Run Product Discovery with Opportunity Solution Trees

cat > CLAUDE.md << 'EOF'
# CLAUDE.md

## Role
Product Manager running continuous discovery.
Framework: Opportunity Solution Trees (Teresa Torres).
Reference: https://www.producttalk.org/opportunity-solution-trees/

## Product Context
- B2B SaaS project management tool
- 2,400 active teams, 85% retention
- Target outcome: Increase weekly active usage from 3.2 to 4.5 days/week

## Discovery State
- Conducting weekly customer interviews
- 12 interviews completed this cycle
- Focus: why teams stop using the tool mid-week

## Output Preferences
- Structure everything as Outcome → Opportunities → Solutions → Tests
- Use markdown trees for visual hierarchy
- Be specific — no vague opportunity statements
EOF

mkdir -p .claude/skills
cat > .claude/skills/discovery-ost.md << 'EOF'
# Product Discovery — OST Skill

Process customer interview data using Teresa Torres'
Opportunity Solution Tree framework.

## Framework Reference
- Source: https://www.producttalk.org/opportunity-solution-trees/
- Book: "Continuous Discovery Habits" by Teresa Torres

## Key Definitions
- **Outcome**: The measurable business/product result we're driving toward
- **Opportunity**: A customer need, pain point, or desire (NOT a solution)
- **Solution**: An idea that addresses one or more opportunities
- **Assumption Test**: A small experiment to validate a risky assumption

## Process
1. Extract raw quotes and observations from interview notes
2. Identify opportunities (needs, pains, desires) — never solutions
3. Group related opportunities into themes
4. For each opportunity, brainstorm 3+ solutions
5. For each promising solution, identify riskiest assumption
6. Design a small experiment to test that assumption

## Opportunity Rules (Critical)
- Opportunities are CUSTOMER problems, not OUR ideas
- "Users want a dashboard" is a SOLUTION, not an opportunity
- "Users can't see their team's progress at a glance" is an opportunity
- Use the customer's language when possible
- Every opportunity must connect to the target outcome

## Output: Markdown Tree
```
🎯 OUTCOME: [target metric]
├── 🔍 Opportunity: [customer need]
│   ├── 💡 Solution: [idea 1]
│   │   └── 🧪 Test: [experiment]
│   ├── 💡 Solution: [idea 2]
│   │   └── 🧪 Test: [experiment]
│   └── 💡 Solution: [idea 3]
├── 🔍 Opportunity: [customer need]
│   ├── 💡 Solution: [idea 1]
│   └── 💡 Solution: [idea 2]
```
EOF

claude "Extract opportunities from these interview notes:

Interview: Sarah, PM at a 50-person startup (Mar 25)
---
'We use the tool Monday and Tuesday for sprint planning, but
by Wednesday everyone's back in Slack. The notifications are
overwhelming so people mute them. Then they miss actual updates.
I wish there was a way to see just the things that changed since
I last looked. Also the mobile app is basically useless — I
can't approve anything on my phone so I have to wait until I'm
at my desk. My team in Berlin is 6 hours ahead so by the time
I approve their PRs it's already their end of day.'

Interview: Jake, Eng Lead at a 200-person company (Mar 27)
---
'The weekly view is fine but I need a daily standup view. I
check the tool first thing in the morning but there's no way to
see just what happened overnight. I end up asking people in
Slack anyway. Also the search is terrible — I can never find
old decisions. We started using a separate wiki just for that.'"

claude "Build the full OST. For each opportunity, brainstorm 3 solutions and identify the riskiest assumption for the most promising one."

claude "Design assumption tests for these two solutions:
1. Smart notification digest (batched, prioritized)
2. Mobile approval flow (1-tap approve/reject)

For each, give me:
- The riskiest assumption
- A test I can run this week
- What 'success' looks like (with a number)
- Estimated effort"

# After your next round of interviews:
claude "Update the OST in discovery/ost-2026-03-29.md with these new findings:

Interview: Maria, Designer at a 30-person agency (Apr 1)
---
'I actually love the notifications but they all look the same.
I can't tell if something is urgent or just a comment. Color
coding or priority levels would help. Also, I realized I
only open the app when someone @mentions me — otherwise I
forget it exists.'"