A site publishes 60 articles across eleven unrelated subjects over two years. Traffic barely moves. Another site publishes 25 articles, all circling one subject, tightly linked. It starts ranking for terms it never explicitly targeted. The difference is not effort or word count. It is coherence.
Search engines build an understanding of what a domain is credible about. Scattered publishing gives them nothing to work with. Concentrated publishing gives them a clear answer, & that answer benefits every page in the group.
Choose the topic before you choose the keywords
The instinct is to open a keyword tool & export everything with volume. That produces a list, not a strategy, & the list will contain terms you have no business ranking for.
Start with a harder question: which single subject could you plausibly become the best resource on within twelve months? The answer is constrained by your actual expertise, your commercial priorities & the competitive reality of the space. It is usually narrower than you would like.
Narrow is correct. Owning a specific subject completely is achievable. Covering a broad subject partially is not, & partial coverage of a broad subject is exactly what most content programmes produce.
Map the topic the way a customer experiences it
Once the topic is chosen, map it properly. Not as a keyword list, but as the set of questions & decisions a person moves through.
Work in three layers:
- Awareness. They have a problem but do not know what solves it. Questions here are symptomatic: why does this happen, is this normal, what does this cost.
- Evaluation. They know the category & are comparing approaches. Questions are comparative: which option suits my situation, what are the trade-offs, what should I ask a supplier.
- Decision. They are choosing a provider. Questions are specific: how does the process work, what is included, how long does it take, what does it cost here.
Sources for the map: your own sales conversations, support tickets, the questions people actually ask on forums & social platforms, & the related searches that appear alongside your core query. Keyword tools help you size & phrase these questions. They should not generate them.
If your cluster contains no pages that answer decision-stage questions, it will produce traffic & no enquiries. That is the most common failure mode.
Structure: one pillar, many supporting pages
The standard structure works because it mirrors how the topic is actually organised.
The pillar page covers the whole subject at a useful depth. It is comprehensive rather than exhaustive: it answers the main question fully & points to detail elsewhere. It targets the broad, competitive head term. It is usually the hardest page to write & should be written last, once you understand the topic through the supporting work.
The supporting pages each answer one specific question thoroughly. Each targets a narrower query with clearer intent. Each links up to the pillar & sideways to the two or three siblings that are genuinely related. Not to all of them; indiscriminate linking dilutes the signal & reads badly.
A workable cluster is one pillar & eight to fifteen supporting pages. Fewer than eight rarely establishes the pattern. More than fifteen usually means the topic should be split.
Internal linking is the mechanism, not the decoration
Clusters work because of the link structure, not because of a content calendar label. Get this part right & the rest follows.
- Every supporting page links to the pillar, in the body, with descriptive anchor text.
- The pillar links out to every supporting page, ideally in a contextual position rather than a list at the bottom.
- Supporting pages link to genuinely related siblings where a reader would benefit.
- Anchor text describes the destination in natural language. Repeating the exact same anchor across fifty links looks manufactured because it is.
- Older content on the same topic gets retrofitted with links into the new cluster. This step is skipped constantly & it is free.
Avoid cannibalisation while you build
The risk with clusters is that several pages end up targeting effectively the same query. Search engines then split signals between them & neither ranks well.
Prevent it by assigning one primary query per page & recording it. Before writing anything new, search your own site for existing coverage. If a page already addresses the question adequately, expand that page rather than creating a competitor to it.
Check quarterly. In Search Console, look for queries where your ranking URL changes between pages over time. That oscillation is the clearest symptom of two pages competing, & the fix is usually consolidation with a redirect.
Depth beats frequency
Publishing weekly is a habit, not a strategy. A cluster of fifteen genuinely thorough pages published over six months will outperform sixty shallow pages published over the same period, & it will cost less to maintain.
Practical markers of depth: the page answers follow-up questions rather than only the headline one, it contains specifics rather than generalities, it includes examples, & someone with expertise wrote or reviewed it. Length follows from depth. It is not a target.
Measure the cluster, not the page
Individual page reporting will mislead you here. A supporting page might attract modest traffic while materially strengthening the pillar's position. Judged alone it looks like a failure.
Report at cluster level:
- Total impressions & clicks across all URLs in the cluster.
- Number of distinct queries the cluster ranks for, which is the clearest signal of growing topical coverage.
- Average position of the pillar for its head term.
- Conversions attributed to any page in the cluster.
Expect a slow start & a compounding curve. Clusters typically show little for the first two to three months & then move, because coverage has to reach a threshold before the pattern registers.
Then maintain it
A finished cluster is not finished. Review it every six months: update anything factually stale, add pages for questions that have emerged, consolidate anything that underperforms, & refresh the pillar to reflect the new supporting content.
Once one cluster is working, build the second. Two well-executed clusters make a site look credible in a way that twenty scattered articles never will. See how we plan this in our content marketing engagements, or tell us which topic you want to own.