AgentSkillsCN

extracting-insights-from-feedback

对非结构化反馈数据进行深入分析，挖掘其中的规律、情感倾向，以及可付诸实践的洞见。适用于批量处理支持工单、应用评价、NPS评论、社交媒体提及，或其他大规模客户反馈时使用。

SKILL.md
--- frontmatter
name: extracting-insights-from-feedback
description: Analyzes unstructured feedback data to identify patterns, sentiment, and actionable insights. Use when processing support tickets, app reviews, NPS comments, social mentions, or other customer feedback at scale.

Extracting Insights from Feedback

Quick start

Collect or infer:

  • •Feedback source(s) (reviews, tickets, NPS, social, etc.)
  • •Volume and time range
  • •Current categorization (if any)
  • •Business questions to answer
  • •Stakeholder needs (product, support, marketing)

Then produce output using TEMPLATES.md. Validate with RUBRIC.md.

Workflow

  1. •Define the business questions: What decisions will this analysis inform?
  2. •Sample or review full dataset depending on volume.
  3. •Develop coding taxonomy: categories, sentiment, severity.
  4. •Code feedback systematically — one pass per dimension.
  5. •Quantify patterns: frequency, trend over time, segment distribution.
  6. •Identify representative quotes for each major pattern.
  7. •Separate signal from noise: prioritize recurring themes over one-off complaints.
  8. •Connect patterns to actionable recommendations.
  9. •Document methodology and limitations.
  10. •Run the rubric check. Revise until it passes.

Degrees of freedom

  • •Low freedom: Accurate representation of feedback, transparent methodology
  • •Medium freedom: Taxonomy design, sentiment classification, quote selection
  • •High freedom: Narrative framing, prioritization, visualization choices

Default: Quantify what can be quantified. Acknowledge limitations of unstructured data. Don't over-interpret sentiment.

References