Topical Maps: Turning Keyword Lists Into Territory

The traditional keyword process produces a spreadsheet sorted by search volume, which gets attacked from the top down. Six months later the site has forty unrelated articles, each competing alone against a domain with a decade of coverage on that subject. The pages are individually fine and collectively meaningless.

A topical map inverts the sequence. It starts from the subject you intend to be the reference for, breaks that subject into its constituent entities, questions and tasks, and only then attaches keywords and volumes. The output is a coverage plan with an end state, which is the part that makes it useful.

Building the map

Start with the core entity — the thing your business is about. For a payroll product it might be "payroll for small businesses". Then expand along four axes.

• Attributes: the properties of the entity. Cost, timing, compliance, accuracy, software.
• Relations: other entities it connects to. Tax authorities, accountants, employment contracts, benefits.
• Tasks: what a person does with it. Run payroll, correct an error, onboard an employee, file a return.
• Situations: the contexts that change the answer. First employee, cross-border staff, contractors, seasonal workforce.

Each intersection is a candidate page. "Correcting a payroll error for a cross-border employee" has essentially no search volume in any tool and will be read by exactly the person who is about to buy payroll software. The map surfaces these; a volume-sorted list buries them.

Structure: hubs and spokes, with real links

Each cluster needs a hub page that covers the subject comprehensively and links to every supporting page, and supporting pages that link back to the hub and sideways to their two or three nearest siblings. This is not a formality. Internal linking is how you tell a crawler which pages belong together and which one is the canonical answer for the broad term.

Page type	Purpose	Typical length	Links out
Hub	Define the subject, route to depth	2,000-3,500 words	All spokes
Task spoke	Answer one procedure completely	900-1,600 words	Hub + 2-3 siblings
Comparison spoke	Decision support between options	1,200-2,000 words	Hub + related tasks
Definition spoke	Capture the entity term itself	600-1,000 words	Hub + task pages

A cluster with twelve pages and no internal links is twelve orphans. The links are the map; the pages are only the territory.

Sequencing: narrow first

The instinct is to publish the hub first because it targets the biggest term. In a young or mid-authority domain that is backwards. The hub competes for the hardest keyword in the cluster and will sit on page four for months, which demoralises everyone.

Publish the narrowest, most specific pages first. They rank within weeks because the competition is thin, they generate the early impression data that tells you whether the cluster is viable, and they give the hub something to inherit authority from when it launches. Then publish the hub and link everything together.

The same logic applies at cluster level. Finish one narrow cluster completely before starting a broader one. Half-covering three subjects is worse than fully covering one, because demonstrated completeness is precisely the signal you are trying to send.

Handling overlap and cannibalisation

Maps produce near-duplicate candidates: "payroll software cost" and "how much does payroll software cost". Do not create both. Merge candidates when the search results for them are more than about 60% identical — that overlap is the engine telling you it considers the two queries the same job.

When cannibalisation appears after publishing (two of your pages swapping positions for the same term), the fix is usually consolidation rather than differentiation: pick the stronger URL, merge the unique content in, redirect the weaker one, and update internal links to point at the survivor.

Measuring a map, not a page

Page-level reporting hides cluster progress. Track three cluster-level series instead: share of mapped pages published, total cluster impressions, and the count of queries where any page in the cluster ranks in the top ten. The third one is the closest available proxy for topical authority, and it typically moves in steps rather than smoothly — flat for weeks, then a jump as the cluster crosses a coverage threshold.

Where maps fail

Two failure modes are common. The first is mapping a subject the business has no right to own — coverage does not substitute for relevance, and a payroll company writing forty pages about general management theory will rank for none of it. The second is treating the map as fixed. Search demand shifts, competitors publish, and new situations appear. Review the map quarterly, add the situations your sales team is actually hearing, and retire branches that never produced impressions.

Done properly, the map becomes the content roadmap, the internal linking spec and the reporting structure at once — which is why it survives the arrival of the next content trend intact.