analyzing-user-feedback
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChineseAnalyzing User Feedback
用户反馈分析
Scope
适用范围
Covers
- Aggregating and normalizing feedback from multiple channels (support, sales, research, reviews, surveys, usage signals)
- Turning raw feedback into themes with evidence and actionable recommendations
- Identifying friction / reasons users won’t use the product (not just validation)
- Producing a repeatable feedback loop (cadence, owners, and handoffs)
When to use
- “Synthesize our user feedback into themes and actions.”
- “Analyze support tickets / feature requests for the top issues.”
- “Create a voice-of-customer report for <area> in the last <time window>.”
- “Summarize churn reasons / cancellation feedback.”
- “Cluster survey open-ends into insights and recommendations.”
When NOT to use
- You need to collect new feedback first (use /
conducting-user-interviews)designing-surveys - You need backlog prioritization as the primary output (use )
prioritizing-roadmap - You need a PRD/spec for a chosen solution (use /
writing-prds)writing-specs-designs - You only need to respond to individual tickets (support workflow, not synthesis)
涵盖内容
- 汇总并标准化来自多渠道的反馈(支持、销售、调研、评论、问卷、使用数据信号)
- 将原始反馈转化为带佐证的主题和可落地的建议
- 识别用户不愿使用产品的痛点/原因(而非仅验证已有假设)
- 建立可重复的反馈闭环(周期、负责人及交接流程)
适用场景
- “将我们的用户反馈汇总为主题和行动项。”
- “分析支持工单/功能请求,找出核心问题。”
- “针对<业务领域>过去<时间范围>的情况,生成客户声音报告。”
- “总结用户流失原因/取消服务的反馈。”
- “将开放式调研回复聚类为洞见和建议。”
不适用场景
- 你需要先收集新的反馈(请使用/
conducting-user-interviews)designing-surveys - 你需要将待办事项优先级作为主要输出(请使用)
prioritizing-roadmap - 你需要为选定的解决方案撰写PRD/规格文档(请使用/
writing-prds)writing-specs-designs - 你仅需回复单个工单(属于支持工作流,而非汇总分析)
Inputs
输入要求
Minimum required
- Product area / workflow to analyze (or “all product”)
- Time window + volume expectations (e.g., “last 90 days”, “~2k tickets”)
- Feedback sources available (tickets, interviews, sales notes, reviews, surveys, community, logs)
- The decision this analysis should inform (roadmap theme, launch readiness, onboarding fixes, messaging, quality)
- Any segmentation that matters (ICP, persona, plan tier, lifecycle stage)
- Constraints: privacy/PII rules, internal-only vs shareable, deadline/time box
Missing-info strategy
- Ask up to 5 questions from references/INTAKE.md.
- If data access is limited, proceed using a small representative sample and label confidence/limitations.
- Do not request secrets. If feedback contains PII, ask for redacted excerpts or aggregated fields only.
最低要求输入
- 待分析的产品领域/工作流(或“全产品”)
- 时间范围及反馈量预期(例如:“过去90天”、“约2000张工单”)
- 可用的反馈来源(工单、访谈记录、销售笔记、评论、问卷、社区内容、日志)
- 本次分析需支撑的决策(路线图主题、发布准备情况、新手引导优化、消息推送、产品质量)
- 任何重要的细分维度(核心目标客户、用户画像、套餐层级、生命周期阶段)
- 约束条件:隐私/PII规则、内部专属 vs 可共享、截止时间/时间限制
信息缺失处理策略
- 可从references/INTAKE.md中选取最多5个问题进行询问。
- 如果数据访问受限,可使用小样本代表性数据推进分析,并标注可信度/局限性。
- 不得索要机密信息。若反馈包含PII,请求提供脱敏片段或仅提供汇总字段。
Outputs (deliverables)
输出成果(交付物)
Produce a User Feedback Analysis Pack in Markdown (in-chat; or as files if requested):
- Context snapshot (scope, decision, time window, segments, constraints)
- Source inventory + sampling plan (what’s included/excluded; why)
- Taxonomy + codebook (tags, definitions, and coding rules)
- Normalized feedback table (tagged items; links/IDs if available; no PII)
- Themes & evidence report (top themes, representative quotes, frequency/severity, confidence)
- Recommendations (actions, owners/time horizon if known, expected impact, open research questions)
- Feedback loop plan (cadence, stakeholders, how engineering participates, how insights are stored)
- Risks / Open questions / Next steps (always included)
Templates: references/TEMPLATES.md
生成Markdown格式的用户反馈分析包(可在对话中直接发送;若有需求也可作为文件发送):
- 上下文快照(范围、决策、时间范围、细分维度、约束条件)
- 来源清单及抽样方案(包含/排除的内容及原因)
- 分类体系及编码手册(标签、定义及编码规则)
- 标准化反馈表格(已标记的反馈项;若有则附带链接/ID;无PII内容)
- 主题及佐证报告(核心主题、代表性引用、频率/影响程度、可信度)
- 建议(行动项、已知的负责人/时间周期、预期影响、待调研问题)
- 反馈闭环计划(周期、相关利益方、工程师参与方式、洞见存储方式)
- 风险、待解决问题及后续步骤(必须包含)
模板参考:references/TEMPLATES.md
Workflow (8 steps)
工作流程(8个步骤)
1) Intake + decision framing
1) 需求收集与决策框架搭建
- Inputs: User context; references/INTAKE.md.
