SaaS topic clusters

Count the spokes under your biggest content hub. If the answer is somewhere between six and twelve, you do not have a cluster. You have a landing page with a reading list attached, and it is not going to rank for the head term you built it around.

The hub is the part everyone polishes and the part that matters least. What decides whether a hub competes is how many genuinely useful pages sit beneath it answering the questions the head term implies, and whether those pages are wired together in a way a crawler can follow.

How many spokes before a hub can compete

Count what the incumbents have. Take the five sites currently ranking in the top five for your head term, run a site search or a site-scoped export in Ahrefs or Semrush restricted to that subject, and count their pages on the topic. Your target is the highest number in that list.

In most SaaS categories that count lands between 24 and 40. Categories with heavy editorial competition, like project management or CRM, routinely run past 60. Narrow technical categories sometimes clear at 15.

The 12 to 20 row is where most SaaS blogs live, and it explains the plateau that every content lead eventually hits. Enough pages to be expensive. Not enough to signal that you own the subject. The traffic curve flattens, someone proposes a redesign, and the actual problem, which is that the cluster is 40 percent built, never gets named.

One qualifier on the counting method. Incumbent page counts tell you what it took to win historically, not what it takes now. Where AI answer engines summarise the head term directly, coverage still matters for citation frequency, but the traffic payoff shifts toward the spokes and the money pages. Size the cluster on coverage, forecast the return on the spokes, and use the organic traffic forecast calculator to test whether the arithmetic survives a lower hub click-through rate.

A worked 40-page cluster for one SaaS category

Take an HR software company whose head term is employee onboarding software. Here is how 40 pages distribute across the cluster, and what each group is for.

Two things about this distribution are deliberate. The informational spokes are the largest group and produce the least direct pipeline, which is the trade every cluster makes: they exist to make the hub and the money pages rankable. And the templates layer is small but carries the link acquisition, because a calculator or a checklist earns links that a 2,000-word guide does not.

Notice what is absent. There is no page targeting hr software, because that belongs to a different cluster with a different money page. Cluster boundaries are defined by which money page the traffic should reach, not by topical adjacency. If two clusters point at the same product page, they are one cluster.

URL and navigation choices that do not break at page 30

Group URLs by content type rather than by cluster: /guides/, /templates/, /research/, /comparisons/. Express cluster membership through internal links, breadcrumbs and a hub index, not through folder depth.

Nesting by cluster looks tidy in a diagram and fails in practice. By spoke 20 you will have a page that belongs to two clusters, and under a /cluster/page/ scheme you must pick one, duplicate it, or restructure. Under a type-based scheme, the page simply receives links from both hubs.

Keep everything on the main domain in a subfolder. A content hub on a separate subdomain starts from scratch on authority and makes internal linking between marketing pages and cluster pages a cross-site problem you have manufactured for yourself. The full argument, including the cases where a subdomain genuinely wins, is in subfolder versus subdomain, and the site-wide version of these decisions sits in SaaS site architecture for SEO.

Navigation carries a second job people forget. A hub reachable only from a blog index gets crawled like a blog post. A hub reachable from the main navigation or the footer gets crawled like a section. If the cluster matters enough to build 40 pages, it matters enough for a navigation slot.

The internal link contract, written down before you draft anything

Define the rules once, put them in the brief template, and enforce them at review. Internal linking decided page by page produces a cluster where six pages have forty inbound links and thirty have one.

The retrofit rule is the one that gets dropped and the one that matters most. A spoke published in month nine with no inbound links from the existing 28 pages is an orphan wearing a cluster badge. Build it into the publishing checklist: nobody marks a post done until two older pages link to it.

The two-siblings rule keeps link equity circulating without turning every page into a link farm. Three is defensible. Ten is not, because at that point the links stop being editorial and start being decoration, and readers learn to ignore them.

Money page distribution is the part that connects this to revenue. Forty spokes each sending one contextual link to the relevant product or pricing page is a substantial internal link signal to a page that usually earns very few links naturally. That is a large share of why clusters improve conversion on pages that were not part of the cluster at all.

One keyword, one page, enforced by a map

Maintain a keyword map where every target query has exactly one canonical owner URL, and refuse to commission a page whose primary keyword is already owned. This is a governance rule, not an SEO tactic, and it is the only reliable defence against cannibalisation.

The map needs four columns to work: the keyword, the owning URL, the search intent, and the date the assignment was made. Anything more elaborate stops being maintained by month four. Our SaaS keyword map template has the structure with the extra columns marked optional for exactly that reason.

Cannibalisation shows up in three recognisable patterns.

• Position swapping. Two URLs trade places for the same query across weeks. Consolidate into the stronger page and redirect the weaker one.
• Neither page ranking. Two pages each cover half a topic and neither has enough depth. Merge them and expand.
• The wrong page ranking. Google prefers a blog post over your product page for a commercial query. Usually an internal link problem: the blog post has links, the money page does not.

Consolidation is uncomfortable and mostly correct. When two pages compete, merging into one and redirecting typically recovers more than keeping both, because the combined page inherits both link profiles and both sets of coverage. Teams resist because deleting published work feels like waste. The waste already happened; the redirect is the recovery.

Sequence the build so the cluster pays before it is finished

Publish in an order that produces revenue early and coverage later, because a cluster that shows nothing for eight months loses its budget in month five. The sequencing rule: money-adjacent pages first, coverage pages second, hub last.

Month 8 is the argument you will have with someone. Publishing the hub last feels backwards, and there is a legitimate counter-case: an early hub gives the spokes somewhere to link up to from day one. A workable compromise is a short hub in month one, 800 words, updated into the real thing in month eight.

Report at cluster level, because per-post traffic hides the effect

Three numbers, reported monthly, at cluster level. Per-post session counts are the wrong instrument and they cause teams to kill spokes that are doing structural work.

Share of voice against three named rivals is the metric that survives a board conversation, because it is competitive rather than absolute. Organic sessions fell 12 percent while share of voice rose 6 points is a coherent story about a market where AI answers absorb clicks. Organic sessions fell 12 percent, on its own, is a story about you failing.

Set the reporting cadence at monthly and the review cadence at quarterly. Cluster effects are slow, and looking weekly produces reactive decisions on noise. When it is time to defend the budget, run the numbers through the SaaS SEO ROI calculator so the payback period is stated rather than implied.

Half-built clusters are the reason most SaaS blogs plateau

The common failure is not bad writing. It is a portfolio of six clusters at 40 percent completion, each one stuck below the coverage threshold where hubs start ranking, competing with each other for the same internal links and the same writer’s time.

The fix is uncomfortable and simple: pick the cluster closest to its target coverage, finish it, and freeze the others. Frozen clusters keep their existing traffic. Three finished clusters outperform eight abandoned ones by a wide margin, on traffic, on conversion and on the effort needed to maintain them.

Two honest costs come with this approach. A 40-page cluster at four posts a month is roughly ten months and, at typical SaaS content rates, somewhere between 40,000 and 120,000 dollars of production depending on the mix of freelancers, agency and in-house writers. And you will publish spokes that never get meaningful traffic on their own. Both are the price of the coverage signal, and both are why clusters should be started deliberately rather than drifted into.

Decide the cluster boundary by money page, count the incumbents, commit to the number, and write the link contract before the first brief. Then start with the seven bottom-funnel pages so something converts by month three. The definition and the shorter conceptual version live in the topic cluster glossary entry and topic clusters for SaaS content, the brief structure that enforces the link contract is in the programmatic page brief template, and the wider prioritisation of clusters against other organic work sits in the SaaS SEO hub.