Loading the tracker…
How it works
A council can declare a climate emergency and change nothing that follows. A government can publish a strategy and fund none of it. This tool reads the public record — the actual meeting papers, votes and policy documents — and makes the gap between what was said and what was done visible, at scale, for anyone to check.
Much of what we track concerns harms carried by parties who cannot vote, organise or take an institution to court — future generations, and the animals inside the food system. Their interests reach a council chamber only through the people who advocate for them. Those advocates deserve to argue from evidence, not impression.
That is the whole purpose here. A harm does not stop existing because an institution left it out of the minutes. A tracker like this exists to stop the ledger's silence being mistaken for silence in the world.
The pipeline
Nothing here is opinion typed into a spreadsheet. Every step starts from a document a public body actually published, and every judgement can be traced back to the exact page it came from.
Gather the PDFs — agendas, minutes, votes, statutory guidance — from hundreds of council and government sites.
Turn the paper into text and find the places where a body actually decided or committed to something.
Line each decision up against a fixed catalogue of specific tracked actions: did they commit to this?
Verify the evidence genuinely supports the claim — a stronger, second reading throws out the weak matches.
Judge how binding the commitment is — a law weighs far more than a press release.
Combine the signals across the tiers of government into the readings below — and publish them.
A person can always challenge a result and correct it, and when a published number is later found wrong it goes into a public corrections record. The method is meant to be argued with — that is a feature, not an embarrassment.
How strong is a promise?
"We will end this by 2030" means one thing in a law and another in a speech. So a commitment is placed on a ladder, from the firmly binding at the top to the merely spoken at the bottom. Where a document sits decides how much it counts.
An old promise also fades if it is never repeated. A pledge made five years ago and never restated counts for less than one made this year — unless the body acts on it again, which resets the clock.
Four places power sits
The same issue travels through four levels, and each has a different kind of doing. Reading them together is what shows where a demand gets stuck.
The public
Pressure
Petitions and signatures — the public asking. This is how an issue gets onto the agenda in the first place.
Parliament
Voice and votes
Debates, motions and division votes. A passed vote is a real decision; a motion merely tabled is not.
Executive
Delivery
Government departments — the tier that can put a commitment into force, with budgets and binding guidance.
Councils
Local decisions
Hundreds of local authorities, whose minutes show what they resolved to do on the ground.
The heart of it
When an institution says a group's interests count but nothing follows, the failure takes one of three shapes. Naming which one is what makes a rating fair — and explainable to the body being rated.
Standing
Never mentioned
The issue is simply absent — not debated, not decided, not named. A silence, not a stance.
Weight
Mentioned, weakly
It is acknowledged — but only in a press release or an aspiration, never anything binding.
Mechanism
Promised, unenforced
A firm commitment exists — but with no budget, no delivery and nothing that makes it happen.
From these three gaps come the three readings a visitor actually sees:
Greenwash risk — How far the claiming outruns the doing. Loud claims with nothing binding behind them is the classic signature.
Public pressure — How hard the public is pushing on an issue, measured from petitions and signatures.
Trajectory — The direction of travel over time — getting stronger, or walking it back. A snapshot can flatter; direction reveals intent.
These are kept as separate readings on purpose. They are different kinds of failure and do not share a unit, so they are never crushed into one score or a league-table rank. A stalled council and a blocking council can look similar in a single number and behave completely differently — and a reader needs that difference.
Two traps we watch for
The honest version of this work has to resist two moves that make an institution look better or worse than it is. Both are being tightened in the design this tool is built on.
Displacement
Right effort, wrong problem
A body answers for one harm with genuine, audited work on a different one — a council meeting a food-system emissions target with fleet upgrades and building retrofits. Every measure is real; none touches the issue at hand. Progress must count only when the action addresses the harm actually named.
The liability shadow
Doing it quietly
Sometimes a body acts but withholds the announcement, because admitting a past harm can cost more than fixing it. So no announcement found must never be reported as nothing was done — silence has to mean the doing was checked and came up empty, not that no press release turned up.
What we are building now
The immediate piece of work is narrow and important. Until now, a rise in tabled motions or in public pressure could nudge an institution toward improving — even though a tabled motion is a plan, not a decision, and public pressure is the public acting, not the institution.
So we are separating the threads. Trajectory will count only real, binding movement by the institution on the issue itself. Public pressure keeps its own reading — it is a vital signal, but it belongs to the people pushing, not to the body being pushed. And the parameters behind every judgement — every threshold and cut-off — are being written into the published method, so anyone can see the trade-offs and argue with them.
This is epistemic infrastructure for people who advocate on behalf of those who cannot advocate for themselves. Its method is published, its classifications are correctable by a person, and its own claims are auditable against what the system actually computes. Some of the judgements described here are still being sharpened — this page includes what is being improved as well as what already runs.