Re-clustering 1,200 keywords cut our planned pages from 90 to 41

Re-clustering 1,200 keywords cut our planned pages from 90 to 41

A content plan listed 90 pages, one per keyword. We re-clustered by SERP overlap instead of by string similarity.

— Keywords in scope: 1,200
— Pages in original plan: 90
— SERP-overlap threshold used: 4+ shared URLs in top 10
— Pages after clustering: ▇▇▇▇ 41

String-based grouping had treated synonyms as separate intents. SERP overlap showed Google already treats them as one. Forty-nine planned pages would have cannibalized each other on day one.

The 41 pages we built averaged position 8; a control batch built on the string method averaged position 13.

So what: cluster by what the SERP shows, not by how the words look. Google has already done the grouping for you.

Benchmark of the week: keywords sharing 4+ top-10 URLs should be one page 88% of the time in our validation set.