Voc insight

VOC Insight / 客户之声洞察

Treat VOC work as evidence synthesis, not just summarization. 将 VOC 工作视为“证据综合”，而不只是摘要。Preserve the user's voice while turning messy feedback into decisions a product, growth, customer success, or operations team can act on.

Default to the user's language. If the request or data is Chinese, respond in Chinese; if it is English, respond in English; if it is mixed, produce bilingual headings or a bilingual summary when useful.

Workflow / 工作流

1. Clarify the business objective / 明确业务目标:
   • product improvement / 产品改进
   • churn reduction / 降低流失
   • onboarding optimization / 新手引导优化
   • pricing or packaging feedback / 定价与套餐反馈
   • competitive intelligence / 竞品洞察
   • service quality improvement / 服务质量改进
   • opportunity discovery / 机会点发现
2. Inspect the data shape / 检查数据形态:
   • source type: interviews, surveys, tickets, reviews, social posts, call transcripts, CRM notes, or mixed sources
   • 来源类型：访谈、问卷、客服工单、评论、社媒、通话转写、CRM 记录或混合数据
   • fields: segment, account, product area, channel, date, rating, revenue, region, status, and free text
   • 字段：用户分群、账号、产品模块、渠道、日期、评分、收入、地区、状态、文本内容
3. Normalize only enough to analyze / 只做必要清洗:
   • keep original wording for evidence / 保留原话作为证据
   • merge exact duplicates / 合并完全重复记录
   • keep rare negative or outlier feedback visible / 不因少数或极端反馈而删除
4. Code feedback into themes / 将反馈编码成主题:
   • pain point / 痛点
   • desired outcome / 期望结果
   • blocker or friction / 阻碍与摩擦
   • feature request / 功能诉求
   • sentiment / 情绪倾向
   • journey stage / 旅程阶段
   • severity / 严重程度
   • segment or persona / 分群或用户画像
5. Quantify themes transparently / 透明量化:
   • mention count and share / 提及次数与占比
   • share within segment / 分群内占比
   • links to rating, NPS, churn, revenue, or ticket severity when available / 与评分、NPS、流失、收入或工单等级的关联
6. Separate facts, interpretations, and recommendations / 区分事实、解读和建议:
   • facts: quotes, counts, ratings, timestamps / 事实：原话、数量、评分、时间
   • interpretations: root causes, hypotheses, opportunity sizing / 解读：根因、假设、机会大小
   • recommendations: actions and expected impact / 建议：行动与预期影响
7. Prioritize with a simple frame / 用简单框架排序:
   • user impact / 用户影响
   • business impact / 业务影响
   • frequency / 频次
   • urgency or severity / 紧急度或严重度
   • confidence / 证据可信度
   • estimated effort / 预估成本

Output / 输出格式

Default to a concise report. 默认输出一份简洁报告：

• Executive summary / 核心摘要：3-5 个最重要发现
• Theme table / 主题表：主题、证据数量/占比、影响分群、严重度、代表性原话、含义
• Top pain points / 主要痛点：按优先级列出证据、可能根因和影响
• Opportunities / 产品或流程机会：改什么、为什么现在改、预期收益
• Open questions / 待确认问题：还缺哪些数据
• Recommended actions / 建议行动：立即、短期、长期

When the user asks for product strategy, add opportunity map, persona implications, roadmap candidates, experiment ideas, and success metrics. 当用户要产品策略时，补充机会地图、人群/画像启示、路线图候选、实验想法和成功指标。

When the user asks for customer success or operations, add escalation triggers, response playbooks, retention risks, service process fixes, and knowledge-base gaps. 当用户要客户成功或运营建议时，补充升级触发条件、响应话术/流程、留存风险、服务流程修复和知识库缺口。

Evidence Standards / 证据标准

• Use short representative quotes, not long transcripts. 使用短代表性原话，不粘贴长转写。
• Label low-confidence themes when sample size is small or source bias is high. 样本小或来源偏差高时，标注低置信度。
• Do not infer demographics, intent, or commercial impact unless the data supports it. 不在证据不足时推断人群属性、意图或商业影响。
• Prefer "mentions" over "users" unless each record maps cleanly to one unique user. 除非记录能唯一对应用户，否则用“提及次数”而不是“用户数”。
• State whether percentages are based on coded records, unique users, accounts, or all feedback. 明确百分比的分母：编码记录、唯一用户、账号或全部反馈。
• Keep contradictory feedback visible; tension often signals segmentation or job-to-be-done differences. 保留矛盾反馈，它通常提示分群或使用场景差异。

Reference / 参考

For a deeper bilingual coding rubric, report templates, and priority scoring guidance, read ${CODEBUDDY_SKILL_DIR}/references/voc-framework.md when the task needs a detailed methodology or reusable scoring model. 如需更完整的双语编码表、报告模板和优先级评分方法，读取该参考文件。