How the data is built: GP Zoo

The build cycle

It is not a scrape. It is a scrape that is then verified, audited and reconciled, with a full change and backup trail.

Sourcing. Practices assembled from primary and government or health directories, then each verified.
Supervised recovery. A cloud agent recovers current GP rosters from each practice's own website and authoritative health directories. Fast, and always supervised.
Audit. Every batch is audited before it is trusted: identity re-checked, counts re-derived, a sample independently re-verified on the live web.
Propose then apply. Nothing an agent produces is written straight into the master. Changes are staged, reviewed and only then applied.
Versioning. The canonical file is dated, the prior version preserved as a wind-back point, and every change backed up first.
Reconciliation. After each change the audit tabs are recomputed so the totals tie out.

What makes it defensible

Identity verification. A roster is accepted for a practice only when identity is confirmed on name plus address or phone. Suburb-only matching is not used; the same street with a different number is a different practice.

Independent versus group. Practices are separated into independent and corporate or group-owned, matched on name and address, never on a shared phone or domain. Group-owned practices are moved out of the independent list, so a practice is never counted in both.

GP counting. The GP count is the number of named GPs on the verified roster. Registrars are counted; specialists, nurses, allied health, students and admin are not. An unusually high count is a flag to re-verify, not a number to trust.

Classification overlays. Every practice is size-banded (solo, small, medium, large independent) and graded on the Modified Monash Model (MMM) from the authoritative locality classification, not a postcode rollup. A confidence flag marks split localities.

Sources

Operator websites, the Australian Business Register, Dun and Bradstreet, LinkedIn, press coverage and direct verification. Where a record cannot be verified it is left blank with a reason, never filled in. 'Still unverifiable' is a valid outcome.

What the data does not hold, and where it stops

We do not sell individual doctor contact details. AMPCo does that, with AMA verification and 97% postal deliverability, and owns that ground. We sell ownership, structure and in-practice attributes, which they do not hold. The two are complementary.

The national list is compiled with technology systems and is not warranted. Roughly 7% of practices still carry no roster, and it is marked as such.
The operator email layer is held back and is not sold in any tier.
Named individual-GP data sits behind a sign-off gate and is not offered as a product without it.
The data does not cover skin clinics, urgent care or other specialty clinics.
It is a maintained, versioned dataset on an audit cadence, not a live feed.

If your organisation's details are wrong, tell us and we will correct them. Corrections from buyers are logged into the next build, which is why the file gets better rather than staler.