Agent Skills: Feedback Loop Setup

Feedback Loop Setup

Position in workflow: v0.9 Launch Metrics → v0.9 Feedback Loop Setup → v1.0 Market Adoption

Consumes

This skill requires prior work from v0.9 Launch Metrics and v0.1-v0.8:

• GTM-* launch channels (from v0.9 GTM Strategy) — Active launch channels (Product Hunt, email, paid ads, etc.) become feedback sources; GTM- messaging and channels inform where feedback will arrive
• MON-* monitoring dashboards and alerts (from v0.8 Monitoring Setup) — MON- thresholds (latency, error rate, performance) define what qualifies as critical feedback; monitoring alerts can trigger deep-dive user research
• KPI-* launch targets and baselines (from v0.9 Launch Metrics) — KPI- thresholds (Day 1/7/30/90 targets) inform feedback urgency and trigger investigation when below target; baseline performance metrics (p95 latency, error rate, conversion rate) provide context for performance feedback
• CFD-* baseline entries (from v0.1-v0.4) — Baseline customer feedback hypotheses (user pain points, value propositions, competitive alternatives) become validation targets post-launch; feedback loop confirms or contradicts CFD- assumptions
• PER-* personas (from v0.4 Persona Definition) — Persona segments (PER-001 Startup Founder, PER-002 Team Lead) enable feedback categorization by user type and prioritization by persona importance

This skill assumes v0.9 Launch Metrics is live with KPI- thresholds established, GTM- channels are active, and MON- dashboards are displaying baseline metrics.

Produces

This skill creates/updates:

• CFD-* post-launch feedback entries (feedback capture specifications, channel/type-based) — Every piece of user feedback becomes a CFD- entry with source, sentiment, impact, and action taken; traced to GTM- channels and user personas
• Feedback processing workflow/matrix — Triage → Categorization → Prioritization → Action mapping showing how feedback flows from capture to ID updates (CFD- → FEA-/BR-/RISK- → EPIC-)
• CFD-* update entries — CFD- entries updated with resolution status, outcome, and follow-up evidence, enabling confidence progression (initial feedback → validated pattern → implemented action → confirmed outcome)

All CFD-* post-launch entries are evidential feedback records, not confidence-based themselves but supporting confidence scoring on OTHER IDs:

• Timestamped (when feedback was received, to track trends and velocity)
• Sourced (channel, user segment, user ID if available for follow-up)
• Categorized (UX | Performance | Feature Gap | Bug | Praise | Confusion for trend analysis)
• Prioritized (Critical/High/Medium/Low with impact justification)
• Actionable (every CFD- either triggers ID creation/update or documents "won't fix" decision)
• Closed-loop (user receives response and can verify resolution)

Example CFD- post-launch entries:

CFD-101: "Can't figure out how to export my data"
Type: Support Ticket
Source: Intercom (GTM-002 email → user support request)
Date: 2025-01-15
User Segment: PER-001 (Startup Founder)

Verbatim: "I've been using the tool for a week and I can't find any way to export my work."

Processed:
  Category: Feature Gap
  Sentiment: Frustrated
  Priority: High
  Frequency: Repeated (3rd request this week)

Impact Assessment:
  Users Affected: ~50 (based on support volume)
  KPI Impact: KPI-104 (D7 Retention) — export needed for team use case
  Revenue Risk: High — multiple users mentioned "dealbreaker"

Action:
  Response: "Thanks for reaching out! Export is on our roadmap."
  Internal Action: Escalated to product team, added to backlog
  Linked IDs: FEA-025 (Export Feature) created, EPIC-05 updated
  Status: In Progress

Resolution:
  Outcome: FEA-025 shipped in v1.2
  Date: 2025-02-01
  Follow-up: Emailed user with release notes

Linked IDs: GTM-002 (email channel source), PER-001 (persona), KPI-104 (affected metric), FEA-025 (action taken), EPIC-05 (implementation)

---

CFD-102: NPS Detractor Response
Type: NPS Response
Source: In-App Survey (MON-005 trigger)
Date: 2025-01-18
User Segment: PER-002 (Team Lead)

Verbatim: "Score: 4. Too slow. Takes forever to load projects and I give up waiting."

