Whether it worked
How government checks its own work
Before a rule is made, someone is supposed to check it will work. Afterwards, someone is supposed to check whether it did. Those checks are done by different bodies and published on different websites. This is all of them, in one place.
No need to use clunky search filters on gov.uk that don't fit what you're looking for. Tell us what your organisation does and we'll bring together the documents most relevant to you.
— of the — records in these four sources are on your subject — — of the whole chain.
Policy by policy
Every policy assessment mapped
One row per policy instead of per body, so you can follow a single rule along the chain — where the chain allows it, which is almost nowhere. Of — policies here, — appear at more than one stage.
These are suggestions from matching words, not from reading the documents. Judge them yourself.
— nothing published at that stage ■□□□ which of the four stages this policy reached Rows are one policy, not one document.
| 00Body | 00Chain | 01Scorecardsimpact assessments | 02RPC opinionssince Sept 2024 | 03Reviewspost-implementation | 04NAO reportssince — |
|---|
No body matches that filter.
| 00Policy | 00Chain | 01Forecastimpact assessment | 02VerdictRPC opinion | 03Reviewpost-implementation | 04AuditNAO report |
|---|
No policy matches that filter.
appears at this stage does not appear — nothing in that source for this body
What the empty cells are telling you.
Only — bodies appear at all four stages. — appear at exactly one. Some of that is legitimate: HMRC makes tax rules, which sit outside the Regulatory Policy Committee's remit, and departments that wrote few regulations five years ago have few reviews due now.
But the pattern that survives those explanations is this. The body doing the forecasting and the body doing the auditing are almost never looking at the same thing, and no document produced at one stage refers to any document produced at another.
How the policies were joined. Scorecards and reviews both carry an IA number and not one number appears in both — they are 2026 document numbers in the same series, one numbering a new rule’s assessment and the other the review document for a rule made years earlier. RPC opinions carry no IA number, and exactly one of them shares a title with a scorecard word for word. Thirteen scorecards quote an RPC reference in their notes, but no RPC opinion carries its own, so there is nothing to match it against. NAO reports carry no key at all. So the joins here are made on wording, with four guards: the titles must overlap, share at least two words, share at least one distinctive word (one appearing in four titles or fewer across all four sources), and agree on any year they name. Without the last two, “Decent Homes Standard” links to “Future Homes Standard”. The rules are in govaudit-page.js; under-linking is the deliberate error.
Where each link goes. Every cell in the table is a link to the document behind it. Impact assessments and reviews open their PDF on legislation.gov.uk, NAO reports open on nao.org.uk. RPC opinions are the exception: that source stores no address against a row, so the link is recovered by matching the opinion’s title to the GOV.UK addresses the weekly job has already collected. Where the match is not confident enough the cell opens this site’s RPC source instead, with a dotted underline to mark the difference.
What is hidden while a description is applied. Everything on this page is counted over records on your subject — except the scorecard coverage figures at stage one. Those count impact assessments that carry no scorecard at all, and a count is not a set of records, so there is nothing to test against a subject. Clear the description to get them back.
How the bodies were matched. The four sources spell departments differently — the NAO writes “Ministry of Housing, Communities and Local Government”, the scorecards write “Ministry of Housing, Communities & Local Govt”. They are matched by hand into a single list of bodies, held in govaudit-page.js with every alias listed beside it. Defunct departments and the NAO's collective headings (cross-government, devolved administrations, regulators) keep their own rows rather than being folded into a successor, because folding them would invent a history our data platform do not contain. When a weekly updater introduces a spelling the list has not seen, it is reported below rather than silently dropped.
Impact assessments —, read — RPC opinions —, read — Reviews —, read — NAO reports —, read —
Analysis
Where it leaks.
Each stage is a real requirement with a real source behind it. Read left to right, the volume falls away — and so does the appetite for a negative answer.
-
Stage 01 · before the rule
The forecast
The department's own impact assessment: what this will cost, who it lands on, what it buys. It is meant to carry a regulatory scorecard — a standard grid rating the effect on business, households, the wider economy and the environment.
—impact assessments listed for the year
—impact assessments on your subject
- — are forward-looking assessments that should carry a scorecard
- — of those carry none at all
- of the — that do, — say the measure makes the country better off and — say it makes no difference
- — say it makes the country worse off
— of the assessments that should be marked against the standard grid are not.
— of the assessments carrying a scorecard are on your subject.
-
Stage 02 · before the rule
The verdict on the forecast
The Regulatory Policy Committee is the independent body that marks those assessments. It gives each a single overall verdict — green means fit for purpose, red means it is not — plus ratings on the reasoning underneath.
—opinions published, —
- — on forward-looking assessments, — on post-implementation reviews
- — green, — red
- the RPC only sees what is sent to it — the — assessments with no scorecard are not in this count
- of the — bodies it has marked, — have never had a red
— of opinions are red. The rest of the record is a pass.
-
Stage 03 · five years later
The department's own review
Most regulations carry a duty to be reviewed after five years, and the department marks its own work. The review sheet asks a direct question: keep, amend, replace or repeal?
—reviews published this year
- — recommend keeping the rule exactly as it is
- — recommend amending it; — replacing it
- none recommends repeal
- on whether the original objectives were met: — yes, — partly, not at all, or unassessable
— of — bodies in the table above filed no review at all this year.
-
Stage 04 · independent
The independent audit
The National Audit Office is the only body in the chain not marking its own work. Its value-for-money studies ask whether the money achieved anything — and unlike the other three stages, nobody has to invite it.
—NAO reports since —
- — are value-for-money studies with findings and recommendations
- — are annual opinions on a set of accounts
- attention is uneven: — alone appears on —
- bodies elsewhere in this table with no NAO report at all: —
Nothing connects an NAO finding back to the forecast that preceded it.
The honest caveat. The four sources do not cover the same window. Impact assessments and reviews are this year's vintage; RPC opinions run from —; NAO reports from —. So the columns in that table are counts within each source's own period, not a like-for-like race. They are still the only place these four sit side by side.