Technical SEO prioritization: which issues should I fix first?

By the severity and confidence encoded in the NovaBrain criterion each finding belongs to, combined with its reach across affected pages, never by how often a problem occurs. The fix-first queue starts with what the standard says matters most and opens each finding to the URLs and the captured field behind it.

Because a finding you cannot open to its pages is a category, and engineers do not fix categories. Opening the affected set shows whether one template explains a thousand pages or a thousand pages need a thousand edits, which changes both the priority and the ticket, and it keeps the audit's evidence attached to the work.

No. A recommendation is not a verified fix, and the workflow will not mark work complete because it was reported complete. Verification requires a fresh crawl in which the failing condition is gone; the same criterion re-runs, updates the finding when the pages pass, and can reopen it later if the condition returns.

It reads the same audit counts and severities and returns an impact-ranked plan in plain language, with deep links to the pages and the tool that hold the evidence, and it declines when the evidence is missing. It is the sentence version of the queue, useful when a founder or a lead wants the order without reading every cockpit.

It counts them as needing evidence rather than grading them, so they neither inflate the health percentage nor enter the queue as findings. A criterion whose required field the crawl did not capture is reported as unmeasured, which tells you what to capture next rather than pretending a pass or a fail.

After each batch of fixes and on a regular crawl cadence otherwise. Each crawl is a timestamped record and each audit keeps its history, so a re-run is a comparison: findings closed on fresh evidence, findings reopened, new findings from new pages, each with its severity and pages, in the order the standard assigns.