A product decision problem, solved before you build.
Teams argue from vibes: loudest founder wins, the roadmap fills up, and three months later nobody can say why that feature shipped. I help product teams pressure-test pricing, ICP, and prioritisation with a synthetic buyer panel — scores, quotes, and a ranked call — before they spend the quarter.
A wrong product bet is not a bad meeting. It is weeks of engineering, a weak launch, and a quiet churn story. Discovery is slow. Aligning co-founders is slower. The cost of deciding badly is the budget you never get back.
A team has three pricing options, two target segments, and one prioritisation meeting next week.
Instead of another opinion fight, they hire a panel (5–10 distinct buyers), interview them in character, and leave with:
- Scores — who wants what, and who doesn’t
- Quotes — lines they can paste into the PRD or the board deck
- A ranked call — including the option that won and who objected
Outcome on a typical run:
- Same panel reusable for follow-ups (“what would make you switch?”)
- A decision the room can defend — not a consensus shrug
- A clear next step: talk to 5–10 real people with sharper questions
The commercial idea: spend discovery after you know what to validate — not before.
| You are stuck on | I help you leave with |
|---|---|
| Pricing page / packaging | Tier shape, lead price, who you’d lose |
| Feature prioritisation | Ranked bets + segment split |
| Co-founder ICP fight | One scoreboard instead of two opinions |
| Positioning & messaging | Which claim lands, which line bounces |
| Stakeholder pushback | Quote bank and “who disagrees” before the room |
Typical engagement
- Decision workshop — one high-stakes call, facilitated with the panel method
- Sprint support — we run several bets in a week (pricing, page, roadmap)
- Team enablement — your PMs learn to run this in Claude Code themselves
Talk to me about a decision → · datafying
- I sell decision evidence, not “AI demos”
- Panels are designed to disagree (uniform 9/10 cheerleading is a fail)
- Outputs are hypothesis-grade — we say so in the report, so you can use them in a room
- You can keep the skill and run it without me after enablement
- Decision memo — ranked takeaways with counts (
6/10) or themes - Scoreboard — avg / median / top–bottom split
- Quote bank — verbatim, with persona ids
- Golden combo — what won across the panel + who objected
- Honesty footnote — synthetic disclaimer for real meetings
Pricing — meal kit at $49/week
10 buyers. Busy parents lean yes; students don’t — avg 6.4.
“I’d pay for three dinners, not seven.” — P2, PM
Decision: Pilot a 3-night plan.
Prioritisation — ship SSO in Q3?
7.8/10 · IT lead 9 vs solo founder 5.
“I can’t sell to security without it.” — P5, IT lead
Decision: Enterprise-first SSO; don’t block self-serve.
Four phases: GENERATE → INTERVIEW → REPORT → SUMMARY.
- You bring the decision brief, panel size
n, and questions - Personas differ on budget power, time pressure, temperament, risk
- Scales 0–10 + open answers in character (refusal and
unsurecount) - Report: summary → scoreboard → themes + quotes → ranked takeaways → golden combos → footnote
Optional scripts/check_report.py enforces report shape and blocks overclaims.
- Hypothesis generation, not market research — validate with real people before real money
- Not for published market-size stats, legal proof, or replacing live interviews
- Best used to sharpen questions and rank options — then talk to humans
mkdir -p ~/.claude/skills/persona/scripts
curl -fsSL https://raw.githubusercontent.com/47096/persona/main/SKILL.md \
-o ~/.claude/skills/persona/SKILL.md
curl -fsSL https://raw.githubusercontent.com/47096/persona/main/scripts/check_report.py \
-o ~/.claude/skills/persona/scripts/check_report.pyRequires Claude Code. Put only SKILL.md + scripts/ in the skill folder — do not git clone this repo into ~/.claude/skills/persona.
Then: /persona, or describe the decision you’re making.
rm -rf ~/.claude/skills/persona # removePattern: hire a panel → interview in character → scores, quotes, and a decision you can defend.
- Pricing page before code — tier structure and who you’d quietly lose
- Co-founder ICP fight — one scoreboard, less politics
- Competitive win themes — objections to answer first on calls
- Message & content test — what they’d click, skip, or mock
- Roadmap cut — three bets, not twelve
- Landing page pressure test — ten seconds, first objection, CTA
- Hiring & job brand — what scares candidates off
Also useful: workshops, fictional casts, debate panels — same loop, lighter tone.
Invent your own: any decision where ten named judges + a written summary would beat a hallway opinion.
If a pricing, ICP, or roadmap call is coming up and the room is already arguing — that is exactly the engagement I run.
Book a decision workshop → · Product decisions with evidence · datafying