Which page type each query wants, and where yours is wrong

A page that does everything right on-page and still sits at position 14 usually has nothing wrong with it. It is the wrong kind of page for the query. Google has already settled that this search wants a comparison table, or a forum thread, or a filterable listing, or a clinician-bylined page — and a long-form article is answering a question no result in that set is answering. This engagement tells you, query by query, what shape the answer has to be, and where yours is a different shape. The question it settles: Rewrite, rebuild, replace, or walk away. Those are the only four verdicts I issue, and every row in the deliverable carries exactly one. Rewrite means the page type is right and the treatment is thin. Rebuild means the URL survives and the page becomes something else. Replace means a new URL and a redirect. Walk away means the top ten belongs to structures you cannot make — a marketplace's inventory, a subreddit, a government register — and any budget aimed there is spent. How the reading is done: Backwards, from the result set. For each query I record the page type of all ten organic results, the publish and update dates showing on them, which SERP features sit above the fold, whether the ranking pages are transactional or editorial, and who owns them: brand, publisher, marketplace, community, or institution. The rule I apply is fixed. If fewer than three of the top ten are a page type you could credibly build and get indexed, the verdict is walk away, whatever the volume says. That rule costs me work and I keep it, because the alternative is charging a client to lose slowly. Read by hand, not by classifier: SERP data comes from DataForSEO and Ahrefs; the classification is mine, one result at a time. Automated page-type labelling fails on exactly the queries where the money is — the ones where a category page and a buying guide look identical to a machine and behave nothing alike in a click. The sets I read are small on purpose. Sixty queries read properly beats six hundred labelled by a script, and the six hundred are what put a page at position 14 in the first place. Delivered as: A sheet with one row per query carrying the verdict, the dominant page type, the evidence counts, and the named competing result that sets the bar. Annotated SERP screenshots for the calls that were close. A wireframe-level outline — sections and their order, no copy — for each page in the rebuild and replace groups. And a written ranking of the whole set by what acting on it costs against what it protects, so a head of growth can take the top five into a planning meeting without reading the rest. Where it gets uncomfortable: Some pages come back as replace that your team is proud of. I have told a client to retire an eight-thousand-word guide that earned links and no rankings, because every URL above it was a filterable listing with faceted navigation. I have told two others that their highest-volume target should be abandoned. If you would rather not hear that about a flagship page, this is the wrong engagement and I would rather say it before you pay. Outside the scope: Copywriting, design, front-end work, schema implementation, keyword discovery, and anything touching links. I read what the result set rewards and hand you the shape. Building it is yours, or your agency's. Not a fit for: New domains with no ranking history, because there is nothing to diagnose. Teams shopping for a second opinion to overrule an agency they have already stopped trusting. And anyone who needs a yes on a query already committed to in a plan: roughly one query in six comes back as walk away, and that verdict is the part of the engagement that pays for itself.