Most keyword research ends with a spreadsheet nobody opens again. The research was fine; what was missing is the four steps between a validated list and a published article.
This assumes you have already killed the keywords you cannot win — if not, run the keyword research checklist first, because clustering a list full of unwinnable terms just produces an unwinnable plan.
Step 1 — Cluster by parent topic, not by theme
The instinct is to group keywords by subject: everything about audits together, everything about links together. That produces tidy folders and tells you nothing about how many pages to build.
Group by parent topic instead — the term driving most traffic to the current #1 page for each keyword. Keywords sharing a parent topic are answered by one page in Google’s view, so they belong on one page in yours.
| Keyword | Volume | Parent topic | Page |
|---|---|---|---|
| seo audit template | 1,800 | seo audit template | Page A |
| website seo audit template | 150 | seo audit template | Page A |
| seo audit format | 200 | seo audit template | Page A |
| seo audit report sample | 1,100 | seo report pdf | Page B |
| seo audit example | 700 | seo report pdf | Page B |
Five keywords, two pages. Build five and three of them cannibalise each other; build one and you leave the second cluster uncovered.
This single rule replaces most cannibalisation debate. It is not judgement — it is Google telling you, through the ranking page, how it has already decided to group the queries.
Step 2 — Map every cluster to a URL
Every cluster gets exactly one destination, and it is either a page that exists or a page you are committing to create.
| Cluster | Primary keyword | URL | Status |
|---|---|---|---|
| seo audit template | seo audit template | /blog/seo-audit-template | Create |
| technical audit service | technical seo audit service | /services/technical-seo-audit | Create |
| keyword difficulty | keyword difficulty | /blog/keyword-difficulty-guide | Exists — improve |
“Exists — improve” is the row most plans are missing. Before committing to a new article, check Search Console for queries where you already rank 11–30. Those pages have Google’s relevance judgement already; moving one from position 14 to position 8 is faster and cheaper than earning a new position from zero, and it needs no new content — usually just better intent match and a few internal links.
A rough rule: if a page already earns impressions for the cluster, improving it beats writing a new one roughly nine times out of ten. Writing a second page for a cluster you already rank for is how sites cannibalise themselves while believing they are expanding.
Step 3 — Sequence by what unblocks the most
Not by volume. Volume is what everyone sorts on and it is the wrong first sort, because the highest-volume terms are usually the hardest and the slowest to pay off.
Sequence on three factors, in this order:
- Can it rank now? Difficulty inside your ceiling, weak domains present in the top 10.
- Does it support something commercial? An article feeding a page that converts is worth more than a higher-volume article feeding nothing.
- Does it unblock later work? Publish the supporting articles before the landing page they support, so the landing page launches with topical support rather than into a vacuum.
Point 3 changes publication order in a way most plans get backwards. If you intend to build a service page for technical audits, publish two or three articles about audits first. When the landing page goes live it already has relevant internal links pointing at it from pages Google has assessed.
Step 4 — Write briefs, not keyword assignments
Handing a writer “target: seo audit template, 1,800/mo, 1,500 words” guarantees a generic article. The brief needs to carry the research, or the research dies at this step.
A brief that works:
Primary keyword: seo audit template — 1,800/mo, KD 6 Secondary: seo audit format, website seo audit template Parent topic: seo audit template (self — safe to target directly) Search intent: artifact. Nine of the top 10 are downloadable templates or structured section lists, not essays. Ship the structure as a table in the first screen. Angle we can defend: the effort column most templates omit — impact-only ranking is why audits do not get implemented. Must link to:
/services/technical-seo-audit(commercial),/blog/technical-seo-checklist(depth, do not repeat its content) Do not cover: the technical checks themselves — that article exists and we link to it instead Length: whatever the structure needs. Do not pad to a target.
The “do not cover” line prevents the most common self-inflicted wound: two of your own articles covering the same ground because nobody told the second writer what the first one already said.
What this produces
At the end of these four steps you have a table where every row is a URL, a cluster, an intent, a destination link and a status. That table is the content plan. The keyword spreadsheet was never the plan — it was raw material.
Review it monthly against Search Console rather than quarterly against the plan. Keywords that were unwinnable at DR 14 become winnable at DR 25, and pages that stalled at position 12 usually need internal links rather than a rewrite.
Where plans fail after this point
Publishing without linking. New articles that link to nothing commercial, and that nothing links to, sit in isolation. Wire each one on the day it publishes — see internal linking and orphan pages.
Abandoning the improve rows. New articles feel like progress; improving an existing page at position 14 does not. The second is usually worth more.
Ignoring intent drift. SERPs change format. A keyword that returned articles last year may return tools now, which quietly invalidates a planned page. Re-check intent for anything sitting in the backlog more than a few months.
If the plan is sound and pages still will not move, the constraint is usually technical rather than editorial — a technical SEO audit establishes whether the site can rank before you spend another quarter on content. For the ongoing version of this work, our SEO services cover research, mapping and briefs as one loop rather than three disconnected exercises.
