How we research and test what we publish

Our editorial policy sets out what we are allowed to say. This page is how we work out whether it is true.

Claims about our own software

These are the easiest to get wrong, because software changes and copy does not. So a claim about what 99 Data Rooms does is written down in a register alongside an automated test that exercises the actual behaviour. If the software stops doing the thing, that test fails and the build stops. We check the claim against the running product rather than against the specification, because those two drift apart.

Statements about UK law

We read the source. For statute that means legislation.gov.uk, for guidance it means the issuing body - gov.uk, the Information Commissioner’s Office, Companies House, the Law Commission - and the page links to it so you can check us. We do not cite other vendors’ blogs as authority for what the law says.

Where the law is genuinely unsettled, or where the answer is “it depends”, we say so rather than picking the tidier answer. Several of our guides end without a clean recommendation for that reason.

Comparisons with other products

We compare against what a competitor publishes about itself - its own pricing page, documentation and feature list - and we date the comparison, because their product will change after we write about it. Where they are better at something, the page says so. We do not publish scores or ratings we have not earned, and we do not emit review ratings in our markup, because we do not have a body of verified reviews to base them on.

How we measure whether any of this works

Search Console and Bing Webmaster Tools for how pages perform, and our own server-side record of which automated crawlers fetch which pages, which is how we know whether AI systems are reading us at all rather than guessing from rankings. We compare like windows with like: Search Console reports on a delay of about three days, so a current week is a short week, and treating it as a full one manufactures a decline that did not happen.

What we cannot verify, and say so

We have no reliable way to measure how often an AI assistant paraphrases us without a citation. We cannot see inside a competitor’s product beyond what they publish. And we do not have enough traffic for Google to report real-world performance data about our own pages, which means some things other sites quote confidently are things we genuinely do not know. Where a number is an estimate, the page says estimate.

When we get it wrong

Tell us at alerts@99developer.com. A substantive correction is noted on the page rather than silently edited, for the same reason we publish the limits of our own product: a record you can check is worth more than a page that has always appeared to be right.