Keyword research template: cluster, prioritise, assign

Collection is easy; organisation is the job

Every keyword tool will hand you thousands of phrases in a few seconds, which is precisely why a raw export is worth so little. The value comes from the three decisions that follow: which of these phrases represent the same underlying question, which page will answer each of those questions, and in what order you will build them. This template exists to hold those three decisions in a form someone else can act on.

Start by writing down the setup — seeds, market, language, and which tool produced which numbers. That last one matters more than it looks. Difficulty scores from different providers are calculated differently and are not comparable, so a sheet mixing them without labels quietly produces incoherent prioritisation. The excluded field is equally useful: recording that you deliberately dropped competitor brand terms or a market you do not serve stops the same conversation repeating every quarter.

Clustering: one cluster, one page

A cluster is a set of keywords that one page can satisfy. The reliable test is to look at the actual results: if the top-ranking pages for two keywords are largely the same pages, search engines consider them the same question and one page should target both. If the results look completely different, they are separate clusters however similar the wording seems.

Once a cluster exists, give it exactly one target URL and put that URL on every row in the cluster. This single discipline prevents the most common self-inflicted SEO problem: two or three of your own pages competing for the same query, splitting internal links and impressions, none of them strong enough to win. When you later find yourself briefing a new page for a keyword that already has a target URL, the sheet tells you to improve the existing page instead.

Keep clusters distinct enough that a reader could tell them apart from the headings alone. If you cannot explain the difference between two clusters in one sentence, they are one cluster and should be merged before anything gets built.

Intent decides format, not word count

Classifying intent is not an academic exercise — it determines what kind of page you build. Informational queries are answered by guides, glossary entries, checklists and templates. Commercial-investigation queries are answered by comparisons, alternatives pages and honest listicles. Transactional queries want a landing page with pricing and a clear next step, not a blog post. Navigational queries are usually about a brand and rarely need new pages at all.

The costliest mistake here is publishing an essay against a query whose results are entirely tools, or a thin landing page against a query whose results are all in-depth guides. The format mismatch caps the page's ceiling regardless of how well it is written. Note SERP features in the sheet for the same reason: a featured snippet changes how you structure the opening, a map pack means local signals matter more than content, and a results page dominated by video tells you the medium matters as much as the words.

Prioritising without kidding yourself

Volume is the worst primary sort. High-volume head terms are usually the hardest to win, often carry vaguer intent, and can take a year to move on a young site. A more useful order is: how close the query sits to revenue, then how close you already are to ranking, then difficulty relative to your own site rather than in the abstract, then volume as a tiebreak.

The 'how close you already are' factor deserves particular attention. Queries where you sit around positions five to twenty are already eligible — search engines have decided your page is a reasonable answer and merely ranked it below others. Improving those pages typically moves faster than creating anything new, which is why the current-position column belongs in the sheet even though it needs Search Console to fill in.

Judge difficulty relative to your own site, not against an absolute scale. A difficulty score of thirty means something different for a domain with years of history and a strong internal linking structure than it does for a site published last month. Use the score as a rough ordering signal within your own list rather than as a threshold copied from someone else's advice, and always note which tool produced it so future readers can recalibrate.

Finally, keep the status column current. A keyword sheet where every row says 'not started' six months later is not a research artefact, it is evidence that the research never connected to the publishing process. Rows should move through briefed, drafting, in review and published — and the moment a cluster's pillar goes live, the supporting rows become the next obvious work.