Inspect → Score → Diagnose → Prioritise → Fix → Verify → Monitor

We crawl the site as a search engine would, render each template at mobile and desktop, diff the raw HTML against the rendered DOM, resolve the full link and redirect graph, and pull field performance data and Search Console coverage where connected.

Output: A complete observed dataset. Nothing self-reported, nothing assumed.

Each of the 45 checks in SW-1:2026.2 is evaluated against its published criterion and returns pass, advisory, fail or not-applicable. Weighted arithmetic produces category scores and an overall score.

Output: A score you can recompute by hand, and a determination that is separate from it.

Every non-passing check gets a finding: the specific URLs, headers, values or measurements observed, plus a plain-language explanation of the mechanism by which it costs you.

Output: Evidence a developer can act on without reproducing the diagnosis themselves.

Findings are ranked by severity × impact ÷ effort. This deliberately favours small high-impact jobs, because the strongest predictor of SEO work getting done is whether the first item is achievable this week.

Output: A single ordered queue. Not nine category lists.

Each item becomes a work order with scope, acceptance criteria and a cost estimate. Do it yourself with the published remedy steps, hand it to your developer as a ready-written ticket, or assign it to a matched specialist.

Output: Work in progress, with scope locked before anyone starts.

We re-run the exact check that failed against its exact criterion. Not a fresh opinion, not a general re-audit — the same test. It either passes now or it does not.

Output: An objective done, and a certificate when the queue is clear.

Continuous re-inspection on a schedule and after deploys. Most severe faults are introduced by a site's own deployments, and the median time to notice one manually is measured in months.

Output: An alert within 24 hours instead of a lost quarter.