---
title: Mining voice-of-customer data and assembling copy from it
slug: voice-of-customer-copy-mining
summary: Joanna Wiebe's method — mine reviews/Reddit for customers' exact words, split findings into a "message" bucket (repeated themes) and a "copy" bucket (weird standout lines), then assemble copy into a framework instead of writing from scratch.
domain: marketing
tags: [copywriting, voice-of-customer, research, headlines]
related: [headline-hooks-and-copy-frameworks, funnel-stage-copy-triggers, ai-landing-pages-with-claude-code, sales-page-audit-checklist, customer-research-for-ads, customer-discovery-calls-at-scale]
sources: [yt:32uLdgfvuQ4, yt:rYn2-fOtdp0, yt:2HDkD7jSbzE]
updated: 2026-10-24
---

# Mining voice-of-customer (VOC) data

## TL;DR
- Customers' own words beat yours. "If you think you need rehab, you do" was lifted straight from an Amazon review of a 12-step book[^1].
- Sources: Amazon reviews (including competitors' and adjacent books), Reddit threads, X, Instagram comments[^2].
- **Two buckets:** *message* = repeated themes → your offer angle; *copy* = the weirdest, most unexpected lines nobody in a boardroom would write → your headlines[^3].
- Don't pick the average line everyone agrees with — hunt for the standout one[^3].
- **Copy is assembled, not written** (Eugene Schwartz): put VOC in a "message matrix" spreadsheet, choose a framework, paste VOC into it, then polish with formulas[^4][^5].

## Details
**Mining.** Joanna Wiebe ([[people/joanna-wiebe]]) found SweatBlock's winning homepage hook ("It doesn't even have to be hot out — my armpits are always wet") in Amazon reviews; leading with that problem produced a 108% lift in revenue[^6]. For a fashion app, the customer phrase "big bum, thick waist, not-so-perky boobs" became the headline that helped it grow[^3].

**The two-bucket sort**
| Bucket | What goes in | Becomes |
|---|---|---|
| A — Message | Themes repeated across many customers | Offer angle, page structure |
| B — Copy | Odd, vivid, specific single lines | Headline, hooks, crossheads |

"Message is the what, copy is the how"[^3].

**Assembly workflow (Joanna's setup).** Two monitors: Google Doc on one, message-matrix spreadsheet of organised VOC on the other. Find the hook in the VOC → choose the framework that fits the hook ([[headline-hooks-and-copy-frameworks]]) → write the framework's sections on the page → paste VOC into each section → polish with formulas (e.g. "even if")[^4][^5]. If you swipe from VOC, specificity comes for free ([[proof-and-believable-claims]])[^7].

Conversion copywriting's core rules, as Joanna defines them: drive the message with data (including qualitative VOC), never start from a blank page (start from a framework), and make every piece move the reader to a specific yes[^8].

**AI caveat.** Will, the AI-workflow guest on Dara Denney's channel ([[people/dara-denney]]), says AI is worst at copywriting; when hiring creative strategists he wants people who go read the subreddit AI surfaces to learn how people actually talk, rather than accept AI-written copy[^9].

## Disagreements / caveats
- Picking single "random" review lines draws objections ("you can't just cherry-pick"); Joanna's answer is that this is the job — but validate the line with a test ([[headline-hooks-and-copy-frameworks]]).
- Verbatim health or body claims from reviews can break ad-platform or regulatory rules; check before using.

## See also
- [[funnel-stage-copy-triggers]]
- [[sales-page-audit-checklist]]

[^1]: yt:32uLdgfvuQ4 @ 03:07 — Joanna Wiebe
[^2]: yt:32uLdgfvuQ4 @ 03:58 — Joanna Wiebe
[^3]: yt:32uLdgfvuQ4 @ 04:49 — Joanna Wiebe
[^4]: yt:32uLdgfvuQ4 @ 17:16 — Joanna Wiebe
[^5]: yt:32uLdgfvuQ4 @ 18:47 — Joanna Wiebe
[^6]: yt:32uLdgfvuQ4 @ 01:13 — Joanna Wiebe
[^7]: yt:32uLdgfvuQ4 @ 23:01 — Joanna Wiebe
[^8]: yt:rYn2-fOtdp0 @ 01:32 — Joanna Wiebe
[^9]: yt:2HDkD7jSbzE @ 21:52 — Will (guest on Dara Denney)