Product Discovery System

"Discovery without delivery = analysis paralysis. Delivery without discovery = feature factory."

This skill covers the Discovery System — continuously discovering which problems matter and which solutions might work. It maintains a living map of customer opportunities and tests solution ideas before committing engineering resources.

Part of: Modern Product Operating Model — a collection of composable product skills.

Related skills: product-strategy, product-architecture, product-delivery, ai-native-product, product-leadership

When to Use This Skill

Use this skill when:

Setting up a weekly discovery rhythm
Building or updating an Opportunity Solution Tree (OST)
Creating interview snapshots after customer conversations
Exploring multiple solution approaches
Designing and running assumption tests
Synthesizing insights across multiple interviews
Running an Opportunity Council meeting

Cadence: Weekly rhythm | Owner: Product Trio (PM + Designer + Tech Lead)

The Problem This Solves

Most teams either:

Do discovery once, then execute for months on stale assumptions
Skip discovery entirely and build what stakeholders request
Do discovery but don't connect it to what actually gets built

The Discovery System creates a weekly rhythm that keeps you close to customers and ensures evidence—not opinions—drives decisions.

Philosophy

Core Beliefs

Weekly rhythm over big research projects — 2-3 interviews per week beats quarterly research sprints
The crossfunctional discovery — Handoffs kill learning
Opportunities are problems, not solutions — "Users need faster onboarding" not "Add a wizard"
Multiple solutions per opportunity — Always explore 3+ options before committing
Test in hours and days, not weeks — If your test takes a month, you're testing too much

What This Framework Rejects

Discovery theater (interviews that don't change roadmap)
Solution-first thinking
PM does interviews alone, hands notes to designers
Building the first idea that comes to mind
Waiting for perfect data before deciding

Framework Components

1. Continuous Discovery Habits

The Weekly Rhythm (Minimum Viable Discovery)

Activity	Frequency	Purpose
Customer interviews	2-3 per week	Stay connected to real problems
Synthesis session	1 per week	Update opportunity map
Assumption test	1 per week	Validate before building

Who Does Discovery: The Product Trio

Role	Contribution
Product Manager	Owns outcome, facilitates, synthesizes
Product Designer	Owns experience, visualizes, prototypes
Tech Lead	Owns feasibility, estimates, identifies constraints

Principle: The trio does discovery together. If the PM does interviews alone and hands notes to designers, you've already lost 50% of the insight.

0→1 Mode:

Founder does interviews personally
10-15 interviews before patterns emerge
Daily cadence if possible
Bias toward speed over rigor

Scaling Mode:

Research ops supports logistics
Systematic interview quotas by segment
Centralized insight repository
Quarterly synthesis reports

2. Interview Snapshots

After each interview, create a snapshot (not a transcript). Capture the essence, not every word.

Snapshot Format:

INTERVIEW SNAPSHOT
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Date: [Date]
Participant: [Role, Company type, Context]
Interviewer(s): [Names]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

KEY OPPORTUNITIES (Unmet needs discovered)
• [Opportunity #1]
• [Opportunity #2]
• [Opportunity #3]

KEY QUOTE (In their words)
"[Memorable statement that captures their experience]"

QUICK FACTS
• [Relevant context about their situation]
• [Current workflow or tools]
• [Constraints or requirements]

JOBS TO BE DONE (If surfaced)
• Functional: [Task they're trying to accomplish]
• Emotional: [How they want to feel]
• Social: [How they want to be perceived]

SURPRISES
• [Anything unexpected]
• [Assumptions challenged]

FOLLOW-UPS
• [Questions for next interview]
• [Things to validate]

AI Integration for Snapshots:

Use AI to draft snapshot from notes
Always review — AI misses 20-40% of important context
Never let AI replace the act of listening

3. Opportunity Solution Tree (OST)

The OST is your living map connecting outcomes to opportunities to solutions to tests.

