| name | afrexai-voc-engine |
| description | Complete Voice of Customer system — collect, analyze, and operationalize customer feedback at scale. Covers NPS/CSAT/CES measurement, customer interview methodology, feedback taxonomy, feature request prioritization, sentiment analysis, closed-loop workflows, and VoC-driven product decisions. Use when building feedback systems, running customer interviews, measuring satisfaction, analyzing feature requests, reducing churn, or closing the feedback loop. Trigger on "customer feedback", "voice of customer", "NPS", "CSAT", "CES", "feature requests", "feedback system", "customer interviews", "satisfaction survey", "churn analysis". |
Voice of Customer (VoC) Engine
Complete system for collecting, analyzing, and operationalizing customer feedback to drive product decisions, reduce churn, and increase expansion revenue.
Phase 1: VoC Program Design
Program Brief
voc_program:
company: "[Company Name]"
product: "[Product/Service]"
stage: "[Pre-PMF | Growth | Scale | Enterprise]"
customer_count: "[approximate]"
current_nps: "[score or unknown]"
current_churn: "[monthly % or unknown]"
primary_goal: "[reduce churn | improve NPS | prioritize roadmap | find PMF]"
segments:
- name: "[Enterprise / SMB / Consumer]"
size: "[count]"
arr_contribution: "[%]"
priority: "[1-3]"
existing_channels:
- "[list current feedback collection methods]"
gaps:
- "[what feedback are you NOT collecting?]"
VoC Maturity Assessment
Score your current program (1-5 per dimension):
| Dimension | 1 (Ad Hoc) | 3 (Structured) | 5 (Operationalized) |
|---|
| Collection | Sporadic, reactive only | Multiple channels, some regularity | Automated, multi-channel, triggered |
| Analysis | Read individual comments | Tag and categorize | Quantified themes, trend tracking |
| Distribution | Stays with support team | Shared in meetings | Real-time dashboards, auto-routed |
| Action | Occasional fixes | Quarterly roadmap input | Closed-loop, feedback drives decisions |
| Measurement | No tracking | NPS or CSAT exists | Full metric suite, benchmarked |
Score interpretation:
- 5-10: Foundation needed — start with Phase 2 basics
- 11-18: Growing — focus on analysis and distribution
- 19-25: Mature — optimize closed-loop and ROI tracking
Phase 2: Feedback Collection Architecture
Channel Strategy
Design feedback collection around the customer journey:
collection_channels:
in_app_widget:
placement: "[specific screens/moments]"
trigger: "User-initiated (feedback button)"
format: "Open text + optional category selector"
volume: "High"
quality: "Medium (context-rich but brief)"
support_tickets:
source: "[Intercom / Zendesk / Help Scout / email]"
tagging: "Auto-tag feedback themes (see taxonomy)"
volume: "High"
quality: "High (real problems, real context)"
feature_request_board:
tool: "[Canny / ProductBoard / custom]"
voting: true
status_updates: true
volume: "Medium"
quality: "High (considered requests)"
nps_survey:
frequency: "Quarterly"
trigger: "Email to active users (>30 days tenure)"
follow_up: "Open text for score explanation"
segments: "[by plan tier, tenure, usage level]"
csat_survey:
trigger: "After key interactions (onboarding complete, support resolved, feature shipped)"
Survey Design Rules
- Max 3 questions per survey (response rate drops 20% per additional question)
- One metric question + one open text is the minimum viable survey
- Timing matters: Send within 24 hours of the trigger event
- Mobile-first: Design for thumb-scrolling
- Never survey angry: Wait 48 hours after a negative support interaction
- Frequency cap: No user gets surveyed more than once per quarter (except event-driven)
- Always explain why: "Help us improve [specific thing]" > "Take our survey"
Sample Size Calculator
| Customer Base | Confidence 95%, ±5% | Confidence 90%, ±5% | Minimum Viable |
