InsightWorks with ChatGPT, Claude, Gemini

Review mining: themes, fixes and copy

Cluster hundreds of reviews into what customers love, hate and say in their own words.

What you need

Export of reviews (CSV or pasted text) with rating and date
Product name(s)
What decision this informs (NPD, CX, creative)

The prompt

You are a consumer insights analyst. Below are [NUMBER] customer reviews for [PRODUCT] from [SOURCES, e.g. our site, Amazon UK, Boots.com].

[PASTE REVIEWS: rating | date | text]

Analyse them and give me:
1. Top 5 things customers love, each with % of reviews mentioning it (approximate), and 2 verbatim quotes.
2. Top 5 complaints or problems, with % and 2 verbatim quotes each. Separate product issues from delivery/service issues.
3. Who seems to buy it (use cases, life stages, gifting) based only on what reviewers say.
4. The exact words and phrases customers use to describe the benefit (for ad and PDP copy).
5. Any trend over time (e.g. complaints rising since a date, which might suggest a batch or formula change).
6. 3 recommended actions each for: product team, CX team, marketing team.

Rules: only use evidence from the reviews; quote verbatim; say "not enough evidence" rather than guessing. Present as a table followed by bullet points.

Replace everything in [BRACKETS]. Use a business or team account with training on your data switched off, and remove customer personal data before pasting.

What you get

A table of loved features and complaints with frequencies and verbatim quotes, customer language for copy, trends over time and actions by team.

Tips

Claude and ChatGPT handle a few hundred reviews per run; for thousands, split by product or quarter and then ask for a combined summary.
Remove customer names and emails before pasting (UK GDPR).

Review mining: themes, fixes and copy

What you need

The prompt

What you get

Tips

Get the brief