Where our data comes from
HypaBase reads only what a company or government chose to publish, or a feed we are licensed to use. Here is the full list of every register and directory we read, grouped by source, with a link to each so you can open it yourself. For what we block and how each rule is enforced, see our compliance page.
Every source is on this page. 80 register, directory and trade-show sites are in use today across 33 live source groups, each linked so you can open it yourself. 278 more identified sources are named on the roadmap below and wired in continuously - held-for-review groups never feed the product and are never counted as in use.
Reading these sources leads the engine to companies’ own websites - enrichment reach, never sources: 1,616 read and verified live at last count (16 Jul 2026), 779 of them quoted as evidence inside the product. We don’t name those companies: a business shouldn’t appear on a vendor’s website just because we read its public pages, and any company can ask to be excluded at any time.
Each card below also states what its reader deliberately leaves alone. Public does not mean ours to take: fields that are personal data - contacts, named individuals - are excluded from register reading by design (contact discovery runs separately, budget-capped and PII-redacted), and carrier relationships are only ever used as evidence on a lead, never assembled into another company’s customer list. The crawler announces itself by name - HypaBaseBot - and honours robots.txt under RFC 9309, fail-closed: if a site says no, or we cannot read its answer, we do not fetch.
Not every source is equal. Several are statutory registers - lists a company is on because the law requires it, published with its ABN - the strongest identity evidence a public source can carry.
Public ARTG register (medical-device sponsors). Sponsor ≠ importer proof ≠ a carrier customer ≠ lead - every row is a seed that must pass qualification. Run the bundled sample or upload a TGA ARTG CSV export.
EESS public register (electrical responsible suppliers). Run the bundled representative sample or upload a real EESS CSV/export. Responsible supplier ≠ confirmed importer ≠ lead - every row must pass qualification.
ACMA RF / comms equipment suppliers. Run the bundled representative sample or upload a real ACMA supplier export. No carrier-customer inference, no contact harvesting - every row must pass qualification.
Approved search APIs only - never scrape Google/LinkedIn or bypass logins. A public carrier mention is freight RELEVANCE only, never carrier-customer proof. Run the bundled representative sample or paste real search/website evidence rows (company,website,evidence).
Public exhibitor lists. Run the bundled representative sample or upload/paste a real exhibitor list. Exhibitor ≠ importer proof ≠ lead - qualification required.
Commercial job boards are NOT read: their crawls are parked (held) pending permitted access - an honest, self-identifying bot is refused and we never spoof a browser - so they are never presented as active coverage. The one real public source, the Workforce Australia vacancy API, disallows automated access in robots.txt, so collection there self-suppresses. Job ad ≠ freight-spend proof; stale ads = low confidence. Run the bundled representative sample or paste real job-ad evidence rows.
Public federal contract + grant awards = DATED why-now triggers (an award funds new equipment / imports). Award ≠ freight-spend proof ≠ lead - confirm the recipient controls replacement logistics; every row passes qualification. Run the bundled representative sample or upload an AusTender/GrantConnect CSV export.
Public export-award finalist lists - strong proof of EXPORT orientation (express-air outbound). Exporter ≠ air-express proof ≠ lead. Run the bundled representative sample or upload a finalist list.
Public recall registers - a recall = urgent replacement-stock timing (a why-now trigger). Recall ≠ freight-spend ≠ lead; never overclaim as carrier spend. Run the bundled representative sample or upload a recall export.
ABN/ABR public registry - the entity BACKBONE (resolves a name → registered entity + industry/state, raising entity-match confidence so seeds clear the entity gate). This button imports the bundled representative sample; the live ABR lookup (needs an ABN Lookup GUID) runs per-lead during enrichment, not here. Entity data ≠ importer proof ≠ lead.
Official keyless AusTender open-contracting API, CC-BY 3.0 AU (licence embedded in every payload). An award is a dated trigger, never freight-spend proof - every row passes qualification.
The vendors' own public case-study libraries (robots allow, live-verified 2026-07-16). A named customer story documents shipping activity at the time of writing; stories are undated, so seeds pass qualification like any other. Deep-audit build order #13.
Attorney-General's statutory register, server-rendered listing read newest-first in a bounded rolling window (robots honoured incl. the PDF disallow; live-verified 2026-07-16). Deep-audit build order #3.
APCO's own public member directory (robots allow-all, live-verified 2026-07-17), read in a day-rotating bounded window (full register covered weekly) filtered to Brand Owner organisations. Co-regulatory NOT-FOR-PROFIT membership, so seeds carry no regulatory signal. Deep-audit build order #6.
Public WordPress REST directory (robots allow-all, live-verified 2026-07-17), read newest-first in a bounded rolling window over the 1,249 Australian projects. Award attribution is evidence a business makes or commissions physical product.
Statutory Commonwealth register; the portal publishes its own full-corpus CSV (host reachable to the honest UA, live-verified 2026-07-17). The register names BRANDS - a brand's REGISTRANT is the importer/manufacturer, so brands resolve to legal entities downstream and seeds carry no regulatory signal until the licence holder is confirmed. Deep-audit build order #4.
