is a much larger extraction than a live window. If you want the history of a scheme, the finder itself has it.
3. About a third of the calls have no closing date at all. Fifty of those 142 said either “Open – no closing date” or “to be confirmed”. Some of those are genuinely open-ended — Innovate UK’s innovation loans, for instance — and some are pre-announcements where UKRI has not settled the date. They are all here, undated, sorting after everything with a deadline, and the page says so above the table. Dropping them would have made the register look tidier and told you less.
4. Soonest first, which is the opposite of every other register here. An archive answers “what happened”, so the newest record goes first. A list of calls answers “what can I still do”, so the nearest deadline goes first, and the panel at the top of the page shows what closes in the next thirty days. Both narrow with the filters and the search box.
5. The funder shown is the lead funder. UKRI writes a funder out in full — “Engineering and Physical Sciences Research Council (EPSRC)” — and a good many calls name three or four councils plus a government department. The filter offers the acronym of the first; the rest, and any co-funding department, are in the detail line under each result. So filtering on MRC will not find every call MRC has a hand in, and the search box will.
6. There is no interface to this, so it is read off the page. UKRI publishes no data feed for opportunities: their site’s own developer interface covers councils, news, publications and events, and stops there. There is a subscription feed of new calls, but it only ever carries the twenty most recent and it orders them by the day they first appeared — so a deadline that moves never shows up in it, which is exactly the change worth knowing about. This register therefore reads the listing pages themselves, once a week. If UKRI redesigns those pages the collector finds nothing and stops, rather than quietly writing an empty register over a good one; the failure is loud on purpose.

Three things to know about the dashboards

1. The four audit sources barely join up, and the joins are deliberately thin. Impact assessments and post-implementation reviews both carry an IA number, and the two sets do not overlap at all. Regulatory Policy Committee opinions carry no IA number. National Audit Office reports carry no key of any kind. So the policy view links records on their wording, behind four guards that between them produce a couple of dozen links across several hundred records. Under-linking was chosen on purpose: a wrong join here would read as a finding.
2. Describing your organisation narrows the whole page, not just the table. The tiles, the chain and the cycle panels all move with your description. A figure that cannot honestly be narrowed to one organisation — a total for the whole record — is hidden rather than left showing a number that no longer means what it says.
3. One date format, because our data platform write dates five ways. Everything is shown as D MMM YYYY, read day-first, which is also what the “newest” sort uses. One bad date is left as it arrives rather than quietly corrected: a Defra consultation on land use in England is published with the year 0025, and the page prints it that way so the error stays visible at the source.