Methodology

What we measure

We use Search Console and privacy-preserving analytics to understand where users are getting stuck. The core measures are impressions, clicks, click-through rate, average position, device, country, triage completions, calculator completions, and outbound clicks to official claim routes.

We separate source-citation clicks from official-action clicks. A source click tells us a reader is checking trust. An official-action click tells us a page helped someone reach the next step.

How search data is used

Search data does not decide the editorial answer. It decides which questions need clearer answers. When a page has strong impressions but weak clicks, we look first at title, description, above-fold answer, and whether the page answers the query directly enough. When a page has weak position, we look first at coverage depth, internal links, and whether a better route exists elsewhere on the site.

Query clusters are tracked with the local `pnpm gsc:analyze` script so the same categories can be compared month to month. The current recovery clusters are benefit cap, Healthy Start, self-employed Universal Credit / Minimum Income Floor, uprating/rates, and council local help.

What we do not optimise for

We do not chase clicks by making a page sound more certain than the law allows. We do not hide official sources below commercial links. We do not publish local pages unless they contain useful local information or a direct official route. We do not use personal triage answers for content targeting.

Review and correction loop

Each substantive update is checked against primary sources, reviewed by a named person, and logged through the same operating model described in our editorial policy. Recent page checks appear in the update log. Errors can be reported through the corrections page.