Semantic and intent clustering

What it is

Keyword clustering is the difference between a content plan and a list of words. The job is to work out which keywords mean the same thing to a searcher, so that one page targets one intent instead of several pages competing with each other.

This goes beyond grouping words that look similar. It clusters by meaning, which frequently disagrees with string similarity.

How it works

• Vectorisation — phrases become vectors using TF-IDF over both word and character n-grams, with optional embeddings when you have a model available.
• Agglomerative clustering — clusters form from the data, with no fixed k. You do not have to guess how many topics exist before the tool can tell you.
• Intent classification — a rules-and-signal model labels each cluster informational, commercial, transactional or navigational.
• Hub and spoke maps — clusters become a pillar page with supporting pages, and topical authority coverage is scored.

What it does for your site

Cannibalisation is the quiet killer. Two pages targeting "crm for small business" and "small business crm software" split their own signals, and neither ranks as well as one strong page would. Clustering finds those collisions before you write the second page.

It also gives you a defensible content plan. Instead of "write more about CRM", you get "these fourteen phrases are one intent, the current page covers nine of them, and here is the hub-and-spoke structure that covers the rest".

Where the results go

Clusters become topical maps, and each cluster is what the forecasting model prices out in traffic and revenue.