Planned studies

This is the register of studies the research section intends to run. Every entry states its question, why the answer would be worth having, the methodological difficulty that stands in front of it, and its status.

Every status below is not started. Nothing on this page has been collected, sampled, tested or measured, and no entry here has produced a figure.

The register is published for the same reason the methods are: a study that exists in public before it runs cannot be quietly dropped when the result is inconvenient, and a study that appears from nowhere with a finding attached has no history a reader can check.

Accessibility of Australian small-business websites

Question. How accessible are Australian small-business websites in practice against WCAG 2.2, and which success criteria fail most often?

Why it is worth answering. The Disability Discrimination Act 1992 has no small-business turnover exemption, so the obligation applies across the whole population in question. The compliance rate is asserted frequently in this industry and measured rarely, and the per-criterion breakdown is directly actionable in a way a headline rate is not.

Difficulty. Automated tooling covers only part of the criteria, commonly estimated at between about a third and a half, so the automated component measures a floor rather than conformance. The manual component is what makes the study meaningful and also what limits its size.

Status: not started. The full method is published at the accessibility audit method.

What Australian web design quotes actually contain

Question. When an Australian small business receives a quote for a website, what is actually in the document — what scope is defined, what is excluded, who owns what at the end, and what is left to be decided later?

Why it is worth answering. Most of the difficulty in buying a website is not the price, it is that two quotes describing different things cannot be compared. A structured account of what these documents contain and omit would let a business owner see what a normal quote leaves out, which is more useful than a benchmark figure.

Difficulty. The sample would come from quotes businesses volunteer, which is self-selection of a severe kind. People who send you their quote are disproportionately people who are unhappy with it. Every document would need de-identification of both the business and the supplier before anything could be published, and no supplier would be named in any circumstance. There is also a conflict here sharper than anywhere else in this register: the author competes with every supplier whose quote would be submitted, and a study by one competitor characterising other competitors’ commercial documents carries an obvious hazard. If this study runs, that constraint shapes the design or it does not run.

Status: not started. No quotes have been collected and no collection mechanism exists.

What happens to Australian small-business websites after launch

Question. Of Australian small-business websites launched in a given period, how many are still being maintained two years later, and three years later?

Why it is worth answering. Websites are bought as projects and behave as assets with running costs. Whether a site is still updated after the launch enthusiasm ends is the difference between an ongoing expense and an abandoned one, and the answer bears directly on how a business should budget for the thing before buying it.

Difficulty. Defining “maintained” is most of the work. A content change, a platform version bump, a certificate renewal and a plugin update are all detectable in different ways and all mean different things. Detection through generator strings, response headers or archived snapshots is inference from a signature rather than observation of the site being worked on, and archive coverage of small Australian business sites is uneven. Establishing a launch date retrospectively is its own problem. This one may end up as a study of what can be detected rather than of what happened, and if so it would be published under that narrower description.

Status: not started.

Whether AI assistants cite Australian sources on Australian web questions

Question. When an AI assistant answers a question about web design, hosting, domains or website obligations in an Australian context, does it cite Australian sources, and how often is the answer grounded in material from another jurisdiction?

Why it is worth answering. Australian rules on domain eligibility, consumer guarantees, email marketing and accessibility differ from the United States material that dominates the web on these subjects. An assistant answering an Australian question from American sources would give confidently wrong answers on things a business can be penalised for.

Difficulty. Assistant outputs are non-deterministic, vary by account and region, and change without notice between model versions, so a citation set is a snapshot that cannot be reproduced by a reader later. Choosing the prompts is itself a design decision that determines much of the result, and a prompt set assembled by someone who publishes Australian web content would be prone to selecting for questions this site happens to answer. Any run would need its prompt set, dates, platforms and model versions published in full, and would be a measurement of a moment rather than of a system.

Status: not started.

What listing a study here does and does not mean

Listing a study is not a commitment to run it. These are questions considered worth answering, subject to whether a defensible method exists, whether the data can actually be obtained, and whether the work is affordable in time.

A study that is defined and then found to be undeliverable will be marked abandoned on this page, with the reason, and will stay on the page. It will not be deleted. A register that only ever shows live and completed items is a filtered view, and the filtering is the interesting part.

The standing rules for how any of these would be measured, including how samples are drawn and how tool-derived figures are treated, are on the research method. How the site’s existing published figures were gathered is a separate question, covered on how this site measures things.