|---|
| 100 | 80 | 74 | 30 |
| 500 | 217 | 176 | 50 |
| 1,000 | 278 | 213 | 75 |
| 5,000 | 357 | 258 | 100 |
| 10,000+ | 370 | 264 | 150 |
Expected response rates by channel:
- In-app: 10-15%
- Email NPS: 20-40%
- Post-support CSAT: 15-25%
- Cancellation: 30-50% (required = 80%+)
Phase 3: Customer Interview Methodology
Interview Types
| Type | When | Duration | Sample | Goal |
|---|
| Discovery | Pre-build, exploring problem | 45-60 min | 8-12 | Validate problem exists |
| Usability | Feature in development | 30-45 min | 5-7 | Test assumptions |
| Satisfaction | Ongoing, quarterly | 30 min | 5-8 per segment | Deep qualitative NPS |
| Churn | After cancellation | 15-20 min | All willing | Understand why they left |
| Win/Loss | After deal close | 20-30 min | 5+ per quarter | Sales process feedback |
| Advisory | Strategic, quarterly | 60 min | 3-5 power users | Co-create roadmap |
Interview Guide Template
interview:
type: "[Discovery / Satisfaction / Churn / Win-Loss]"
participant: "[name, role, company, segment]"
date: "[YYYY-MM-DD]"
interviewer: "[name]"
pre_interview:
- Review account history (usage data, support tickets, NPS scores)
- Check CRM for relationship context
- Prepare 3 hypotheses to test
warm_up:
- "Tell me about your role and what a typical week looks like."
- "How long have you been using [product]?"
- "[Acknowledge something specific about their usage]"
context:
- "What problem were you trying to solve when you found us?"
- "What were you using before? What worked/didn't?"
- "Walk me through how [product] fits into your workflow."
core_exploration:
-
[]
[]
[]
[]
[]
Interview Techniques
The Mom Test (Rob Fitzpatrick):
- Talk about their life, not your idea
- Ask about specifics in the past, not hypotheticals
- Talk less, listen more
- Bad: "Would you use a feature that does X?" (they'll say yes to be nice)
- Good: "Tell me about the last time you tried to do X. What happened?"
Five Whys for Root Cause:
- "That's frustrating. Why is that a problem for you?"
- "Why does that matter to your work?"
- "What happens when that goes wrong?"
- "Why can't you work around it?"
- "What would change if we fixed this?"
Silence technique: After they answer, count to 5 silently. They'll often add the most valuable insight in the silence.
Mirror technique: Repeat their last 3 words as a question. They'll elaborate without you leading.
Interview Synthesis Template
After every 5 interviews in a batch:
synthesis:
batch: "[Interview batch name]"
dates: "[range]"
participants: "[count, segments]"
themes:
- theme: "[Name]"
frequency: "[X of Y participants mentioned]"
severity: "[Critical / High / Medium / Low]"
representative_quotes:
- "[exact quote]" — [participant type]
- "[exact quote]" — [participant type]
implications: "[what this means for product/business]"
recommended_action: "[specific next step]"
surprises:
- "[Things you didn't expect]"
segments_divergence:
- "[Where different segments disagreed]"
confidence_level: "[High / Medium / Low]"
next_steps: []
Phase 4: Feedback Taxonomy & Classification
Category Hierarchy
feedback_taxonomy:
product:
usability:
- "Confusing UI / navigation"
- "Feature hard to find"
- "Too many steps to complete task"
- "Mobile experience issues"
functionality:
- "Feature doesn't work as expected"
- "Missing capability"
- "Performance / speed"
- "Integration issues"
reliability:
- "Bugs / errors"
- "Data loss / corruption"
- "Downtime / availability"
experience:
onboarding:
- "Setup too complex"
- "Documentation unclear"
- "Time to value too long"
support:
- "Response time"
- "Resolution quality"
- "Self-service gaps"
communication:
- "Product updates unclear"
- "Billing confusion"
- "Status page / transparency"
value:
pricing:
Sentiment Scoring
For every piece of feedback, score:
| Dimension | Scale | Guide |
|---|
| Sentiment | -2 to +2 | -2=angry, -1=frustrated, 0=neutral, +1=positive, +2=delighted |
| Urgency | 1-5 | 1=nice-to-have, 3=important, 5=blocking/churning |
| Frequency | Count | How many unique customers mention this |
| Revenue at risk | $ | ARR of customers mentioning this theme |
| Effort to fix | S/M/L/XL | Engineering estimate |
Auto-Tagging Rules
When processing feedback, apply tags automatically:
IF contains("cancel", "churn", "leaving", "switching") → tag: churn_risk, urgency: 5
IF contains("bug", "error", "broken", "crash") → tag: bug_report, urgency: 4
IF contains("wish", "would be nice", "if only") → tag: feature_request, urgency: 2
IF contains("love", "amazing", "best") → tag: positive_signal, sentiment: +2
IF contains("competitor", "alternative", "other tool") → tag: competitive_intel
IF contains("price", "cost", "expensive", "cheaper") → tag: pricing_feedback
IF contains("confusing", "hard to", "can't figure") → tag: usability_issue, urgency: 3
Phase 5: Metric Systems
NPS (Net Promoter Score)
Question: "How likely are you to recommend [product] to a colleague? (0-10)"
Scoring:
- 0-6: Detractor
- 7-8: Passive
- 9-10: Promoter
- NPS = % Promoters - % Detractors (range: -100 to +100)
Benchmarks by industry:
| Industry | Poor | Average | Good | Excellent |
|---|
| SaaS B2B | <20 | 20-40 | 40-60 | 60+ |
| SaaS B2C | <10 | 10-30 | 30-50 | 50+ |
| E-commerce | <20 | 20-40 | 40-60 | 60+ |
| Financial Services | <10 | 10-30 | 30-50 | 50+ |
| Healthcare Tech | <15 | 15-35 | 35-55 | 55+ |
NPS follow-up questions (by score):
- Detractor (0-6): "We're sorry to hear that. What's the #1 thing we could improve?"
- Passive (7-8): "What would it take to make us a 9 or 10?"
- Promoter (9-10): "That's great! What do you value most about us?"
NPS action rules:
- Score dropped >10 points quarter-over-quarter → immediate investigation
- Detractor from Enterprise account → CSM alert within 24 hours
- Promoter → trigger referral/review/case study ask within 7 days
CSAT (Customer Satisfaction Score)
Question: "How satisfied were you with [specific interaction]? (1-5)"
Score: (Satisfied responses [4-5] / Total responses) × 100
Benchmarks: 75% = decent, 85% = good, 90%+ = excellent
When to use: After specific interactions (support, onboarding, feature launch)
CES (Customer Effort Score)
Question: "How easy was it to [specific task]? (1-7, 1=very difficult, 7=very easy)"
Score: Average of all responses
Benchmarks: <4 = high effort (problem), 4-5 = acceptable, 5+ = low effort (good)
Why CES matters: CES is the strongest predictor of future purchasing behavior. High-effort experiences drive 96% of disloyalty.
Composite Health Score
Combine metrics for overall VoC health:
voc_health_score:
nps:
weight: 30
score: "[0-100 normalized: (NPS + 100) / 2]"
csat:
weight: 20
score: "[CSAT %]"
ces:
weight: 15
score: "[(CES / 7) × 100]"
feedback_volume:
weight: 10
score: "[trend: increasing = good]"
response_rate:
weight: 10
score: "[survey response rate %]"
closed_loop_rate:
weight: 15
score: "[% of feedback items with documented response/action]"
total: "[weighted sum / 100]"
grade: "[A: 80+, B: 65-79, C: 50-64, D: 35-49, F: <35]"
Phase 6: Feature Request Prioritization
RICE + VoC Framework
Score each feature request:
| Factor | Formula | Weight |
|---|
| Reach | # customers requesting (or affected) per quarter | 25% |
| Impact | Score 0.5 (minimal), 1 (low), 2 (medium), 3 (high) | 25% |
| Confidence | % confidence in estimates (50-100%) | 20% |
| Effort | Person-weeks to build | 15% |
| Revenue Signal | ARR at risk or expansion ARR | 15% |
RICE-V Score = (Reach × Impact × Confidence × Revenue Signal) / Effort
Prioritization Decision Tree
1. Is this blocking revenue? (churn risk from top accounts)
→ YES: Fast-track (Sprint 1)