Statutory Treasury register; the register publishes its own full XLSX and its terms state that downloading constitutes acceptance. Imported on demand (the register updates ~fortnightly) - first run 2026-07-17: 3,494 new entities. Deep-audit build order #2.
The certifier's own public client list (robots honoured, live-verified 2026-07-17). Kosher certification is bought by producing food businesses - each name is a manufacturer with a published certification as evidence. Kosher Australia (the other certifier) stays on the roadmap: its pages time out today.
HACCP Australia's public "check the register" directory (robots allow-all, live-verified 2026-07-17). Certified suppliers make or distribute physical product into the food industry. Private certification, so seeds carry no regulatory signal.
Open-data CSV (CC-BY); the newest monthly snapshot is resolved via the data.nsw.gov.au API at run time and its date is stamped into every seed. Mirrors the working VIC liquor adapters.
Public association member list - cold-chain and urgent clinical freight verticals. Membership ≠ lead - every row passes qualification.
Public association member list - hygiene, personal-care and specialty-chemical importer-manufacturers with overseas supply chains.
Public association member directory - complementary medicines are the canonical high-value low-weight air vertical. Server-rendered list; the published PDF is not used.
Public CESA member directory (consumer-electronics suppliers and importers), read by the scheduled discovery workflow. Member ≠ importer proof ≠ lead - every row must pass qualification.
Public ElectronicsOnline company directory, read by the scheduled discovery workflow. Directory listing ≠ importer proof ≠ lead - every row must pass qualification.
Public LabOnline company directory, read by the scheduled discovery workflow. Directory listing ≠ importer proof ≠ lead - every row must pass qualification.
Public FoodProcessing.com.au company directory - the fourth Westwick-Farrow directory, same open pattern as the three already in use (live-verified 2026-07-17). Directory listing ≠ importer proof ≠ lead - every row must pass qualification.
Public ProcessOnline company directory, read by the scheduled discovery workflow. Directory listing ≠ importer proof ≠ lead - every row must pass qualification.
Public MTAA corporate-member directory (medical technology), read by the scheduled discovery workflow. Member ≠ importer proof ≠ lead - every row must pass qualification.
Public CSV of endorsed AU export meat establishments (freight-heavy exporters), read by the scheduled discovery workflow. Endorsed establishment ≠ air-freight proof ≠ lead - every row must pass qualification.
Public VGCCC liquor licence-register CSVs (pre-retail importers/wholesalers + producers) - businesses that move physical product. Licensee ≠ importer proof ≠ lead - every row must pass qualification.
ACT open-data register of licensed waste and transport operators, read by the scheduled discovery workflow. Licensed operator ≠ importer proof ≠ lead - every row must pass qualification.
Public APVMA PubCRIS dataset (registered ag/vet product holders) via data.gov.au, read by the scheduled discovery workflow. Registrant ≠ importer proof ≠ lead - every row must pass qualification.
These sources are identified but not read today: some are held until a licence or legal review clears them, others are parked because the publisher blocks or disallows automated access and we will not work around that. None of them feed the product today.
Adapter built + tested (CC-BY per-category CSVs) but PARKED honestly: data.gov.au robots.txt disallows all automated access and the agency hosts wall datacenter IPs, so the scheduled run self-suppresses (it probes robots nightly and activates the day the portal opens). Active path today: a manual CC-BY CSV export upload. Deep-audit build order #1.
Adapter built + tested (CC-BY 4.0 Facilities CSV, 4,000+ ABN-verified industrial facilities) but PARKED honestly: data.gov.au robots.txt disallows all automated access, so the scheduled run self-suppresses (probes robots nightly, activates if the portal opens). Active path today: a manual CC-BY CSV export upload. Deep-audit build order #5.
PARKED pending permitted access - the ABF edge refuses an honest, self-identifying bot (403) and we never spoof a browser UA. The workflow stays registered so the run report keeps surfacing the honest error; it collects nothing today and is never presented as active coverage.
ROBOTS-SUPPRESSED - workforceaustralia.gov.au disallows automated access in robots.txt (User-agent: * / Disallow: /), so the workflow honours that and self-suppresses: it collects nothing until permitted access is arranged, and is never presented as active coverage.
BOL proves shipment / import activity, NOT air-express or carrier usage. Confirm the licence; unknown licence is blocked. Upload a licensed/allowed BOL CSV to run.
Licensed providers only. Marketing claims are never trusted - small sample only. Unknown licence = blocked.
Real, identified Australian public sources we intend to wire in, named here so the plan is inspectable. None of them feed the product yet, and they are never counted as coverage - each one moves up this page when its adapter ships and clears review.
This list is generated from the product’s own source registry, so it stays in step with what the platform actually reads. A source is only shown as in use once it has been cleared, the roadmap names are the registry’s own entries - not a marketing list - and it is read at working volume: a single overnight run of the adapter fleet imported 7,241 records (17 Jul 2026).