mattgierhart/prd-v05-risk-discovery-interview

Risk Discovery Interview

Position in workflow: v0.4 Screen Flow Definition → v0.5 Risk Discovery Interview → v0.5 Technical Stack Selection

This is an interactive interview skill. The AI asks questions, the user reflects and decides. The goal is to surface risks so the user can mitigate or accept them—not to kill ideas.

Consumes

This skill requires prior work from v0.1-v0.4:

• CFD-* all customer feedback entries (from v0.1-v0.2) — User research foundation; reveals confidence tier in market assumptions
• BR-* all business rules (from v0.2-v0.3) — Constraints on what can change (pricing, moat, product type constrain risk responses)
• FEA-* feature entries (from v0.3) — Feature complexity and priorities signal technical risks
• PER-* persona entries (from v0.4) — Persona distribution and behaviors reveal adoption risks and churn signals
• UJ-* journey entries (from v0.4) — Journey complexity signals friction points; long journeys increase adoption risk
• SCR-* screen entries (from v0.4) — Screen count and design complexity informs technical resource risk

This skill assumes v0.1-v0.4 work is complete and serves as context for interview discovery.

Produces

This skill creates/updates:

• RISK-* entries (risk discovery, owner-assigned severity) — Identified risks with Impact/Likelihood scoring (raw score from 1-9), response type (Mitigate/Accept/Avoid/Transfer), specific mitigations, and owners
• README Risk Scorecard — Baseline risk profile aggregated by category (Market/User/Technical) with total scores and risk level assessment
• Risk mitigation summary — Top 3-5 risks requiring active mitigation before v0.6 architecture work

All RISK- entries are created through user decision during the interview; they reflect explicit owner choices on severity, not AI assumptions:

Example RISK- entry (user-scored):

RISK-001: Market — Competitor Feature Parity
Description: Competitor X launches report scheduling feature (our FEA-003 planned) within 60 days
Trigger: Competitor announces roadmap; sees our landing page
Impact: High (3) — User severity assessment based on competitive urgency
Likelihood: Medium (2) — User assessment of competitor execution speed
Raw Score: 6 (3 × 2)
Status: open
Effective Score: 6.0

Early Signal: Competitor job postings for feature area, beta announcement
Response: Mitigate
Mitigation: Accelerate FEA-003 launch by 30 days; add scheduling as P0 (links to FEA-003, KPI-002)
Owner: Product Lead
Linked IDs: FEA-003 (report scheduling), KPI-002 (activation rate), BR-042 (undercut positioning)
Review Date: Weekly during v0.6 (architecture phase)
Added: v0.5

Note: Confidence scores are NOT part of RISK- entries. Risks are binary facts (discovered or not); severity is user-decided. Confidence applies to CFD-/FEA-/KPI- entries, not risks.

Design Principles

1. Interview, not inquisition — Facilitate discovery, don't interrogate
2. Inform, not kill — Surface risks so user can mitigate, not abandon
3. User owns decisions — AI facilitates, user assigns severity and response
4. Actionable outputs — Every risk has a mitigation path or explicit "accept"

Risk Categories

Discovery Categories (Interview Prompts)

| Category | Focus Area | Example Questions | |----------|------------|-------------------| | Market | Competitors, timing, demand | "What if [competitor] launches this feature next month?" | | Technical | Complexity, unknowns, dependencies | "Which feature has the most technical uncertainty?" | | Adoption | User behavior, activation, retention | "What's the biggest friction point in onboarding?" | | Resource | Team, budget, time | "If you had to cut scope by 50%, what stays?" | | Dependency | External factors, integrations, partners | "What external factor could block launch?" | | Timing | Deadlines, market windows, seasonality | "Is there a deadline we must hit? Why?" |

Scoring Categories (README Scorecard)

Each RISK- entry maps to one of 3 scoring categories for the README Risk Scorecard:

| Scoring Category | Discovery Categories | Measures | |---|---|---| | Market | Market, Timing | Will anyone buy this? | | User | Adoption, Dependency | Will users succeed with this? | | Technical | Technical, Resource | Can we build and run this? |

Interview Flow

Phase 1: Context Review

Before asking questions, AI reviews:

• CFD- evidence from v0.1-v0.2
• FEA- features and their priorities
• UJ- journeys and their complexity
• BR- business rules and constraints

Phase 2: Guided Questions

Ask questions from each category, adapting based on product context:

Market Risks:

• "What happens if [competitor] launches something similar in 60 days?"
• "What market assumption are you least confident about?"
• "What would cause users to choose a competitor instead?"