→ NO: Continue ↓
2. Does it affect >20% of customers?
→ YES: High priority (next quarter)
→ NO: Continue ↓
3. Is it from a strategic segment we're targeting?
→ YES: Medium-high priority
→ NO: Continue ↓
4. Does it align with product vision?
→ YES: Backlog with planned quarter
→ NO: Decline with explanation (close the loop!)
Saying "No" Framework
Not every feature request should be built. Decline gracefully:
Template: "Thanks for this suggestion! We've heard this from [X] customers. Right now, our roadmap is focused on [theme] because [reason]. We're not planning to build this in the next [timeframe], but we've logged it. Here's what we ARE building that might help: [alternative]. We'll update this request if our plans change."
When to say no:
- Request serves <5% of customers
- Conflicts with product vision
- Better solved by integration/partner
- Effort/value ratio is poor
- Customer is better served by a different product
Phase 7: Closed-Loop Workflow
The Feedback Loop (4 stages)
COLLECT → ANALYZE → ACT → COMMUNICATE
↑ |
└──────────────────────────────┘
Closed-Loop Response Templates
Acknowledgment (within 24 hours):
"Thanks for sharing this feedback. We've logged it as [category] and it's being reviewed by our [product/engineering] team. We'll update you when we have a plan."
In Progress:
"Quick update on your feedback about [topic] — we're actively working on this. Expected [timeframe]. Here's what's changing: [brief description]."
Shipped:
"Remember when you told us about [original feedback]? We fixed it! Here's what changed: [specific improvement]. Try it out and let us know what you think."
Declined (see "Saying No" above)
Response SLAs by Feedback Type
| Type | Acknowledge | Triage | Resolve/Respond |
|---|
| Bug report | 4 hours | 24 hours | Per severity SLA |
| Feature request | 48 hours | 1 week | Quarterly review |
| Complaint | 4 hours | 24 hours | 72 hours |
| Praise | 24 hours | N/A | Share internally |
| Churn feedback | 24 hours | 48 hours | 1 week (win-back) |
| NPS detractor | 24 hours | 48 hours | 1 week |
Closed-Loop Tracking
feedback_item:
id: "FB-[YYYY]-[####]"
source: "[channel]"
customer: "[name, segment, ARR]"
date_received: "[YYYY-MM-DD]"
category: "[from taxonomy]"
sentiment: "[-2 to +2]"
urgency: "[1-5]"
status: "[New | Acknowledged | Triaging | In Progress | Shipped | Declined | Won't Fix]"
assigned_to: "[team/person]"
date_acknowledged: "[YYYY-MM-DD]"
date_resolved: "[YYYY-MM-DD]"
resolution: "[what was done]"
customer_notified: "[yes/no + date]"
customer_satisfied: "[yes/no/unknown]"
linked_items:
jira: "[ticket ID]"
roadmap: "[feature/initiative]"
other_feedback: "[related FB IDs]"
Phase 8: VoC Reporting & Dashboards
Weekly VoC Summary
weekly_voc:
period: "[week of YYYY-MM-DD]"
volume:
total_feedback: "[count]"
by_channel:
support: "[count]"
in_app: "[count]"
surveys: "[count]"
interviews: "[count]"
trend: "[↑/↓/→ vs last week]"
metrics:
nps: "[score] ([↑/↓] [X] from last period)"
csat: "[%] ([↑/↓] [X]%)"
ces: "[score] ([↑/↓])"
top_themes:
- theme: "[#1 theme]"
mentions: "[count]"
sentiment: "[avg]"
revenue_at_risk: "[$]"
action: "[what we're doing]"
- theme: "[#2]"
- theme: "[#3]"
alerts:
- "[Enterprise detractor: Company X, NPS dropped from 8 to 3]"
- "[New theme emerging: Y mentioned by 5 customers this week]"
closed_loop:
items_received: "[count]"
Monthly VoC Report (for leadership)
Structure:
- Executive summary — 3 bullet headline (NPS trend, top risk, top opportunity)
- Metric trends — NPS/CSAT/CES with month-over-month and quarter-over-quarter
- Theme analysis — Top 5 themes with frequency, sentiment, revenue impact
- Segment breakdown — Feedback patterns by customer segment
- Competitive signals — What customers say about competitors
- Roadmap alignment — How feedback maps to current roadmap
- Action items — What we're doing about top themes
- Closed-loop performance — % acknowledged, % resolved, avg time
Quarterly Strategic VoC Review
Go deeper quarterly:
- Cohort analysis: Do newer customers have different feedback patterns?
