Overview
Voice of Customer (VOC) comes from multiple sources such as reviews, inquiries, CS, and SNS comments. As a VOC umbrella, this skill provides two modes in one place.
| Mode | Example trigger utterances | Deliverables |
|---|---|---|
| Review aggregation mode | "Gather and analyze reviews", "Coupang review keywords", "Multichannel review integration" | 4-channel review integration → 4-stage analysis of sentiment, keywords, insights, and action plan |
| VOC classification and triage mode | "Classify VOC", "Customer inquiry priority", "Emergency VOC classification" | 3-axis classification × KTAS 5-level urgency → processing order + response templates |
Since reviews are one source of VOC, the two modes connect naturally: reviews and inquiries gathered in review aggregation mode are passed to VOC classification and triage mode to assign processing priority. The mode is determined by the utterance, not by a CLI flag.
Review aggregation mode (multichannel review 4-stage analysis)
Integrates and analyzes multichannel reviews from Naver Smart Store, Coupang, own mall, YouTube, Instagram, etc., to generate a 4-stage analysis report of sentiment, keywords, insights, and action plan.
One-line responsibility: Product ID or channel-specific review data input → 4-stage analysis → (optional) PPT automation output.
4-stage analysis
- Sentiment analysis — Positive, negative, neutral ratio (%) + extraction of strong emotional expressions + sentiment trend over time (monthly/quarterly)
- Keyword analysis — Top 20 keywords (by frequency) + positive/negative classification + 1 representative sentence per keyword
- Insight derivation — 3 product strengths + 3 weaknesses + separate service (shipping, CS) evaluation + competitor mentions
- Top 5 action plan priorities — Immediate (1 week), short-term (1-3 months), mid-term (3-6 months), long-term (6 months+), monitoring
Collection guide by channel
| Channel | Collection method | Cost | Automation possibility |
|---|---|---|---|
| Naver Smart Store | API (Naver Commerce API) | Free (authentication required) | Automation possible |
| Coupang Wing | Wing API or Wing dashboard CSV | Free (sellers only) | Partially automated |
| Own mall (Cafe24/Imweb) | API or admin CSV | Free | Automation possible |
| YouTube | YouTube Data API | Free (within limits) | Automation possible |
| Instagram Graph API (brand accounts only) | Free | Automation possible |
Data integration: If there is a connected commerce MCP connector, query it directly; if not, generate the same output using export files or screenshots attached by the user or Claude's own analysis.
Review aggregation output format
{
"product": "비건 세럼 50ml",
"total_reviews": 1247,
"channels": {"smartstore": 580, "coupang": 412, "self_mall": 255},
"sentiment": {"positive": 78, "negative": 12, "neutral": 10},
"top_keywords": [
{"keyword": "흡수력 좋음", "freq": 234, "polarity": "+"},
{"keyword": "향이 별로", "freq": 87, "polarity": "-"}
],
"insights": {
"strengths": ["흡수력·발림성", "비건 인증 신뢰", "용기 디자인"],
"weaknesses": ["향 호불호", "용량 부족 (50ml)", "가격 부담"],
"service": "배송 빠름, CS 친절"
},
"action_plan": [
{"priority": 1, "term": "즉각", "action": "상세페이지에 향 정보 보강 (라벤더 향)"},
{"priority": 2, "term": "단기", "action": "100ml 대용량 SKU 추가"}
]
}
Review aggregation usage example:
"비건 세럼 50ml, 스마트스토어·쿠팡 통합 리뷰 분석"
→ 두 채널 리뷰 1,000+ 통합 → 4단 분석 → (선택) doc-pptx PPT 자동화
VOC classification and triage mode (3 axes + KTAS)
For sellers who cannot handle all opinions during a flood of VOC (Voice of Customer), 3-axis classification + KTAS emergency room classificationgenerates a processing priority matrix. It is more accurate when the 4-stage analysis results of review aggregation mode (negative reviews and repeated complaints) are provided as input.
