The limits of this platform

What we don’t know.

Every other page here tells you something. This one tells you where those pages stop, why they stop there, and what it would take to move the line.

The numbers below are read out of the system, not written down. If a gap closes, the figure moves on its own.

Who decides

Have the communities in your registry agreed to be in it?

No community has yet recorded a decision, and no community record has been published. Everything visible is assembled from public sources, not released by the community it describes.

0 authority records · 0 of 42 community records published

Why: The publish gate requires a verified authority record from the community and has no administrator override, so we cannot publish on anyone’s behalf. That mechanism is built and has never been used.
What would change it: Communities recording their own decisions. This is a relationship, not a feature.

Has anyone checked these impact figures with the community they describe?

Every figure carries a source we can show you. None has been signed off by the community it is about.

0 of 133 figures signed by their community

Why: Sourced is not the same as agreed. We chose the vocabulary, the categories and which figures count as outcomes; the community whose work it describes has not been asked.
What would change it: Community sign-off, which the schema already has a field for and no rows in.

Who decided which organisations are Aboriginal-controlled?

Mostly a register, and for a residue, us. Most flags now rest on ORIC registration, Supply Nation certification, or the organisation’s own statement. A small number still rest on our reading of the organisation’s name, and some carry no recorded basis at all.

21 of 2927 flags rest on our own inference · 198 carry no recorded basis

Why: An earlier process inferred Aboriginal identity from organisation names. That is us classifying somebody else’s identity by our own method, which our Indigenous Data Sovereignty assessment named as a failure. Where a register agrees, we now defer to it; where a register is silent we keep the flag rather than strip it, because silence is not disagreement.
What would change it: A human reading the residue. It contains both organisations a register wrongly overlooks and at least one government department that should never have been flagged.

Did these organisations agree to be listed?

No. The organisation records come from public registers. That makes holding them lawful; it does not make it consented.

0 of 104,426 organisation records claimed by the organisation

Why: There is no consent step for the register layer, only for the claim process, and almost nobody has claimed.
What would change it: A consent position for the organisation layer, which is a governance question rather than an engineering one.

What is in the data

Are the pins on your map in the right place?

Nothing on the map has been verified against a source. Some pins are a town centre standing in for an address, and they are marked rather than drawn as precise.

4 verified · 338 marked as a locality stand-in

Why: Earning the highest precision grade means checking a coordinate against an authoritative record, one at a time. No row has been through that yet.
What would change it: The review pass, then a bulk load from the ABS locality boundaries.

Can you tell me what is happening in my community?

Only roughly, and we will usually undercount. Organisations are located by their registered address, and a service delivered in a remote community is frequently administered from a regional centre hundreds of kilometres away.

Why: Verified on 2026-07-31 against one community: the largest organisation delivering services there is registered in a city, so every place-keyed view of that community missed it entirely, while the shared postcode pulled in unrelated organisations from the mainland. An address says where the paperwork sits, not who a service is for.
What would change it: A record of who a service is FOR, rather than where its office is. That is a question only the community and the organisation can answer, not one we can derive.

Do you know where every program is?

No. Most programs correctly have no location at all, because they name a state or the nation rather than a place. Of the rest, we can resolve a minority against the ABS locality boundaries.

18 locality names resolvable with a stated source

Why: Our gazetteer holds only entries whose coordinate has a stated source. Typing town coordinates from memory is how this dataset acquired its wrong ones in the first place.
What would change it: A human review pass over the resolved candidates, then a bulk load from ABS Suburbs and Localities.

What about places whose names appear more than once?

We refuse them rather than pick the likelier one. Denmark is a town in Western Australia and a country; nothing in the record says which is meant.

5 names held as ambiguous by name alone

Why: Choosing between same-named places without evidence is guessing, and a confident wrong location is worse than a missing one.
What would change it: A state or region in the source record, or a person who knows.

Can you tell machine-derived records from human-checked ones everywhere?

Not everywhere. Two of our largest tables carry a provenance flag. The evidence library, the media table, the outcomes table and the funding table do not have the column at all.

Why: The house rule that filters machine-generated rows cannot be applied where the field does not exist.
What would change it: Adding the column and backfilling it honestly, where unknown means unknown rather than verified.

What the method can carry

Can you compare these numbers to each other?

Often not. Roughly half of the recorded figures cannot be turned into a number at all, because the source hedged: "more than 200", "nearly three quarters", "up to 90 a night".

55 of 133 figures carry a usable number

Why: Turning a hedged phrase into a number invents precision the source declined to give. Those figures keep their words and no number, which means they can be read but not charted.
What would change it: Nothing we can do. The source said what it said.

Do your justice reinvestment metrics line up with the national framework?

We do not know. The Measurement and Evaluation Framework for the National Justice Reinvestment Programs was finalised in January 2026 and is not published.

0 of 69 metric keys mapped to the national framework

Why: A mapping written without the document would be a guess dressed as an alignment. Our mapping table is deliberately empty and a guard test keeps it that way.
What would change it: Reading the framework. It can be requested from justicereinvestment@ag.gov.au.

What the sources will tell us

Have you read the documents in the public record layer?

No. Relevance came from titles and subject lines. Every record carries a flag saying the text has not been read.

Why: Thousands of tabled papers and questions on notice, each a PDF. Titles are cheap; reading is not.
What would change it: Opening them. Never quote a figure from that layer without doing so first.

Do your watch house numbers match the official ones?

No, and ours are lower. The feed we read reports who is in the cells at a point in time; the tabled official figure is an averaged daily count of everyone in custody at any point that day.

Why: They are different measures of different things. Over a fortnight where both exist, ours ran 20 to 42 per cent below the official figure on every watch house.
What would change it: Nothing. Both are correct for what they measure. Our figure understates crowding and must never be compared to capacity.

Why is there no data on medical incidents or staffing in watch houses?

Because it was asked for and refused. A question was put to Queensland Police covering medical incidents and staffing; those sections were declined.

Why: A question’s subject line says what was asked, not what was answered. The data does not exist publicly.
What would change it: The agency answering, or a different route to the same facts.

Why publish this

A platform that never says what it doesn’t know is asking to be taken on faith. We are asking communities to trust us with their records, so the limits belong in public alongside the findings.

What we got wrong How we check things