Processed:
  Category: Performance
  Sentiment: Negative
  Priority: Critical
  Frequency: Trending (NPS dropped 10 points this week)

Impact Assessment:
  Users Affected: ~200 (20% of NPS responses mention speed)
  KPI Impact: KPI-103 (Activation), KPI-104 (Retention) — both trending down
  Revenue Risk: High — performance is activation blocker

Action:
  Response: N/A (anonymous survey)
  Internal Action: Performance spike investigation started (MON-001 latency breach detected)
  Linked IDs: RISK-012 (Performance Degradation) escalated, EPIC-06 prioritized for optimization
  Status: In Progress

Resolution:
  Outcome: Database query optimization deployed, latency restored to baseline
  Date: 2025-01-22
  Follow-up: Next NPS cycle (Day 30) will measure improvement

Linked IDs: MON-005 (dashboard source), PER-002, KPI-103, KPI-104, MON-001 (latency baseline), RISK-012, EPIC-06

---

CFD-103: Community Feature Request (Dark Mode)
Type: Community Post
Source: Discord #feature-requests (GTM-005 community channel)
Date: 2025-01-20
User Segment: Power Users (multiple PER-)

Verbatim: "Thread: 47 messages discussing dark mode. Summary: 15 unique users requesting."

Processed:
  Category: Feature Gap
  Sentiment: Neutral (constructive)
  Priority: Medium
  Frequency: Repeated (ongoing, 15 users vocal)

Impact Assessment:
  Users Affected: 15+ vocal, likely more silent
  KPI Impact: Minor — nice-to-have, not activation blocker; may reduce churn for night users
  Revenue Risk: Low

Action:
  Response: Community manager acknowledged, added to public roadmap
  Internal Action: Added to backlog as P2 feature
  Linked IDs: FEA-030 (Dark Mode) created, posted on public roadmap
  Status: Acknowledged

Resolution:
  Outcome: Pending — scheduled for Q2 release
  Date: N/A
  Follow-up: Posted on public roadmap

Linked IDs: GTM-005 (community channel), PER-* (multiple personas), FEA-030, public roadmap

Feedback → ID Flow

Each CFD- post-launch entry triggers cascading updates:

| Feedback Type | Creates/Updates | Confidence Impact | Example | |---------------|-----------------|-------------------|---------| | Feature Request | FEA-, BR-FEA- | Increases FEA- confidence (user interview → beta validation) | CFD-101 (export request, 3rd this week) → FEA-025 (confidence: 2→3, source: support-requests-2025-01) | | Performance Complaint | MON- threshold, RISK- escalation | Triggers MON- investigation; may update RISK- severity | CFD-102 (slow, 20% mention) → MON-001 threshold validation → RISK-012 escalation | | UX Confusion | SCR-, UJ- refinement | Informs screen redesign without changing foundational journey | "Can't find export" → SCR-005 (export button placement) update | | Bug Report | RISK- or direct fix | RISK- frequency increases → triggers prioritization | Critical bugs → P0 RISK- entry | | Praise/Testimonial | CFD- (evidence), GTM- (social proof) | Confirms CFD- hypothesis; can become GTM- case study | "Love this feature!" → CFD- entry → GTM-015 (testimonial) |

This feedback loop enables evidence-driven iteration: feedback patterns → ID updates → implementation → launch validation → next iteration.

Downstream Connections

| Consumer | What It Uses | Example | |----------|--------------|---------| | v1.0 Market Adoption Planning | CFD- feedback patterns inform roadmap | 10× CFD- export requests → FEA-025 move to P1 | | Product Development | CFD- → FEA-, BR- updates feed next EPIC | CFD-102 performance complaints → EPIC-06 optimization prioritized | | Sales/Marketing | CFD- testimonials become GTM assets | CFD-103 community enthusiasm → GTM-015 case study | | Support Team | CFD- patterns become FAQ and onboarding | Repeated "can't export" → FAQ article | | Risk Management | CFD- negative trends escalate RISK- | NPS dropping → RISK-012 escalation | | KPI Accountability | CFD- confirms KPI- achievement | KPI-104 (D7 Retention) gaps trigger CFD- investigation |

Purpose

Establish systematic channels for capturing, processing, and acting on post-launch user feedback—closing the loop between user experience and product iteration.

Core Concept: Feedback as Fuel

Feedback is not a task to complete—it is fuel for iteration. Every piece of feedback should flow into the ID graph, informing future CFD-, BR-, FEA-, or RISK- entries. If feedback sits in a spreadsheet, it's not feedback—it's noise.

Feedback Channels

| Channel | Type | Best For | Response Time | |---------|------|----------|---------------| | In-App | Prompted | Contextual reactions | Real-time | | Support | Reactive | Issues, requests | <24h | | Community | Proactive | Discussion, ideas | Ongoing | | Surveys | Scheduled | Structured data | Periodic | | Analytics | Passive | Behavior signals | Continuous |

Execution

1. Map feedback touchpoints

   • Where do users already reach out?
   • Where should we actively prompt?
   • What channels from GTM- are active?

2. Design feedback capture

   • In-app widgets (NPS, CSAT, feature requests)
   • Support ticket taxonomy
   • Community moderation workflow
   • Survey schedule and instruments

3. Define processing workflow

   • Who triages incoming feedback?
   • How does it become CFD- entries?
   • What triggers action?

