Original Data Studies Built From Public Records

The studies that get cited in these three sectors are built from records that are already public and simply have not been assembled. Every Fair Work Commission decision is published. AustLII carries judgment text across most Australian jurisdictions. data.gov.au holds departmental datasets nobody has cross-tabulated, and a Commonwealth agency has thirty days to answer an FOI request under the Freedom of Information Act 1982, extendable by another thirty with your agreement. Plan on ten to fourteen weeks from brief to release. Very little of that is writing. The extraction: One study per engagement. We agree the question first and the sources second, and the question has to be one where the answer is not already knowable: how long a general protections matter actually runs from application to conclusion, what share of graduate program postings now names a salary band, how many providers changed registration status across a defined period. Extraction is a scripted pull where the source allows it and a manual read where it does not. Either way the code or the query goes into the appendix with the extraction date, because a study whose numbers cannot be regenerated in two years is not evidence, it is a graphic. The rule that decides most of it: We do not report a subgroup below one hundred responses or one hundred records. That rule alone removes most of what is sold as a data study in this market: a thousand-person panel survey sliced until the headline rests on thirty-eight people in one state. When the data cannot carry the headline, we change the headline. We also do not commission panel surveys and describe the result as research, and if a survey is the only available route to your question, we will tell you the question is wrong for this format. Publication: The study sits at a permanent URL on your own domain, never on a microsite, with a method section naming every source, the extraction date, the exclusions and the licence. The dataset is published as CSV under CC BY 4.0 with an attribution string written for you. Charts are SVG with the underlying figures printed in the caption, because a journalist redrawing your chart in house style will cite the source they can read from the caption. There is one page, and it is the page every link points at. Release: One masthead on the beat gets it first under a dated embargo, with the dataset and a named person available for questions. General release follows to a list built for this study, not a standing list. Where a sector body, a regulator or an academic group would find the numbers useful, they receive them with no link request attached. Those citations arrive later and they outlive the news cycle. Who this is not for: Anyone who needs the conclusion agreed before the extraction. If the data comes back against the brief, you get it that way and you may decline to publish; the fee and the dataset are yours either side of that decision. It is also the wrong engagement if your legal or compliance review cannot turn a document around inside five business days, because an embargo date does not move for internal sign-off. And if the study would sit next to a commercial position you hold in the answer, that position gets stated inside the study or we do not run it.