The customer-voice cluster.
Reviews are the cheapest market research that exists. Most teams read them like complaints, not signals.
The 4-star tab.
A founder showed me his Yelp tab. He'd been opening it for two years. Read every review. Replied to almost all of them. Tracked the bad ones. Felt anxious every Monday.
"What's the category telling you?" I asked.
He paused. "That a few people are jerks?"
He had 380 reviews and no map.
What a customer-voice cluster does.
One review is anecdote. 200 reviews clustered by theme is a category map.
The cluster surfaces three things you can't get from any other signal source for free:
All three are public. All three are free. Most teams skip them because they read each review as a complaint instead of a category signal.
Reviews are the only market research where buyers wrote it for you without being asked.
How to build a customer-voice cluster.
Two hours of focused work. Three competitors minimum + your own.
- Pull 100-300 reviews from your own listings + 3 competitors. Yelp, Google, G2, Trustpilot, App Store, Amazon. Whichever your category lives on.
- Don't read for sentiment. Read for pattern. Tag each review with 1-3 theme labels (e.g. pricing-clarity, setup-friction, support-response, quality-vs-expectation).
- Count by theme. Sort themes by frequency. The top 3 themes are the category's shared concerns.
- Pull verbatim quotes for each top theme. These are your sales-page words.
- Note objections. The negative reviews on competitors' listings are gold. They tell you what kills the consult.
- Cross-reference with searches. Top themes should also show up in "near me" + question searches. If they do, the cluster is real.
When NOT to trust the cluster.
If you have under 50 reviews total. The pattern needs volume. Under 50, you're reading noise. Pull from competitors to make up the gap.
Back to the 4-star tab.
We built his cluster in two hours. 380 reviews became 7 themes. The top three: setup-time anxiety, pricing surprise after consult, and provider-rotation worry. Three things he had never explicitly addressed on his site.
He rewrote the consult page around those three. Booking rate lifted 28% in the first month. The reviews weren't complaints. They were instructions. He'd been reading them as criticism for two years.
[TODO B · Mechanism/why]
[TODO: Explain WHY the thing in A happens. Cite mechanism, data, evidence.]
[Short italic pull-quote that crystalizes the mechanism]
[TODO C · Application/the move]
[TODO: What to do with the insight. Concrete steps.]
- [Step 1] description
- [Step 2] description
- [Step 3] description
[TODO: When NOT to do this / counter-case]
[TODO: One paragraph showing edge case or when the move is wrong.]
[TODO A' · Callback to scene]
[TODO: Return to the opening scene with new meaning. 2-3 sentences. Don't over-resolve.]
What is a customer-voice cluster?
A grouped reading of public buyer language (reviews, comments, transcripts, DMs) that surfaces the category's shared concerns, expectations, and proof requirements. Pattern-level instead of anecdote-level.
How many reviews do I need?
Minimum 50 across your own + competitors. Ideal 100-300. Under 50, you read noise. Pull from competitors to fill the volume gap.
Where do I pull reviews from?
Wherever your category lives. Yelp + Google for local. G2 + Trustpilot for SaaS. App Store for mobile. Amazon for product. Pull from your own listings + 3 competitors at minimum.
How do I tag review themes?
Read for pattern, not sentiment. Tag each review with 1-3 theme labels (pricing-clarity, setup-friction, support-response, etc.). Count by theme. Top 3 = category's shared concerns.
What do I do with the cluster once I have it?
Three moves: rewrite your sales page using the verbatim words from top themes; preempt the top 3 objections directly on your consult page; surface proof in the format the cluster asks for (guarantees, before/afters, screenshots, named clients).
Last updated May 31, 2026. Field notes by Alex Lamb, LoopWorker.