Topic Clusters Start With a Map, Not a Word Count

The map is the deliverable

Most content plans arrive as a list of titles. Twenty rows in a spreadsheet, a publish date beside each one, and an assumption that volume will sort out the rest. That plan can be executed perfectly and still produce nothing, because the part that makes a cluster work is not the writing. It is the structure the writing sits inside.

The structure is a map. On one side sits a pillar page for a service the business actually sells. Underneath it sits a set of supporting pages, each one holding a single question or a single narrow search that the pillar mentions but does not fully answer. Every supporting page links back up to the pillar. The pillar links down to each of them. That is the entire shape, and it can be drawn on paper before a single sentence exists.

Drawing it first is what makes the writing cheap and fast. When the map exists, whoever opens a supporting page already knows the exact query that page owns, what the pillar has already covered, and which sibling pages it must not repeat. Without the map, the writer is guessing, and the guess almost always lands on top of something already published. That is how a site ends up with nine articles about pool resurfacing, all of them competing with each other, none of them ranking.

So the map is the product. The writing that follows is execution against a decision that was already made.

Drawing one pillar per service that sells

The first pass is subtraction. A business usually names eight or ten things it does. Two or three of those pay the bills. The pillars get drawn for the ones that pay, not the ones that sound complete on a capabilities list.

A marine contractor in Naples might do seawalls, boat lifts, docks, pilings, minor repairs, inspections, and permitting help. Seawalls and lifts are where the invoices come from. Those get pillars. Permitting help gets a supporting page under whichever pillar it feeds. The temptation to give every service its own pillar produces a wide, shallow site where nothing has enough underneath it to matter.

Once the pillars are set, the pillar page itself gets defined by what it does not cover. It answers the broad buying query well enough to convert, then hands off the narrow questions to the pages below it. A pillar that tries to answer everything leaves its supporting pages with nothing to own, and the cluster collapses into one long page with links stapled to the bottom.

One query per page, and killing overlap on paper

Every supporting page gets exactly one assigned query. It is written into the map as a sentence, in the words a customer would actually use, and no other page in the cluster is allowed to claim it. That constraint does more work than any writing rule.

Overlap is detected by reading the assigned queries side by side and asking a simple question about each pair: would a person searching one of these be satisfied by the page built for the other? If the answer is yes, one of the two pages does not exist. It gets merged into its neighbor, or it gets re-cut into a narrower question that stands on its own.

This is also where a plan gets honest about size. A cluster built from real distinct questions is usually smaller than the spreadsheet everyone started with, and every page left standing has a reason to exist. Cutting eight duplicate ideas at the map stage is a better outcome than publishing them and spending a year wondering why traffic is flat. The same discipline governs how the silo is arranged on the site itself.

There is a second reason one query per page matters, and it shows up in AI answers. When several pages across a site each cover one facet of the same subject, and they agree with each other, an engine assembling an answer sees the same position stated consistently from multiple directions. That repetition reads as a site that is a source on the subject rather than a page that happened to mention it once. Contradictory or near duplicate pages produce the opposite: the same subject stated three slightly different ways, with nothing to corroborate.

The link contract between a pillar and its spokes

Links inside a cluster are not decoration and they are not "related posts" pulled by a plugin. They are a contract written into the map, and each page has obligations on both ends of it.

• Every supporting page links up to its pillar, in the body copy, using words that describe the service rather than the phrase "click here". One link, high on the page, is enough.
• The pillar links down to every one of its supporting pages, from the section of the pillar that raises that question. A supporting page the pillar never mentions is an orphan.
• Supporting pages link sideways only when a reader genuinely needs the neighbor. Two or three sideways links per page, not a wall of them.
• Nothing in a cluster links into a different pillar's supporting pages. Cross traffic between clusters goes pillar to pillar, so the boundaries stay legible.

Written this way, the link structure states which page is the authority on the broad service and which pages are the detail work underneath it. That is a claim about the site, made in structure rather than in adjectives, and it is the same principle that governs internal linking across a local site. Where a cluster serves specific cities, the same contract runs alongside the local content strategy so that city pages and topic pages are not fighting for the same searches.

Build order runs outward from the money

Once the map is clean, order matters more than speed. The sequence that works is the one that gets a complete, converting unit live early, then extends it.

1. The pillar for the highest revenue service, plus the three or four supporting pages a buyer actually reads before calling. Cost, process, and the objection that kills most deals. That group can produce leads on its own.
2. The rest of that pillar's supporting pages, working down from the questions people ask most often to the ones that come up occasionally.
3. The second pillar and its core group, built the same way, with the pillar to pillar link added in both directions.
4. The long tail of both clusters, the narrow questions that bring in small volumes of very qualified readers and give the cluster its depth.
5. The refresh pass, where early pages get updated against what search data has since revealed about how people phrase these questions.

Building depth first on a service that rarely closes is the most common way a good map produces a bad year. The pages get published, the counts look healthy, and nothing that arrives is worth answering the phone for. Ordering by revenue means the first thing that ranks is the thing you wanted ranking. The service page copy at the center of each cluster is what turns that ranking into a call, so it gets written before the supporting pages, not after.

What a cluster built backwards looks like

The failure is easy to recognize because it looks like progress. A long, careful pillar page goes live. It is thorough, it is well written, and it sits there for six months doing nothing, because nothing supports it.

A pillar earns a broad, competitive search partly on the strength of what surrounds it. On its own it is one page making a large claim with no corroboration behind it, competing against sites where the same claim is backed by a dozen pages of detail. Meanwhile the narrow questions that the business could win quickly, the ones with less competition and readier buyers, are still unwritten, so there is no traffic at all while everyone waits for the pillar to move.

The other version of backwards is publishing supporting pages with no pillar. Ten articles, each ranking a little, each linking nowhere in particular, and no page positioned to catch a reader who moves from a specific question to hiring someone. Traffic arrives and leaves. The fix in both cases is the same and it is not more writing: go back to the map, put the missing half in, and check the link contract before adding a single new page. The shorter explanation of how pillars and spokes fit together is a useful reference while you do it.

If you want to see what your current pages already cover and where they are quietly competing with each other, that is exactly what the cluster map exposes. Start with a free SEO audit and you will get the map drawn against your real site, alongside the rest of the content and copywriting work.

Frequently asked questions