Structure:

                    OUTCOME
                    (Metric we're trying to move)
                         │
          ┌──────────────┼──────────────┐
          │              │              │
     OPPORTUNITY    OPPORTUNITY    OPPORTUNITY
     (Unmet need)   (Unmet need)   (Unmet need)
          │              │              │
     ┌────┴────┐    ┌────┴────┐    ┌────┴────┐
     │         │    │         │    │         │
  SOLUTION  SOLUTION SOLUTION SOLUTION SOLUTION SOLUTION
  (Idea)    (Idea)  (Idea)   (Idea)  (Idea)   (Idea)
     │         │
  ┌──┴──┐   ┌──┴──┐
  │     │   │     │
 TEST  TEST TEST  TEST

OST Rules:

Rule	Why
One outcome per tree	Don't try to solve everything at once
Opportunities are problems, not solutions	"Users struggle to..." not "Add a feature..."
Multiple solutions per opportunity	Always explore 3+ before committing
Evidence-backed	Each opportunity has interview/data support
Living document	Update weekly as you learn

Good Opportunity Statements:

"Users struggle to understand which metrics matter during their first week"
"Managers can't quickly see which team members need attention"
"New users don't know what to do after signup"

Bad Opportunity Statements (These are solutions):

"We need a dashboard"
"Add an onboarding wizard"
"Send email reminders"

Target Opportunity Selection:

Use compare-and-contrast to select focus:

Opportunity	Pain Severity	Frequency	Strategic Fit	Evidence Strength
A	High	Daily	Core	8 interviews
B	Medium	Weekly	Adjacent	3 interviews
C	High	Monthly	Core	12 interviews

Principle: Choose ONE target opportunity at a time. Complete focus beats scattered effort.

4. Solution Exploration

For every target opportunity, generate at least 3 solution approaches before committing.

The Three Solution Types:

Type	Description	Example
The obvious solution	What everyone expects	"Add an onboarding wizard"
The 10x harder solution	If effort were no constraint	"AI-powered personalized setup"
The non-product solution	Pricing, process, partnership, or service	"White-glove onboarding call"

Solution Categories:

Category	When to Consider
Product changes	Features, UX improvements
Pricing/packaging changes	How value is captured
Enablement changes	Documentation, training, support
Process changes	How work gets done internally
Partnership solutions	Integrate vs. build

Principle: The best solution to a product problem is often not a product change.

Thin-Slice MVP:

Don't build the whole solution. Build the smallest thing that tests your riskiest assumption.

Full Solution	Thin Slice
"Complete onboarding wizard with 10 steps, progress tracking, and personalization"	"Single welcome screen that asks one question and shows one recommendation"
"Full analytics dashboard with customizable widgets"	"One pre-built view showing the top 3 metrics"
"AI-powered recommendation engine"	"Rule-based suggestions for top 5 use cases"

5. Assumption Testing

Every solution has assumptions. Find the ones that would kill it if wrong.

Assumption Categories:

Category	Question
Desirability	Will users want this?
Viability	Will this work for the business?
Feasibility	Can we build this?
Usability	Can users figure it out?

Assumption Test Format:

ASSUMPTION TEST
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Assumption: [What we believe is true]
Risk level: [High / Medium / Low]
Test method: [How we'll test]
Success criteria: [What would confirm]
Failure criteria: [What would disprove]
Timebox: [Hours/days, not weeks]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Test Methods (Fastest to Slowest):

Method	Time	When to Use
Desk research	30 min	Does evidence already exist?
One-question survey	1 hour	Quick signal from existing users
Fake door test	1 day	Measure interest before building
Concierge test	1-3 days	Manually deliver the value
Wizard of Oz	1 week	Fake backend, real frontend
Prototype test	1-2 weeks	Clickable prototype with users
A/B test	2-4 weeks	Live code, statistical significance

Principle: Test in hours and days, not weeks and months. If your test takes a month, you're testing too much at once.

