The four data sources behind every Demiton public preview

The Demiton public preview pages — Cross River Rail, Bruce Highway Upgrade, and the project pipeline being added next — render real data, not narrative. Every figure is sourced. Every contract is attributable.

The sourcing comes from four feeds. There are no others.

This post documents what each one carries, what it doesn't, and how we structure them.

---

1. CKAN — state-level open data portals

Most Australian infrastructure spending sits at the state level. The state-level disclosure portals all run on CKAN, an open-source data catalogue platform. The endpoints are stable JSON APIs and don't require an account.

State	Portal
Queensland	data.qld.gov.au
New South Wales	data.nsw.gov.au
Victoria	data.vic.gov.au
Western Australia	data.wa.gov.au
South Australia	data.sa.gov.au

For Cross River Rail, the relevant package is the CRRDA Contract Disclosure Report. It carries supplier name, contract value, award date, contract reference, category, and procurement method, plus commence and finish dates. We pull the live feed at 03:00 UTC daily.

What CKAN gives you:

• The contracts that have actually been awarded
• The agency-level totals
• The contract reference numbers (so you can trace them back)

What CKAN does not give you:

• Subcontractor lineage (the head contractor's contracts only)
• Planning approvals (those live in separate state planning portals, mostly HTML)
• Tender pipeline (only awards are disclosed, not the work going to market)

Data quality varies by portal and category. For CRR, ABNs are populated on roughly 6% of records — the rest report supplier name only. We surface that figure on the page rather than pretending the gap doesn't exist.

---

2. AusTender — Commonwealth procurement

For federal contracts above $10,000, the disclosure feed is AusTender (tenders.gov.au).

Same shape as CKAN at the field level: supplier, value, dates, agency, procurement method. Different feed.

The audience for this distinction matters. State-funded infrastructure (most rail, road, water) is on CKAN. Defence work, federal infrastructure programs, and inter-jurisdictional rollouts are on AusTender. A civil firm bidding into both will see different patterns in each.

We covered the five things AusTender reveals in a separate post.

---

3. ABR — entity validation

The Australian Business Register is the source of truth for legal entity records. Free public web service, GUID-authenticated, JSONP responses.

We use ABR for one job: confirming that the supplier name disclosed in a CKAN or AusTender record resolves to a real, active ABN.

The ABR record carries:

• Legal entity name (often different from the trading name)
• Entity type (company, trust, partnership)
• Active / cancelled status
• GST registration

What it adds to the public preview is small but load-bearing — it's how we mark a contract as ABN verified on the page. Without it, the contract value is just a string we read from a feed. With it, the supplier is a known legal entity in good standing.

ABR doesn't tell you anything operational. It doesn't tell you who works there, what they've delivered, or whether they're a panel front for someone else.

---

4. BOM — site weather

The Bureau of Meteorology publishes free, station-level observation feeds. We use the geohash-addressed half-hourly observations for the project's nearest weather station.

For Cross River Rail, that's Eagle Farm. For corridor projects, it's the station nearest the works precinct.

What it adds:

• Temperature, humidity, wind, rainfall — the conditions on site
• A historical baseline of what weather is normal for that location at that time of year
• The number of viable working days per month, derivable from the observations

For the public preview, this is largely ambient context. For the paid tiers, it's how Demiton infers when wet-weather delays will land and what they'll cost in standby plant.

---

What this gives you, structured

Four feeds is not many. The work is not in acquiring the data — anyone with a HTTP client can hit these endpoints. The work is in keeping them aligned.

A contract published on CKAN has a supplier name. ABR will tell you which ABN that name resolves to. BOM will tell you whether the contract's commence date sat inside the dry season for that station. AusTender will tell you whether the same supplier has been winning work federally too.

None of those joins exist in the source data. They have to be made.

That's what the platform does on every public preview page:

1. Pull the relevant CKAN package (or AusTender query) for the project
2. Resolve every supplier ABN against ABR where present
3. Bind every contract to the project's geographic context (BOM station)
4. Persist the structured result against a versioned snapshot

The page renders that snapshot. The structuring is the value.

---

What's not in the public preview

Subcontractor relationships, internal cost data, the unit rates that won the contract, the variations that were issued during delivery, the weather impact actually claimed — none of that is in public data. It's not a research problem; it's a structural absence.

That's what the customer-data tiers (Insights and Connected) bind into. Public data tells you what happened externally. Customer data tells you what happened on your jobs. The memory layer is what compounds across both.

For now, four feeds. Structured. Sourced. Attributable.