- Actions: Confirm the decision, scope, time window, audience, and constraints. Define what “good” looks like.
- Outputs: Context snapshot.
- Checks: A stakeholder can answer: “What decision will this analysis change?”
- 输入: 用户提供的上下文;references/INTAKE.md。
- 行动: 确认决策方向、范围、时间范围、受众及约束条件。定义“合格成果”的标准。
- 输出: 上下文快照。
- 检查项: 相关利益方可明确回答:“本次分析会影响哪项决策?”
2) Inventory sources + define the sampling plan
2) 梳理来源并制定抽样方案
- Inputs: List of sources + access constraints.
- Actions: Create a source inventory, decide inclusions/exclusions, and pick a sample strategy (random, stratified, top-volume buckets).
- Outputs: Source inventory + sampling plan.
- Checks: Sampling plan covers the highest-volume and highest-risk segments (or explicitly explains why not).
- 输入: 反馈来源列表及访问约束。
- 行动: 生成来源清单,确定包含/排除的内容,并选择抽样策略(随机抽样、分层抽样、高容量分组抽样)。
- 输出: 来源清单及抽样方案。
- 检查项: 抽样方案覆盖了高容量及高风险的细分群体(若未覆盖需明确说明原因)。
3) First-pass read-through (open coding)
3) 初次通读与开放式编码
- Inputs: Sampled feedback items.
- Actions: Read/annotate items manually to surface what’s “wrong” and why users struggle or churn. Write raw notes before building categories.
- Outputs: Initial codes/notes + candidate themes list.
- Checks: Notes capture rejection reasons and friction, not just feature ideas.
- 输入: 抽样后的反馈项。
- 行动: 手动阅读/标注反馈项,找出用户的痛点及流失原因。在建立分类前先记录原始笔记。
- 输出: 初始编码/笔记及候选主题列表。
- 检查项: 笔记记录了用户拒绝使用的原因和痛点,而非仅功能想法。
4) Build the taxonomy + codebook
4) 构建分类体系及编码手册
- Inputs: Initial codes; product context.
- Actions: Define a tagging schema (topic, lifecycle stage, severity, user segment, root cause, sentiment). Write clear tag definitions and rules.
- Outputs: Taxonomy + codebook.
- Checks: Two people could tag the same item similarly using the codebook.
- 输入: 初始编码;产品上下文。
- 行动: 定义标签体系(主题、生命周期阶段、影响程度、用户细分、根本原因、情感倾向)。撰写清晰的标签定义及编码规则。
- 输出: 分类体系及编码手册。
- 检查项: 不同人员使用该编码手册可对同一反馈项做出相似的标记。
5) Normalize and tag the feedback table
5) 标准化并标记反馈表格
- Inputs: Raw items; taxonomy/codebook.
- Actions: Create a normalized table, tag each item, and capture evidence fields (source, date, segment, verbatim excerpt, link/ID).
- Outputs: Normalized feedback table (tagged).
- Checks: No PII; every row has at least 1 primary theme tag + a severity/impact signal.
- 输入: 原始反馈项;分类体系/编码手册。
- 行动: 生成标准化表格,标记每个反馈项,并记录佐证字段(来源、日期、细分群体、原文摘录、链接/ID)。
- 输出: 已标记的标准化反馈表格。
- 检查项: 无PII内容;每一行至少包含1个核心主题标签及1个影响程度信号。
6) Synthesize themes + quantify carefully
6) 汇总主题并谨慎量化
- Inputs: Tagged table.
- Actions: Summarize top themes, quantify frequency by segment/source, identify severity and “why it happens”, and call out unknowns/bias.