AI Integration for Testing:

Use AI to build prototypes faster (vibe coding)
Use AI to analyze survey responses
Use AI to synthesize test results
Don't use AI to decide — Humans interpret, AI assists

6. Opportunity Council (Scaling Mode)

Weekly cross-functional meeting that turns discovery into decisions. Prevents discovery theater.

Participants:

PM (facilitator)
Design lead
Engineering lead
Sales/CS representative (input, not veto)
Marketing representative (for GTM alignment)

Agenda (60 min max):

Segment	Time	Focus
New evidence review	15 min	2-3 key findings from this week
Opportunity prioritization	20 min	Promote, kill, or park opportunities
Solution shaping	15 min	Review prototype/test results
GTM/tech flags	10 min	Early visibility on constraints

Council Rules:

Decisions are recorded with rationale
Single decider (PM) — council advises, PM decides
No side quests — if it's not on the OST, it waits
Evidence required — no "I think users want..."

0→1 Mode: Skip formal council. Founder + team informal sync.

Primary Outputs

Output	Description	Update Cadence
Opportunity Solution Tree	Living map of outcome → opportunities → solutions	Weekly
Interview snapshots	Library of customer evidence	After each interview
Test results	What we learned, what we decided	After each test
Target opportunity	Current focus area	Weekly review
Solution candidates	Prototypes ready for prioritization	Ongoing

Templates

This skill includes templates in the templates/ directory:

interview-snapshot.md — Post-interview capture format
opportunity-solution-tree.md — OST structure and rules
assumption-test.md — Test design and tracking

Using This Skill with Claude

Ask Claude to:

Set up discovery rhythm: "Help me design a weekly discovery cadence for [team size/stage]"
Create interview guide: "Create an interview guide for understanding [JTBD/opportunity]"
Draft snapshot: "Turn these interview notes into a snapshot"
Build OST: "Help me build an OST for the outcome: [metric]"
Reframe solutions as opportunities: "These are solutions — help me reframe as opportunities"
Generate solution options: "Generate 5 solution approaches for [opportunity]"
Design thin slice: "What's the thin-slice MVP for [solution]?"
Create assumption test: "Design an assumption test for [hypothesis]"
Synthesize interviews: "Find patterns across these [X] interview snapshots"
Prepare Opportunity Council: "Create an agenda for Opportunity Council with these findings"

Connection to Other Skills

When you need to...	Use skill
Define ICP and strategic context	`product-strategy`
Convert opportunities to bets	`product-architecture`
Plan delivery and measurement	`product-delivery`
Discover for AI products	`ai-native-product`
Scale discovery across teams	`product-leadership`

Quick Reference: Discovery Quality Checklist

Before concluding discovery on an opportunity:

Evidence breadth — 5+ interviews mentioning this opportunity
Evidence depth — Understand functional, emotional, social jobs
Multiple solutions explored — 3+ approaches considered
Thin slice identified — Know the smallest testable version
Riskiest assumption named — Know what would kill this
Test designed — Ready to validate before building
Strategic fit confirmed — Aligns with strategy canvas

Anti-Patterns to Avoid

Anti-Pattern	Why It Fails	Instead
Discovery theater	Interviews don't change roadmap	Evidence → decisions
AI replaces listening	Miss 40% of insight	AI augments, humans decide
Solution-first thinking	Build wrong thing	Opportunity-first
Homer Simpson car	Feature bloat from asking "what do you want?"	Ask about problems, not solutions
PM does discovery alone	Handoffs kill learning	Trio does discovery together
One solution per opportunity	Miss better approaches	Always 3+ options
Month-long tests	Too slow to learn	Test in hours/days

Sources & Influences

Teresa Torres — Continuous Discovery Habits, Opportunity Solution Trees
Marty Cagan — INSPIRED, EMPOWERED
Rob Fitzpatrick — The Mom Test
Cindy Alvarez — Lean Customer Development

Part of the Modern Product Operating Model by Yannick Maurice

product-discovery