- Segment analysis: Which segments are happiest/unhappiest?
- Journey mapping: Where in the journey does feedback spike?
- Trend analysis: Which themes are growing/shrinking?
- Competitive landscape: Are we gaining or losing on key dimensions?
- ROI of changes: Did features we shipped based on feedback improve metrics?
Phase 9: VoC-Driven Product Decisions
Evidence Hierarchy
When making product decisions, rank evidence:
- Behavioral data (what they DO) — usage analytics, conversion, retention
- Revealed preferences (what they PAY for) — upgrade triggers, pricing experiments
- Qualitative interviews (what they SAY in depth) — discovery, satisfaction
- Survey data (what they REPORT) — NPS, CSAT, feature votes
- Support tickets (what they COMPLAIN about) — pain points
- Feature request votes (what they ASK for) — stated preferences
- Social/review feedback (what they TELL others) — public sentiment
Rule: Never build based on Level 6-7 alone. Always triangulate with Level 1-3.
Product Decision Template
product_decision:
what: "[Feature/change being considered]"
voc_evidence:
quantitative:
customers_requesting: "[count]"
arr_represented: "[$]"
nps_impact: "[detractors mentioning this]"
support_tickets: "[count related]"
qualitative:
interview_quotes:
- "[quote]" — [segment]
themes_connected: "[from taxonomy]"
behavioral:
usage_data: "[relevant metrics]"
churn_correlation: "[if applicable]"
decision: "[Build / Decline / Investigate further]"
confidence: "[High / Medium / Low]"
expected_impact:
nps_change: "[estimated]"
churn_reduction: "[estimated]"
expansion_revenue: "[estimated]"
success_metrics:
- "[How we'll know this worked]"
review_date: "[When to check impact]"
Phase 10: Advanced VoC Patterns
Customer Advisory Board (CAB)
Structure:
- 8-12 customers across segments
- Quarterly 90-minute sessions
- Mix: 60% product roadmap feedback, 20% industry trends, 20% relationship building
- Incentive: Early access to features, executive access, influence on roadmap
Selection criteria:
- Represents target segment
- Engaged user (not just big logo)
- Constructive (not just complainers or cheerleaders)
- Diverse use cases
- Willing to commit time
Competitive Feedback Mining
When customers mention competitors:
competitive_mention:
competitor: "[name]"
context: "[switching to / comparing / switched from]"
feature_gap: "[what competitor has that we don't]"
our_advantage: "[what they like about us vs competitor]"
customer_segment: "[type]"
deal_size: "[$]"
action: "[product response needed?]"
Track quarterly: "Competitor mention frequency" — rising mentions of a specific competitor = early warning.