Technical Risks:

• "Which feature has the most technical uncertainty?"
• "What technology choice are you least confident about?"
• "Is there anything you've never built before?"

Adoption Risks:

• "What's the biggest friction point in [UJ-001 onboarding journey]?"
• "What behavior change are you asking users to make?"
• "What would cause a user to churn in the first week?"

Resource Risks:

• "If you had only 2 developers, what would you cut?"
• "What skill does the team lack?"
• "What's your runway for validation?"

Dependency Risks:

• "What external API or service could break your product?"
• "What partner relationship is critical?"
• "What regulatory requirement could block launch?"

Timing Risks:

• "Is there a hard deadline? What happens if you miss it?"
• "Is there a market window closing?"
• "What seasonal factor affects launch timing?"

Phase 3: Risk Documentation

For each identified risk, create RISK- entry with user input on severity and response.

Phase 4: Priority & Review

• Force-rank risks by Impact × Likelihood
• Identify top 3-5 that require active mitigation
• Document "accept" decisions explicitly

Interview Techniques

| Technique | How to Use | When to Use | |-----------|------------|-------------| | Pre-mortem | "It's 6 months from now and the product failed. Why?" | Opening question | | Constraint forcing | "If you only had [X], what would you cut?" | Resource discovery | | Dependency mapping | "What external factor could block launch?" | Dependency discovery | | Assumption surfacing | "What must be true for this to work?" | Any category | | Devil's advocate | "Let me argue the opposite—what if [X]?" | Challenge weak evidence |

RISK- Output Template

RISK-XXX: [Risk Title]
Scoring Category: [Market | User | Technical]
Discovery Category: [Market | Technical | Adoption | Resource | Dependency | Timing]
Description: [What could go wrong]
Trigger: [What would cause this to happen]
Impact: [High | Medium | Low] — User assessed
Likelihood: [High | Medium | Low] — User assessed
Raw Score: [Impact × Likelihood, 1-9]
Status: [open | mitigating | mitigated | resolved | accepted]
Effective Score: [Raw Score × Status Weight]

Early Signal: [How we'd know this is happening]
Response: [Mitigate | Accept | Avoid | Transfer]
Mitigation: [Specific action if Response = Mitigate]
Owner: [Who is responsible for monitoring]

Linked IDs: [FEA-XXX, UJ-XXX, BR-XXX affected]
Review Date: [When to reassess this risk]
Added: [PRD stage when discovered, e.g., v0.5]

Status weights: open=1.0, accepted=1.0, mitigating=0.5, mitigated=0.25, resolved=0.0

Example RISK- entry:

RISK-001: Primary API Dependency (Stripe) Outage
Scoring Category: Technical
Discovery Category: Dependency
Description: Stripe API outage would block all payment processing
Trigger: Stripe infrastructure failure or rate limiting
Impact: High (3) — All revenue blocked during outage
Likelihood: Low (1) — Stripe has 99.99% uptime SLA
Raw Score: 3
Status: mitigating
Effective Score: 1.5

Early Signal: Stripe status page, payment failure rate spike
Response: Mitigate
Mitigation:
  - Implement graceful degradation (queue payments for retry)
  - Add status page monitoring alert
  - Document manual billing fallback process
Owner: Tech Lead

Linked IDs: FEA-020 (payments), UJ-005 (checkout), BR-030 (pricing)
Review Date: Before launch, quarterly thereafter
Added: v0.5

