For government staff
This page is for people who work in an Australian government agency and want to use a dataset from this site in their work. It says what the site is, what it is not, and how to check anything on it against the source.
It is not an official source
publicdata.au is a private website run by National Digital, a software company in Burleigh Heads, Queensland. No government agency runs it, funds it or has endorsed it. The official source for every dataset is the publisher's own portal, and each dataset page links to it. When a figure matters, cite the publisher and use this site for the file.
What you can rely on
Every version keeps the publisher's own file beside it, byte for byte, with its SHA-256 hash. The manifest beside it records the portal URL, the date the file was fetched and the licence the portal stated at that moment. You can download the source file from here or from the portal and compare the hashes.
The rows are the publisher's rows. Cells are typed, headers are renamed to a plain form with the original header kept beside each one, and the encoding is made UTF-8. A cell the publisher suppressed, such as "<5", becomes null with a flag that says so. Nothing is added, removed, ranked, joined or summarised. The count charts on a dataset page are counts of rows worked out in the build, and the caption says so.
A URL you can put in a brief
A dated version never changes. A link of the form /d/<dataset>/v/<date>/ returns the same bytes for as long as the site exists, so a reader of your brief can open the file you used. The latest/ link moves to the newest release, so use a dated link when the figure must stay the same.
Each dataset page has a citation ready to copy in author-date form, the form most departmental style guides ask for, as well as plain text, HTML, Markdown and BibTeX. The licence requires the publisher's attribution, which is inside every file. We ask that you also name publicdata.au and link to the version, and you are free to decline.
Before you use a dataset
Each dataset page has a box under the description that states the publisher's own caveats, such as which months are preliminary and which years are not comparable. Read it before you quote a figure. The publisher's release notes on the portal remain the authority.
On a departmental network
The site is plain HTTPS. Nothing needs a login, a key, a plug-in or JavaScript to download a file. The default download on every dataset page is Excel when the workbook is a size an office machine opens, and CSV otherwise. Excel and Power BI can read a CSV or a workbook straight from its URL through Data, then Get Data, then From Web, and the dataset page shows the exact steps for the format you pick.
The site sets no cookies and runs no advertising or third-party tracking. The only analytics is a first-party Cloudflare page count, and there are no accounts. The privacy page lists everything the site records. Cloudflare hosts the site and serves it from its network in Australia and elsewhere. The disclosure policy for security reports is at /.well-known/security.txt.
Who runs it
National Digital, ABN 13 744 838 758, Unit 11, 76 Township Drive, Burleigh Heads QLD 4220. The terms of use set out the liability position, including the Australian Consumer Law. Questions and corrections go through the contact page.
If a value looks wrong
It is almost always wrong in the publisher's file as well, because the site publishes what the publisher publishes. Send it to the publisher, whose page is linked from the dataset. If the serialisation is wrong, for example a column typed badly, tell us and we will fix it, publish a new build and record the correction in the version's notes.
Asking for a dataset
Every dataset listed on an Australian government open-data portal can be searched from the box at the top of any page, and one with an open licence and a download can be voted for. Votes set the build order. Nobody is asked who they are. If your agency would like a dataset built sooner, or wants the same treatment for data it has not published yet, the publishers page says how that works.
If your agency publishes
A dataset listed on your portal with an open licence can be served here without a request, and you can ask for it to be removed at any time. Your attribution travels inside every file, your original file sits beside every version, and every page states that you have not endorsed the site. The publishers page has the detail.