Marketer pack
Claude Skill
Voice-of-Customer Researcher
Mines interviews, reviews, tickets, and communities into personas, jobs-to-be-done, and the exact language customers use.
What it does
Works in two modes: analyzing research you already have (interviews, surveys, support tickets, win/loss notes, NPS) and gathering new signal from where customers actually talk (Reddit, G2, forums, review sites). Extracts jobs-to-be-done, pains, trigger events, objections, and verbatim language so positioning and copy are grounded in evidence, not assumption.
When to use
- ✓Before writing copy or positioning and you need real customer language
- ✓You have transcripts, reviews, or tickets and need to extract the signal
- ✓You need personas or JTBD grounded in evidence, not a brainstorm
When not to use
- ✗You have zero research and won't gather any — it surfaces signal, it can't invent it
- ✗You need to act on findings to fix a page — that's a CRO job
Install
Download the .zip, then unzip into your Claude skills folder.
mkdir -p ~/.claude/skills
unzip ~/Downloads/voice-of-customer-researcher.zip -d ~/.claude/skills/
# Restart Claude Code session.
# Skill is now available — Claude will use it when relevant.SKILL.md
SKILL.md
---
name: voice-of-customer-researcher
description: Use when conducting, analyzing, or synthesizing customer research. Triggers on "customer research", "VOC", "voice of customer", "analyze transcripts", "review mining", "build personas", "jobs to be done", "JTBD", "what do customers say".
---
# Voice-of-Customer Researcher
Get to what customers actually think, say, and struggle with — so positioning, product, and copy are grounded in reality. Two modes, usually combined: extract signal from research you have, and go find signal you don't.
## Mode 1 — Analyze existing assets
For each asset, segment before you conclude — don't average across different customer types or churn reasons.
- **Interview / sales-call transcripts** — pains, triggers, desired outcomes, objections, alternatives considered. Look for the moment they decided to look for a solution.
- **Surveys** — segment by tier/use-case/tenure; watch where open-ended answers contradict multiple-choice.
- **Support tickets** — separate bugs from confusion from missing features from expectation mismatch. Categorize before analyzing.
- **Win/loss & churn notes** — segment by reason; don't blend price, fit, and timing.
- **NPS** — detractors and passives carry more improvement signal than promoters; pair every score with its verbatim.
## Mode 2 — Go find research
Know where the customer talks and what to extract: Reddit threads, G2/Capterra reviews (yours and competitors'), niche forums, communities, review sites. Pull recurring pains, the language they use unprompted, and the alternatives they compare against.
## Extraction framework
For each source, extract:
1. **Jobs to be done** — functional (the task), emotional (how they want to feel), social (how they want to be seen)
2. **Pains** — prioritize the ones mentioned unprompted and with emotional language
3. **Trigger events** — what changed that made them seek a solution
4. **Objections & alternatives** — what almost stopped them; what they compared you to
5. **Verbatim language** — exact phrases, for the copy bank
## Output
Personas grounded in evidence + a verbatim language bank (their words, not yours) + the top JTBD and pains, ranked by frequency and intensity.
## Anti-patterns
- Averaging across segments that should be analyzed separately
- Treating every support ticket as equal signal
- Inventing quotes or personas when the research isn't there — say what's missing instead
- Rewriting customer language into marketing language (the point is to keep theirs)
Example prompts
Once installed, try these prompts in Claude:
- Analyze these 12 sales-call transcripts — extract jobs-to-be-done, top pains, objections, and the exact phrases prospects use.
- Mine G2 and Reddit for how people talk about [category] — build 2 personas with their real language.