Search intent, not search volume: how to group keywords properly

The volume trap

Keyword tools sort by volume because volume is easy to sort by. That single default has probably caused more wasted content budget than any algorithm update. Volume tells you how many people typed something. It says nothing about what they expected to find, and matching that expectation is the whole job.

Two queries with similar volume can require completely different pages. One wants a definition and a diagram. The other wants a price comparison table. Publish the wrong shape and the page will not rank, no matter how well written it is.

Four intents worth separating

Informational

The reader wants to understand something. These pages need clear structure, examples and a definition early. Selling here reads as noise.

Comparison

The reader has narrowed the field and wants help choosing. Tables, trade-offs and honest weaknesses do the work. A comparison page with no negatives is not trusted, by readers or by search engines.

Transactional

The reader is ready to act. Get out of the way: price, availability, what happens next. Long preambles cost conversions.

Navigational

The reader is looking for a specific brand or page. Rarely worth a new article, frequently worth a fix to your site structure.

How to group a keyword list properly

Start with the search results, not with the spreadsheet. For each candidate query, look at what already ranks. If the top five results for two queries are broadly the same pages, those queries belong on one page. If they are entirely different, they need separate pages, however similar the words look.

This one habit prevents the most common structural error in content programmes: two pages built for the same job, splitting links and impressions between them, neither strong enough to win.

• Group by overlapping results, not by string similarity
• Name the job of each cluster in one sentence
• Pick one primary query per cluster and let the rest support it
• Record the decision so nobody rebuilds the cluster next quarter

Spotting cannibalisation before you publish

Cannibalisation is easier to prevent than to fix. Before a new page enters the queue, check whether an existing page already targets that intent. If it does, the honest options are to update the existing page or to change the angle of the new one. Publishing both is the option that feels productive and performs worst.

If you cannot explain in one sentence how a new page differs from an existing one, you are about to compete with yourself.

Measuring the right thing

Sessions are a lagging, noisy signal. Query-level impressions and average position tell you far more about whether a page is doing its job. A page gaining impressions but not clicks usually has a title or intent mismatch, which is a cheap fix. A page losing impressions steadily is decaying and belongs in a refresh queue.

Set a review date at publication time. Ninety days is a reasonable default for competitive topics, six months for stable ones. The point is that the review is scheduled work, not a rescue mission.

Putting it into practice this week

Export your last hundred queries. Group them by intent using the results-overlap test. Mark every cluster that already has more than one page targeting it. Fix those first, before writing anything new. Most teams recover more traffic in that exercise than in the following two months of publishing.