Technical audit of a publishing estate, ranked by revenue risk

Large content estates lose their index coverage in three or four places, not fifty. Across the last four estates we examined, ranging from nine thousand to a hundred and twenty thousand published URLs, the biggest single cause was the same every time: pages that were crawled, rendered, then dropped, because something in a template told Google the page was a duplicate of a listing. This engagement names your version of that, puts the findings in order of what each one costs you, and hands your engineers a queue they can start on Monday. Evidence we work from: Server logs, twenty-eight days minimum, taken from the CDN or the origin, whichever sees Googlebot unfiltered. Without logs this is inference with a cover page, so we ask for them before quoting rather than after. Index state comes from Search Console URL Inspection plus the Coverage export, not from a rank tracker. Rendering is checked twice, once through the live test and once through a headless Chrome pass matched to Googlebot's viewport, because those two disagree more often than people expect. The internal link graph comes from a full crawl with the JavaScript renderer enabled. Field performance comes from CrUX, not from a Lighthouse score generated on our own connection. Thresholds that trigger a finding: TTFB above 800ms at the seventy-fifth percentile. INP above 200ms. Any template where more than thirty percent of published URLs sit in Coverage as excluded. Any hreflang cluster missing return tags between en, uk and pl. Any URL parameter absorbing more than a tenth of Googlebot's requests. Numbers below those thresholds go in an appendix and we tell you not to spend engineering time on them. Three things this has actually found: A media estate spending sixty-two percent of crawl requests on faceted archive pages that had never once received a session from search. A B2B publisher whose best four thousand articles moved to client-side rendering during a framework upgrade and lost forty percent of impressions over eleven weeks, which nobody connected to the upgrade because the deploy had been six weeks earlier and had shipped a visible feature everyone was pleased with. A catalogue whose canonical tags were correct in the source HTML and rewritten by an edge worker that predated everyone currently on the team. What arrives: A written diagnosis of twelve to twenty pages, ordered by commercial exposure rather than by a severity label a tool assigned. A ticket set written into your own tracker, Jira or Linear, each ticket carrying acceptance criteria an engineer can close against and sized by your team in a session with us. One hour with your engineering lead and one with whoever owns the roadmap, because these tickets compete with product work and lose that argument unless somebody senior has seen the number attached. You will not receive a spreadsheet of nine hundred issues sorted by colour. Where we will disagree with you: Sometimes the honest conclusion is that the technical layer is not your problem. Twice last year the finding was that the pages were indexed, ranked, and simply not worth reading next to what sat above them. That is an editorial problem wearing technical clothes, and we say so in the document. We do not then sell the remediation as an add-on, because taking payment to tune crawl efficiency on an estate that needs editing is how audits earned their reputation. Outside scope: We do not implement fixes on the two lower tiers, we do not write or rewrite pages, we do not do link work, and we do not manage your CMS migration. We will read your migration plan and mark the parts that will cost you traffic, which is a different and much cheaper thing to buy. Not for you if: Your estate is under roughly five hundred URLs. A free crawler and one focused afternoon will tell you the same, and we will say that on the intro call rather than take the entry fee. You need a document to show a board that a decision already taken was correct; we have declined that twice and would again. You cannot get server logs released by your infrastructure team, in which case fix that first and come back, because everything downstream of the log file is weaker without it. Languages and timing: Delivery in English, Ukrainian or Polish. Sessions run in your working hours across European and North American time zones. We hold two audit slots per month and do not oversell them; if the next opening is five weeks out we give you the date instead of starting late.