- Outputs: Themes & evidence report with confidence levels.
- Checks: Each theme includes representative evidence (quotes/examples) and is not purely speculative.
- 输入: 已标记的表格。
- 行动: 总结核心主题,按细分群体/来源量化频率,识别影响程度及“问题产生原因”,并指出未知项/偏差。
- 输出: 带可信度等级的主题及佐证报告。
- 检查项: 每个主题都包含代表性佐证(引用/示例),而非纯推测。
7) Translate into actions + learning plan
7) 转化为行动项及学习计划
- Inputs: Themes report; constraints.
- Actions: Convert themes into actions (bugs, UX fixes, comms, product bets) and open questions (what to research next). Tie each action to evidence and expected impact.
- Outputs: Recommendations + learning plan.
- Checks: Recommendations are concrete enough to execute next sprint/quarter (clear owner/time horizon if known).
- 输入: 主题报告;约束条件。
- 行动: 将主题转化为行动项( bug修复、UX优化、沟通策略、产品尝试)及待调研问题。将每个行动项与佐证及预期影响关联。
- 输出: 建议及学习计划。
- 检查项: 建议足够具体,可用于下一个迭代/季度的执行(若已知则明确负责人/时间周期)。
8) Share out + establish the feedback loop + quality gate
8) 分享成果 + 建立反馈闭环 + 质量校验
- Inputs: Draft pack.
- Actions: Propose the share-out format (doc + review). Define cadence, owners, and storage (where insights live). Run references/CHECKLISTS.md and score with references/RUBRIC.md. Add Risks/Open questions/Next steps.
- Outputs: Final User Feedback Analysis Pack.
- Checks: Pack is shareable as-is; limitations are explicit; follow-up actions are scheduled.
- 输入: 分析包草稿。
- 行动: 建议分享格式(文档+评审)。定义周期、负责人及存储方式(洞见的存放位置)。使用references/CHECKLISTS.md进行检查,并通过references/RUBRIC.md进行评分。添加风险、待解决问题及后续步骤。
- 输出: 最终版用户反馈分析包。
- 检查项: 分析包可直接共享;局限性已明确说明;后续行动项已安排。
Quality gate (required)
质量校验(必填)
- Use references/CHECKLISTS.md and references/RUBRIC.md.
- Always include: Risks, Open questions, Next steps.
- 使用references/CHECKLISTS.md和references/RUBRIC.md进行校验。
- 必须包含:风险、待解决问题、后续步骤。
Examples
示例
Example 1 (support tickets): “Analyze the last 60 days of onboarding-related tickets. Output a User Feedback Analysis Pack and top 10 recommended fixes.”
Expected: source inventory + sampling, taxonomy, tagged table, themes with quotes, and ranked actions.
Expected: source inventory + sampling, taxonomy, tagged table, themes with quotes, and ranked actions.
Example 2 (survey + reviews): “Synthesize survey open-ends and app store reviews for our new pricing change. What are the biggest friction points and why?”
Expected: themes split by source/segment, severity signals, and recommendations (incl. messaging/UX changes).
Expected: themes split by source/segment, severity signals, and recommendations (incl. messaging/UX changes).
Boundary example: “Read all our feedback and tell us what to build next.”
Response: ask for scope/time window/decision + a sample dataset; otherwise produce a sampling plan + a minimal first-pass synthesis with explicit limitations.
Response: ask for scope/time window/decision + a sample dataset; otherwise produce a sampling plan + a minimal first-pass synthesis with explicit limitations.
示例1(支持工单): “分析过去60天内与新手引导相关的工单。输出用户反馈分析包及排名前10的修复建议。”
预期输出:来源清单及抽样方案、分类体系、已标记表格、带引用的主题报告,以及排序后的行动项。
预期输出:来源清单及抽样方案、分类体系、已标记表格、带引用的主题报告,以及排序后的行动项。
示例2(调研+评论): “汇总关于新定价调整的开放式调研回复及应用商店评论。找出最大的痛点及原因。”
预期输出:按来源/细分群体划分的主题、影响程度信号,以及建议(包含沟通策略/UX优化)。
预期输出:按来源/细分群体划分的主题、影响程度信号,以及建议(包含沟通策略/UX优化)。
边界示例: “阅读所有反馈并告诉我们接下来要开发什么。”
回复:询问范围/时间范围/决策方向及样本数据集;若无法获取则生成抽样方案及带明确局限性的初步汇总分析。
回复:询问范围/时间范围/决策方向及样本数据集;若无法获取则生成抽样方案及带明确局限性的初步汇总分析。