Risk Response Types

| Response | When to Use | Example | |----------|-------------|---------| | Mitigate | Can reduce impact or likelihood | Add fallback provider, implement retry logic | | Accept | Low impact or unavoidable | "Competitor might copy us—we accept" | | Avoid | Change plan to eliminate risk | Remove feature with high technical uncertainty | | Transfer | Someone else owns the risk | Use managed service instead of self-hosting |

Severity Matrix

| | Low Impact | Medium Impact | High Impact | |---|---|---|---| | High Likelihood | Monitor | Mitigate | Mitigate urgently | | Medium Likelihood | Accept | Monitor/Mitigate | Mitigate | | Low Likelihood | Accept | Accept/Monitor | Monitor |

Anti-Patterns to Avoid

| Anti-Pattern | Signal | Fix | |--------------|--------|-----| | Risk theater | 50+ risks documented | Focus on top 10 that matter | | All high severity | Everything is critical | Force rank; max 3-5 "High" | | No owner | Risks without accountability | Every RISK- needs an owner | | Mitigation = "be careful" | Vague responses | Require specific, testable actions | | Interview becomes lecture | AI talks more than user | Ask, listen, summarize | | Killing ideas | Every risk leads to "don't do it" | Frame as "how to succeed despite" |

Phase 5: Score & Scorecard

After documenting all risks:

1. Assign scoring categories: Map each RISK- to Market, User, or Technical
2. Calculate scores: Raw Score = Impact × Likelihood; Effective = Raw × Status Weight (all start as open)
3. Sum by category: Add effective scores within each scoring category
4. Determine risk level: Total score → Low (0-12), Moderate (13-25), Elevated (26-40), High (41+)
5. Update README scorecard: Fill in the Risk Scorecard table in README.md

Quality Gates

Before proceeding to Technical Stack Selection:

• [ ] All 6 risk categories explored
• [ ] Maximum 10-15 RISK- entries (focused, not exhaustive)
• [ ] Force-ranked by priority (Impact × Likelihood)
• [ ] Top 5 risks have specific mitigation plans
• [ ] "Accept" decisions are explicit, not accidental
• [ ] Every RISK- has an owner
• [ ] Every RISK- has a scoring category (Market/User/Technical)
• [ ] README Risk Scorecard updated with baseline scores

Continuous Risk Management

v0.5 establishes the baseline risk register, but risk discovery does not end here. New RISK- entries can be added at any stage:

| Stage | Typical New Risks | Score Impact | |-------|-------------------|--------------| | v0.6 Architecture | Infrastructure complexity, integration unknowns | Technical score rises | | v0.7 Build | Implementation blockers, test coverage gaps | Technical score rises | | v0.8 Deployment | Operational risks, security findings | Technical score rises | | v0.9 GTM | Market timing shifts, competitive moves | Market score rises | | v1.0 Growth | Real adoption data contradicting assumptions | User score rises |

Protocol when adding risks after v0.5:

1. Use the same RISK- template (assign next available number)
2. Set Added: v0.X to record which stage surfaced it
3. Recalculate category and total scores
4. Update README Risk Scorecard

Protocol when risk status changes:

1. Update the RISK- entry status in PRD.md
2. Recalculate effective score (Raw × new Status Weight)
3. Update README Risk Scorecard totals

Downstream Connections

RISK- entries feed into:

| Consumer | What It Uses | Example | |----------|--------------|---------| | README Risk Scorecard | Aggregated scores by category | Total score determines project risk level | | v0.5 Technical Stack Selection | RISK- constraints affect tech choices | RISK-003 (latency) → choose edge hosting | | v0.6 Architecture Design | Risk mitigations become architecture requirements | RISK-005 → add circuit breaker | | v0.7 Build Execution | Risk monitoring in EPIC | Track RISK-001 early signals | | KPI- Thresholds | Kill criteria from risks | "If RISK-002 triggers, evaluate pivot" |

Detailed References

• Interview question bank: See references/question-bank.md
• RISK- entry template: See assets/risk.md
• Example risk register: See references/examples.md