A Claude Code skill for persona research. Generates synthetic personas, conducts user interview simulation with structured questions, and produces a data-driven Markdown report with actionable recommendations.
Built for Australian market research contexts — ABS demographics, ACCC-compliant language, ESOMAR-inspired methodology.
Synthetic personas are fictional characters created by AI to represent real customer segments. Instead of recruiting 10 actual people for interviews (which takes weeks and costs thousands), this skill generates 10 realistic profiles — with jobs, budgets, pain points, and decision-making patterns — and interviews them on your behalf.
Think of it like a flight simulator for customer research. A flight simulator doesn't replace real flying, but it lets pilots test scenarios safely and quickly. Synthetic personas do the same for product decisions: they let you explore how different customer types might react to your pricing, features, or messaging — before you spend money building or selling.
Synthetic personas are useful, but they're not a replacement for real customer research. Here's what to trust and what to validate:
| Trust | Validate |
|---|---|
| Directional signals (e.g., "price-sensitive segment exists") | Exact numbers (e.g., "73% prefer $49/mo") |
| Pattern detection across personas | Individual persona opinions |
| Feature prioritisation and ranking | Final pricing or positioning decisions |
| Hypothesis generation | Hypothesis confirmation |
Bottom line: Use synthetic personas to narrow your options and sharpen your questions — then validate the final decision with 5-10 real customers. The skill produces insights, not proof. Treat it as the start of your research, not the end.
Two co-founders disagree on which customer segment to target first. Generate 8 personas across both segments, interview them on the same value propositions, and let the data decide — not the loudest voice in the room.
Impact: Turns subjective founder debates into objective scorecards. Prevents the #1 startup killer: building for the wrong audience because co-founders couldn't agree.
You're about to spend three weeks building a pricing page with four tiers. Before that, generate personas matching your buyer profiles, present each tier configuration, and find the "golden price anchor" — the tier structure that maximises perceived value across segments.
Impact: Saves 3 weeks of dev time if the data says your 4-tier model confuses people. You launch with the right structure from day one.
Map your top 3 competitors as personas with their real traits (budget, priorities, objections). Then generate your ideal customer personas and interview them on switching triggers, deal-breakers, and the "golden argument" that wins them over.
Impact: You walk into sales calls knowing exactly which objection to address first and which competitor weakness to exploit — without waiting for real lost deals to teach you.
You're planning a 6-month content calendar but writing for everyone means writing for no one. Generate personas by role, seniority, and content preference. Interview them on what they'd actually click, share, and save.
Impact: Each blog post, email, and LinkedIn thread targets a specific persona's trigger. Content becomes a conversion tool instead of a vanity metric generator.
Your board wants "user evidence" before approving Q3 features. Generate 10 personas representing your core segments, run them through your proposed features, and produce a report showing which features score highest and why.
Impact: Turns "we think users want X" into "8/10 personas scored Feature A at 8.5/10 with these specific reasons." Boards approve faster when they see numbers, not hunches.
| Without Persona | With Persona |
|---|---|
| Weeks of recruiting for interviews | Instant synthetic respondents |
| Expensive research tools (Qualtrics, Maze) | Zero additional cost, runs in Claude Code |
| Subjective "I think users want..." | Scored, ranked, quote-backed insights |
| No methodology documentation | ESOMAR-inspired, ACCC-aware reports |
- Speed: 15-minute research cycle vs. 2-4 week traditional recruitment
- Cost: $0 vs. $2k-10k for research panels or tools
- Decision quality: Quantitative scores + qualitative themes reduce assumption-driven mistakes
- Repeatability: Run the same questions across different contexts to compare segments
Requires Claude Code with skills support.
Copy SKILL.md to your Claude Code skills directory:
mkdir -p ~/.claude/skills/persona && cp SKILL.md ~/.claude/skills/persona/SKILL.mdOr clone this repo and link it:
git clone https://github.com/wsamuelw/persona.git ~/.claude/skills/personaIn Claude Code, invoke with:
/persona
Or describe what you need:
- "Create 5 personas for a B2B accounting app targeting small business owners in Sydney"
- "Generate personas and interview them on pricing sensitivity"
- "Run persona research on a meal kit delivery service in Melbourne"
- Generate — Creates N distinct personas grounded in realistic demographics
- Interview — Persona research in action: each persona answers your questions in character (scale 0-10 + open-ended)
- Report — Aggregates into a structured Markdown report with tables and themes
- Summary — Delivers top 3 recommendations and "golden combos" per question
Click to expand — Sample report excerpt (B2B SaaS pricing research, 10 personas)
| Question | Avg (0-10) | Median | Top/Bottom Split |
|---|---|---|---|
| How appealing is the $49/mo Starter tier? | 7.2 | 7 | Solo founders (8.3) vs. Ops managers (5.8) |
| Would you upgrade to the $149/mo Pro tier? | 6.1 | 6 | Agency owners (8.0) vs. Freelancers (4.2) |
| How clear is the pricing page layout? | 5.4 | 5 | Technical buyers (7.1) vs. Non-technical (3.9) |
- Price anchoring matters more than tier count: 7/10 personas compared pricing relative to a "reference point" rather than evaluating each tier independently. → "I didn't pick the best plan — I picked the one that didn't feel like a rip-off compared to the cheap one." (P3, Agency Owner)
- Feature gating creates anxiety: Non-technical buyers feared hitting limits mid-project. → "What happens when I hit 500 contacts? Do I get a warning or does the system just stop?" (P7, Marketing Manager)
- Annual discount needs context: The 20% annual discount didn't resonate without a monthly equivalent. → "Show me the monthly price first, then tell me I save by going annual. Don't hide the monthly behind a toggle." (P2, Freelancer)
- Golden tier structure: 3 tiers with the middle tier highlighted as "Most Popular" — 8/10 personas preferred this over a 4-tier layout
- Golden price anchor: Lead with $49/mo, position $149/mo as "everything you need", make $299/mo feel premium without being absurd
- Golden CTA: "Start free trial — no card required" scored 9.1/10 vs. "Get started" at 5.3/10
Findings based on 10 synthetic personas generated per ABS demographic distributions and provided context. Synthetic data supports ideation and prioritisation; validate with 5-10 real prospects before commercial rollout. Inspired by ESOMAR research principles. Outputs are synthetic — do not use in advertising claims. Refer to Australian Privacy Principles (APPs) for data handling.
MIT