One-line responsibility: VOC list + core customer definition + brand USP input → automatically generate 3-axis classification + KTAS 5-level urgency + processing order + response templates.
3-axis classification matrix
Axis 1: Customer Fit
| Classification | Definition | Weight |
|---|---|---|
| Core customer | Top 20% by LTV / VIP / 5+ repeat purchases | ×3 |
| General | Regular purchase / new member | ×2 |
| Non-core | One-time purchase / non-member | ×1 |
Axis 2: Frequency
| Classification | Definition | Weight |
|---|---|---|
| Repeated | 5+ instances of the same VOC | ×3 |
| Intermittent | 2-4 instances of the same VOC | ×2 |
| One-time | 1건 | ×1 |
Axis 3: Brand Value Relevance
| Classification | Definition | Weight |
|---|---|---|
| Direct USP hit | Related to the brand's core USP | ×3 |
| Related | General features and services | ×2 |
| Unrelated | External factors (shipping, payment, etc.) | ×1 |
Priority score
우선순위 점수 = 고객 핏 × 빈도 × 핵심 가치 관련성
예시:
VIP 고객 + 반복 + USP 직격 = 3×3×3 = 27 (최우선)
일반 + 일회성 + 무관 = 2×1×1 = 2 (후순위)
KTAS 5-level urgency mapping
Applying the emergency room patient classification system (KTAS):
| Level | Name | VOC score | Processing deadline | Example |
|---|---|---|---|---|
| Level 1 | 🔴 Immediate (Resuscitation) | 27 | <1 hour | VIP refund refusal / threat to report to the Ministry of Food and Drug Safety |
| Level 2 | 🟠 Emergent | 18-26 | <4 hours | Repeated negative reviews / core USP complaints |
| Level 3 | 🟡 Urgent | 12-17 | <24 hours | Repeated complaints from general customers |
| Level 4 | 🟢 Less Urgent | 4-11 | <72 hours | One-time general complaint |
| Level 5 | 🔵 Non-Urgent | 1-3 | <1주 | Praise and information inquiries |
Processing order + response templates
Response tone and content guide for each level:
- Level 1: Direct apology from the CEO + immediate refund, exchange, or compensation + promise of a callback
- Level 2: CS Team Lead response + 1:1 solution + follow-up
- Level 3: CS standard response + resolution within 24 hours
- Level 4: Automated response + processing within 72 hours
- Level 5: Automated response + expression of appreciation
Usage examples
"VOC 50건 입력, 핵심 USP '한국 농가 직거래' — VOC 분류해줘"
→ 50건 × 3축 분류 + KTAS 5단계 매핑 + 처리 순서 + 응답 템플릿
"스마트스토어·쿠팡 리뷰 모아서 분석해줘"
→ 리뷰 집계 모드: 4채널 통합 → 감정·키워드·인사이트·액션플랜 4단 분석
→ (이어서) 부정 리뷰·반복 불만을 VOC 분류·트리아지 모드로 우선순위화
Follow-up chain (text output review)
The text outputs of the two modes (insights, action plans, and service evaluations in review aggregation mode; response templates Level 1-5 and processing order guidance in VOC classification mode) are moai-coworker:ai-slop-reviewerautomatically post-processed and chained. If VIP/emergency response messages or external reports contain AI patterns (excessive adjectives, translationese), credibility collapses, so review is essential.
리뷰 집계 모드 → moai-coworker:ai-slop-reviewer → (선택) moai-officer:doc-pptx
VOC 분류·트리아지 모드 → moai-coworker:ai-slop-reviewer → 사용자 검토 → CS 발송
Related skills
cs-channel-message— Copywriting based on response messages and action plansmoai-officer:doc-pptx— PPT automation for review analysis resultsmoai-coworker:ai-slop-reviewer— AI slop review of text outputs (automatic chain, HARD)moai-writer:korean-humanize— Removing Korean AI tells (after slop review, required)