How we handle evidence

A named study or dataset, cited with author and year, where the number in the sentence is the number in the paper. Where the study is small, old, from another country, or measured something narrower than the claim implies, we say so in the same breath. A statistic without its limitations is not really a statistic.

No single decisive study, but a consistent direction across a literature — for example that sustained stress affects sleep, cognition and immune function. Used when the claim is well supported in general but would be misrepresented by attaching it to one paper and one number.

Patterns noticed in how hiring works, in how people often respond to financial pressure, and in what tends to happen as a search becomes prolonged. Nobody has measured these under controlled conditions, and they are offered as informed observation rather than as findings — which means they may hold for most people and still not hold for you, and that is a reasonable thing to conclude.

Something worth attempting, which may or may not work for you. Not dressed up as a finding. If it does nothing for you after a fair attempt, that is information about the technique, not about you.

The obvious approach to a repository like this is to delete anything that cannot be footnoted to a randomised controlled trial. That would strip out most of what people actually find useful — the observation that confidence falls faster than market value, that silence usually says more about the volume an employer is handling than about the application, that the question "so what are you doing now?" stops being an ambush once you have an answer ready. None of those are empirical claims. All of them help.

Labelling rather than deleting lets the site keep more material, not less — while making sure nobody mistakes a practical observation for a clinical finding. It also means that when something is a research finding, that label carries weight.

Every statistic used here is checked against its original source rather than against the article that repeated it. Numbers that circulate widely are often mis-attributed — a figure from one meta-analysis attached to another, a US survey presented as UK data, an effect size converted into a percentage it never was. Where a well-known claim turned out not to survive that check, it has been removed or replaced rather than softened.

Two recurring problems are flagged throughout the site rather than quietly smoothed over. First, a great deal of job-loss and mental-health research is American, and is signposted as such rather than presented as though it described the UK. Second, several of the best-known "UK" services are England-only — NHS Talking Therapies self-referral, the Jobs and Careers Service (formerly the National Careers Service), the main Citizens Advice phone line — so every entry in the directory states where it actually applies.