4. Establish feedback → ID flow

   • Feedback → CFD-
   • CFD- → BR-, FEA-, RISK- updates
   • Updates → EPIC- for implementation

5. Set up monitoring

   • Volume metrics
   • Sentiment tracking
   • Response time SLAs

6. Create CFD- entries for post-launch feedback

CFD- Output Template (Post-Launch Feedback)

CFD-XXX: [Feedback Title]
Type: [Support Ticket | Feature Request | Bug Report | NPS Response | Community Post | Survey Response]
Source: [Intercom | Zendesk | Discord | In-App | Email | Twitter]
Date: [When received]
User Segment: [PER-XXX if identifiable]

Verbatim: "[Exact user quote or description]"

Processed:
  Category: [UX | Performance | Feature Gap | Bug | Praise | Confusion]
  Sentiment: [Positive | Neutral | Negative | Frustrated]
  Priority: [Critical | High | Medium | Low]
  Frequency: [One-off | Repeated | Trending]

Impact Assessment:
  Users Affected: [Count or estimate]
  KPI Impact: [KPI-XXX affected if applicable]
  Revenue Risk: [High | Medium | Low | None]

Action:
  Response: [How we responded to user]
  Internal Action: [What we're doing about it]
  Linked IDs: [BR-XXX, FEA-XXX, RISK-XXX created/updated]
  Status: [New | Acknowledged | In Progress | Resolved | Won't Fix]

Resolution:
  Outcome: [What happened]
  Date: [When resolved]
  Follow-up: [Did we close the loop with user?]

Note: See Produces section above for detailed CFD- examples with full traceability links.

Feedback Collection Methods

In-App Feedback

| Method | When to Use | Question | |--------|-------------|----------| | NPS | After activation, monthly | "How likely to recommend?" (0-10) | | CSAT | After support interaction | "How satisfied?" (1-5) | | CES | After key action | "How easy was this?" (1-7) | | Feature Request | Persistent widget | "What's missing?" | | Bug Report | Error states | "What went wrong?" |

Survey Cadence

| Survey | Frequency | Purpose | |--------|-----------|---------| | NPS | Monthly | Overall sentiment tracking | | Onboarding Exit | After churn signal | Why didn't they activate? | | Feature Satisfaction | Post-release | Did this solve the problem? | | Annual Deep Dive | Yearly | Strategic feedback |

Passive Signals

| Signal | What It Indicates | Action Trigger | |--------|-------------------|----------------| | Rage clicks | Frustration | UX investigation | | Drop-off | Confusion or friction | Funnel analysis | | Feature abandonment | Poor value delivery | User interview | | Error rates | Technical issues | Bug investigation |

Feedback Processing Workflow

CAPTURE → TRIAGE → CATEGORIZE → PRIORITIZE → ACTION → CLOSE LOOP

1. CAPTURE
   - All channels → central inbox

2. TRIAGE (Daily)
   - Critical: <4h response
   - High: <24h response
   - Medium/Low: Weekly review

3. CATEGORIZE
   - Apply CFD- template
   - Link to existing IDs

4. PRIORITIZE
   - Frequency × Impact × Revenue Risk
   - Weekly prioritization meeting

5. ACTION
   - Create/update IDs (BR-, FEA-, RISK-)
   - Add to EPIC- backlog
   - Communicate internally

6. CLOSE LOOP
   - Respond to user
   - Update CFD- status
   - Verify resolution

Sentiment Monitoring

Track aggregate sentiment over time:

| Metric | Calculation | Target | |--------|-------------|--------| | NPS | % Promoters - % Detractors | >30 | | CSAT | % Satisfied (4-5) | >80% | | Support Volume | Tickets per 100 users | <5 | | Response Time | Median first response | <4h | | Resolution Rate | % resolved within SLA | >90% |

Anti-Patterns

| Pattern | Signal | Fix | |---------|--------|-----| | Feedback graveyard | Collect but never act | Mandate weekly triage meeting | | Only negative | No positive feedback captured | Celebrate wins, capture praise | | No closing loop | Users never hear back | Require follow-up on High+ priority | | Volume without insight | "We got 500 tickets" | Categorize and trend analysis | | Building in silence | Ship features, don't validate | Post-release surveys | | Anecdote-driven | "One user said..." | Require frequency data |

Quality Gates

Before proceeding to v1.0 Market Adoption:

• [ ] All feedback channels identified and configured
• [ ] In-app feedback widgets deployed
• [ ] Support ticket taxonomy defined
• [ ] Community monitoring active
• [ ] Processing workflow documented and assigned
• [ ] Feedback → ID flow established
• [ ] Sentiment metrics baselined

Detailed References

• Feedback channel setup: See references/channel-setup.md
• CFD- post-launch template: See assets/cfd-feedback-template.md
• Survey question bank: See references/survey-questions.md
• Sentiment analysis guide: See references/sentiment-guide.md