About · Data
Download the data
Everything on this site comes from one database, and you can download all of it. Use it for your own analysis, or give it to a tool that can read SQLite or CSV files, rather than collecting it page by page.
Data as at 28 October 2026.
- spending-sqlite.zip (110 MB): the SQLite database the site runs on, with every table. SHA-256 47ac6c97f4756642b6d831d5c4f92dd229de9615230368373b63e18fe2b5560c
- spending-csv.zip (48 MB): CSV files of the main tables, listed below. They open in Excel. SHA-256 9492f4446a50b1a7002d4fb858735f74f8a6561dda30837698f9a68cd9c7ae3c
Each file includes a README with the notes on this page. The files are rebuilt whenever the data is reloaded, and keep the same addresses.
Reading the figures
- Amounts are whole Australian dollars.
- Budget estimates and actuals are different series. Every amount carries its estimate type (Budget estimate, forward estimate, estimated actual or actual) and the document it came from.
- Contract and grant values are whole-of-life values as published, not annual spending. The annual estimates files spread each value evenly over its term, and are estimates.
- Program line items don’t add up to the whole-of-government total. The difference is the whole-of-government adjustments.
- Columns marked AI come from a language model and may be wrong. See how contracts are classified and how suppliers are grouped.
- Signals are patterns worth a closer look, not findings of wrongdoing.
Suppliers who are people
Some suppliers and grant recipients are individuals, such as sole traders. Their names are as the government published them. This site leaves them out of rankings and signal lists, and we ask that you do the same. The files flag them, and the methodology explains how they are identified.
Sources and credit
Our own work on the data is licensed under Creative Commons Attribution 4.0 (CC BY 4.0). That covers the database as a whole, the AI classifications and groupings, the signals and the annual estimates. The underlying figures keep their publishers’ licences, and Sources and changes lists every file and API response the data was loaded from. Please credit the original publishers and the Federal Spending Explorer (spending.mchugh.au).
What’s in the CSV files
| File | Rows | What it holds |
|---|---|---|
| contracts.csv | 256,164 | Every AusTender contract notice, at its latest amendment. value_whole_of_life is the contract's value over its whole term, not annual spending. ai_category is an AI classification of professional services contracts and may be wrong. |
| contract_amendments.csv | 340,439 | Every published version of each contract notice, with the value it gave. |
| contract_annual_estimates.csv | 465,414 | Each contract's value spread evenly by day from start to end date, by financial year. An estimate, not reported spending. |
| suppliers.csv | 35,901 | Suppliers named in contract notices, with their ABR entity type and corporate group. individual_basis is set when the supplier is, or may be, a person. |
| program_expenses.csv | 12,875 | Portfolio Budget Statements program expense line items, by financial year. treatment says whether a line counts as an expense or was set aside to avoid double counting; treatment_rule names the rule, explained on the methodology page. |
| whole_of_government.csv | 165 | Total expenses by function and by type, from Budget Paper No. 1 and the Final Budget Outcome. |
| subfunctions.csv | 404 | Expenses by sub-function (Budget Paper No. 1, Appendix A), and the general revenue assistance itemised in Budget Paper No. 3. |
| loans_and_investments.csv | 249 | Net cash flows from loans and equity provided for policy purposes, which are not expenses. Negative is net cash paid out. |
| grants.csv | 219,127 | GrantConnect grant awards, at their latest version. value_whole_of_term is the award's total over its term. recipient_key is the recipient's name normalised (case, punctuation, company suffix), which groups the site's recipient pages; GrantConnect publishes no ABN. |
| grant_annual_estimates.csv | 397,225 | Each grant's value spread evenly over its term, by financial year. An estimate, not reported spending. |
| signals.csv | 15,191 | Contracts matching a "worth a closer look" rule, explained on the methodology page. A pattern worth checking, not a finding of wrongdoing. Suppliers who are, or may be, individuals are left out, as they are on the site's signal lists. |
| sources.csv | 50,878 | Every source file and API response the data was loaded from, with its SHA-256 fingerprint. |