Painpoint research

Painpoint Research

Surface direct quotes from places your customers already talk, organized so you can use them in copy and validation.

When to trigger this skill

The user wants any of:

• "Find me painpoints for [ICP]"
• "Research what [audience] is complaining about"
• "Pull customer language from social media for [segment]"
• "What are [persona] saying about [problem]?"

Do NOT trigger this skill when:

• The user wants generic web research (use a web search directly)
• The ICP is vague ("small business owners" — too broad). Ask for sharpness first.
• The user wants competitive intel about specific vendors (use a different workflow)

Inputs you need

Required:

• A sharply-defined ICP. Examples: "Family caregivers in the sandwich generation taking care of elderly parents." "Solo founders building their first SaaS without a technical background." "YouTube creators with under 10K subscribers trying to monetize."

If the user gives a vague ICP, ask them to narrow it before researching. Bad: "marketing managers." Good: "growth marketing managers at 20-50 person B2B SaaS startups who report to the CEO and have no agency budget."

Optional but useful:

• The problem space you're focused on (saves time vs. open-ended research)
• Any platforms to exclude (e.g. "skip Reddit, my customers aren't there")

Workflow

1. Sharpen the ICP

If the ICP is too broad, ask one question to sharpen it. Then proceed.

2. Run the research

Search across the platforms where the ICP actually talks. Default platforms:

• Reddit — search relevant subreddits for the ICP. Pull from the comments, not just top-level posts.
• Twitter / X — search the exact phrases the ICP would use to describe their pain.
• Forums and niche communities — Indie Hackers, Hacker News, Stack Exchange, professional subreddits.
• Facebook Groups (rarely searchable directly, often skipped).
• LinkedIn — pull from comments on posts about the problem space, not from profile bios.
• Product review sites — G2, Capterra, TrustRadius for the "what I hate about my current tool" reviews.

For each painpoint you find:

• Capture verbatim. Don't paraphrase. The customer's exact language is what's valuable.
• Tag emotional intensity. Mild / moderate / strong, based on emotion words and intensifiers.
• Note source URL. Always. No quote goes in without a source.
• Group by theme. Cluster related painpoints. Five to seven themes is usually right.

3. Generate the research artifact

Use template/research-output.md. Fill every section.

Save to ~/painpoint-research/<icp-slug>/painpoints.md.

4. QA check

Before reporting done, verify:

• At least 15 direct quotes total
• Every quote has a source URL
• Themes are clustered (not 15 standalone quotes with no grouping)
• At least one quote with strong emotional intensity per major theme
• No fabricated quotes. If you can't find something, say so.

5. Present

Tell the user:

• The themes you found, ranked by signal strength
• Two or three quotes worth using directly in copy or outreach
• Any themes that surprised you, with what they might mean
• The platforms that yielded the most signal (helps with channel planning)

What to look for, not just collect

• Repeated language. When five different people describe the same pain in similar words, that's the copy you want on your landing page.
• Emotional intensifiers. "I literally," "I hate," "every single time," "I can't even." These mark strong painpoints worth leading with.
• What they've tried. Workarounds are gold. They tell you which adjacent solutions to compare against and where the solution space already failed them.
• What they're afraid of. Painpoints + fear of trying again is a buying signal.

What to flag as caveats

• A quote from a single thread isn't a pattern. Need at least 3 independent sources.
• Highly upvoted posts can be selection bias. The loudest pain isn't always the most common.
• Some ICPs don't post about their pain publicly (regulated industries, senior executives). Don't force a research artifact when the ICP isn't online about it. Recommend interviews instead.