Feedback-Driven Pricing Insights
Customer feedback reveals pricing issues:
| Signal | Meaning | Action |
|---|
| "Too expensive" with no specifics | Perception problem, not price problem | Improve value communication |
| "Feature X should be in my plan" | Packaging issue | Review tier boundaries |
| "Competitor is cheaper" | Price positioning | Competitive analysis |
| "I'd pay more for Y" | Expansion opportunity | Test premium tier/add-on |
| "Not sure what I'm paying for" | Value clarity gap | Improve billing page, add ROI dashboard |
Multi-Product VoC
If you have multiple products:
- Separate NPS per product (not one company NPS)
- Cross-product feedback: "I wish [Product A] did what [Product B] does"
- Unified customer view: all feedback across products in one place
- Cross-sell signals in feedback: mentions of adjacent problems
VoC for Pre-PMF Companies
Simplified system when you have <50 customers:
- Talk to every customer — monthly calls, 20 minutes
- Track in a spreadsheet — name, date, feedback, theme, action taken
- One metric: Sean Ellis test — "How would you feel if you could no longer use [product]?" (>40% "very disappointed" = PMF)
- Ship weekly, ask weekly — tight feedback loop
- Don't survey — sample size too small, just talk
Phase 11: Edge Cases
Negative Feedback Spike
- Don't panic. Check: is it a bug (concentrated), a change (adoption curve), or a trend (systemic)?
- Triage: Is it one loud customer or multiple independent voices?
- Respond: Acknowledge publicly if visible (social, reviews). Fix if bug. Explain if change.
Feedback Contradictions
- Segment A wants simplicity, Segment B wants power features
- Resolution: Check which segment is your ICP. Build for ICP, offer workarounds for others. Consider tiered product.
Survey Fatigue
- Response rates dropping below 10% = fatigue
- Fix: Reduce frequency, make surveys shorter, show impact ("You told us X, we built Y"), vary channels
Vocal Minority Problem
- 3 customers scream for feature X, 300 silently use feature Y
- Rule: Never prioritize by volume of complaints alone. Cross-reference with usage data and revenue.
Cultural Differences (International)
- NPS scores vary by culture (Japan averages lower, US averages higher)
- Fix: Benchmark within regions, not globally. Adjust expectations by market.
Feedback from Free Users
- Lower weight than paying customers for prioritization
- BUT valuable for: conversion barriers, first-impression issues, market understanding
- Rule: Track separately. Flag "conversion blockers" for growth team.
Quality Scoring Rubric
Rate your VoC program (0-100):
| Dimension | Weight | 0-25 | 50 | 75 | 100 |
|---|
| Collection coverage | 15% | 1 channel | 3 channels | 5+ channels | Full journey coverage |
| Analysis depth | 15% | Read comments | Categorize | Quantified themes | Predictive insights |
| Response time | 10% | Weeks | Days | Hours | Real-time SLA met |
| Closed-loop rate | 15% | <20% | 40-60% | 60-80% | >80% with satisfaction check |
| Decision influence | 15% | Ignored | Occasional input | Regular roadmap input | Systematic decision driver |
| Metric tracking | 10% | None | NPS exists | NPS + CSAT + CES | Full suite + benchmarks |
| Segmentation | 10% | None | By plan tier | Multi-dimensional | Persona-level insights |
| ROI measurement | 10% | None | Anecdotal | Feature-level impact | Revenue-attributed |
Natural Language Commands
Use these to interact:
- "Set up VoC program" → Generate program brief + collection channels + first survey
- "Design NPS survey" → Create NPS survey with follow-ups + distribution plan
- "Prep customer interview" → Generate interview guide for specific customer/type
- "Analyze this feedback" → Categorize, score sentiment, tag, recommend action
- "Synthesize interview batch" → Create synthesis from multiple interview notes
- "Prioritize feature requests" → RICE-V score + prioritization recommendation
- "Generate VoC report" → Weekly/monthly/quarterly report from data
- "Score our VoC program" → Run maturity assessment + improvement recommendations
- "Close the loop on [item]" → Generate appropriate response for feedback item
- "Compare our NPS to benchmarks" → Industry benchmark comparison + trend analysis
- "Find churn signals in feedback" → Analyze feedback for churn risk patterns
- "Build customer advisory board" → Selection criteria + program structure + first agenda
Built by AfrexAI — Turning customer voices into revenue decisions.
