Keyword research that ends in a page map, not a list

Most keyword research stops one step before the decision. You get four thousand rows, a difficulty column, and no answer to the question the roadmap actually needs: how many pages do we build, which one wins which query, and in what order. This engagement ends in a page map — one line per page, the query that page must win, the queries it absorbs, the query it must not take because another page already owns it, and a build sequence. On a mid-size site the map usually lands between 30 and 60 pages, and it is normally fewer pages than the brief assumed. Who does the work: I do, start to finish. Six years in-house running organic search for a fintech and then a marketplace before going fractional, which mostly means I have already been the person defending a page count to a CFO in January and explaining the traffic in June. Nobody junior touches the clustering and no part of it is outsourced. How pages get decided: Clustering runs on SERP overlap, not string similarity and not an embedding score. Two queries share a page when their top ten organic results share three or more URLs, pulled the same day, same country, same device, through the DataForSEO SERP API. At two shared URLs they get separate pages. That threshold is arbitrary in the way all thresholds are arbitrary, but it is fixed before I see your data and it is written into the file, so you can rerun it and get my clusters back without me. What arrives: A workbook with one tab per decision — the page map, the clusters with their overlap evidence, the queries I dropped and the reason each was dropped, and a redirect list where existing URLs collapse into a new page. A written brief for every page in the first build wave: the query it must win, the page type, what has to exist on it for the current top ten to be beatable, and the internal links it needs on day one. A ninety-minute call with whoever signs off the roadmap, recorded. And a one-page note written for a CEO, because that is usually the conversation that decides whether any of this gets built. When the answer is inconvenient: It often is. The most common finding is that a planned section supports far less demand than the plan assumes, and that two or three pages you already have are competing for the same result and each losing. I will tell you to build fewer pages and delete some of what exists. I will also tell you when a query belongs to a marketplace or a publisher and no page you can build will take it, which is a cheaper sentence to hear now than after a quarter of writing. If the honest conclusion is that the money belongs in a different channel, the deliverable says that. What this is not: No content writing, no briefs beyond the first wave, no link acquisition, no CMS implementation, no technical fixes, no rank tracking afterwards. I do not run monthly reporting against the map. If you want it maintained as the site changes, that is a separate arrangement and I will say so rather than quietly extending scope. Who should not buy this: Teams that have already decided what to build and need research to support it. Anyone who needs the file by Friday for writers who start Monday — the SERP pulls alone take days and the arguing takes longer. Sites where nobody in the room has the authority to delete pages, because half the value here is subtraction. And anyone whose measure of success is how many keywords are in the file; that number goes down here, deliberately. Why it is bought: The cost of this work is small against the cost of building forty pages that split their own demand. Most clients buy it before a rebuild, a migration, or a funded content push — the moment the page count is about to be locked and changing it afterwards means changing URLs that already exist and already rank.