Topic Clusters & Revenue
Keyword Clustering That Drives Results: 10 Lessons from Andy Chadwick
Turn huge keyword lists into a small set of pages that rank, avoid cannibalization, and map cleanly to your architecture. These lessons are grounded in Andy Chadwick’s talk for Keyword Insights and real site examples.
1) What clustering solves
Clustering groups similar queries so one well-structured page can answer many variations. The result is fewer pages that rank better and no more self-competition. In most niches you need far fewer pages than keywords, which keeps crawl budget and internal links focused on the workhorses.
2) SERP vs NLP clustering
There are two main paths. SERP-based clustering checks overlap in Google’s top results and groups terms that share URLs. It mirrors what Google already serves, which means it respects intent splits that NLP might miss. NLP or semantic clustering is cheaper and fast for early passes, but it can blend distinct intents into one bucket.
| Approach | Strengths | Risks | When to use |
|---|---|---|---|
| SERP-based | Follows real results, protects intent splits, great for competitive niches | Higher compute cost, needs live SERP checks | Final planning and enterprise migrations |
| NLP/semantic | Fast, inexpensive, good for first pass | May merge different intents | Exploration and early grouping |
Keyword Insights supports SERP-based clustering with adjustable overlap and intent labeling. Try it for strict, intent-aware clusters.
3) Enterprise case study: 50M URLs, heavy cannibalization
In the talk, Andy shares a U.S. real-estate site with about 50 million URLs stuck on page 2 because crawl budget was spread thin and near-duplicate pages competed with each other. By finding roughly 70 percent SERP overlap in categories like “homes for sale in California”, “houses for sale in California”, and similar variants, the team merged categories, redirected about 15 million URLs, and saw traffic rise around 110 percent in 3 to 4 months. Watch the talk for context.
Enterprise playbook
- Export a massive keyword set, run SERP-overlap clustering
- Group near-duplicates at the category level
- Pick canonical versions and redirect the rest
- Fix internal links so equity flows to the canonicals
Risk control
- Phase redirects in batches, monitor Search Console
- Keep a change log and revert only if needed
- Capture old slugs in redirect rules to avoid 404 spikes
4) Small-site consolidation
Smaller sites often carry yearly variants and “near-same” topics. Convert sitemap URLs into their target keywords, cluster them, and merge duplicates like “most-followed Instagram 2022” vs “2023”, or overlapping posts like “best time to post on Instagram”. Keep the evergreen page and redirect the rest to build a durable winner.
5) Franchise duplicates
Decentralized teams often publish the same guide under city or region folders, which causes large-scale cannibalization. Clustering reveals stacks of near-identical posts that should become one master with local appendices, proper canonicals, or location parameters. The search signal concentrates and rankings stabilize.
6) Fast keyword-to-URL mapping with “tokenized” slugs
Tokenize your existing URLs into pseudo-keywords and cluster those alongside your keyword list. Give the tokenized items a special marker and very high volume so they always remain visible inside the clusters. This shows which keywords map to existing URLs and which clusters need new pages. Andy highlighted this in a PPC and SEO mapping workflow for Air Wick.
Copyable checklist
7) “Un-merge” terms that look the same but are not
Google often treats similar phrases differently. Plural vs singular or sibling nouns can live in different intent buckets. Examples from the talk include “vaporizer accessories” vs “vaporizer parts” and “skateboard wheel” vs “skateboard wheels”. If SERPs do not overlap, split the cluster and plan separate pages.
8) Find content gaps at scale
Pull a very large keyword universe for your topic, cluster it, then cross-check rankings to filter clusters where your domain does not show up. This surfaces net-new pages without hours of manual filtering. Prioritize by intent and business value, then brief.
9) Beat broad authorities with focused pages
Export all keywords a single giant page ranks for, cluster them, and split into multiple tightly focused articles. Andy showed an example of a page ranking for hundreds of terms that broke cleanly into a couple dozen clusters. The smaller, focused articles can outrank the broad piece because each one matches intent and depth better.
10) Zero-volume wins from forums and PAA
Scrape threads on Reddit to find recurring questions, cluster those variants, enrich with People Also Ask prompts, then publish concise answers. A skincare brand used this approach to grow quickly because questions with low reported volume still have strong intent and weak competition. Keyword Insights offers helpers like large Search Console exports that unlock far more queries for clustering. Give it a try.
Templates and quick wins
Consolidation plan
Franchise clean-up
Zero-volume sprint
Brief fields per cluster
FAQ
How strict should my SERP overlap be
Start moderate so obvious families group, then tighten for competitive topics. If top results share many URLs, one page usually serves the whole set.
Do I always map one cluster to one page
Most of the time yes. If the SERP shows two clear intents, split into two pages. Re-check after publishing and adjust links and anchors.
How often should I re-cluster
Quarterly is a good cadence. Add new queries, merge overlaps, and refresh briefs. Watch cluster-level rank distribution and traffic to catch decay early.
Where do I start if I have no tools
Begin with sitemap tokenization and a small NLP pass to spot obvious merges. Then move to a SERP-based tool like Keyword Insights for the final plan.
Case study stats and tactics summarized from Andy Chadwick’s talk. Use clusters to plan helpful pages that answer people’s questions and reduce duplicate targeting.
