Community

survey-analysis

Open Agent Skill · research · works with Claude Code, Codex, Cursor, and compatible agents

Analyze survey results into themes, segments, and decision-ready findings. Use whenever the user shares survey responses, NPS or CSAT comments, form exports, or asks what the survey says — with counts and caveats rather than vibes.

View on GitHub

Survey Analysis

Surveys mix quantitative items and open text. Report n, base rates, and selection bias. Do not overclaim causality.

Workflow

  1. Clarify decision the survey should inform.
  2. Profile sample: n, response rate if known, segments.
  3. Quant: distributions, top-box, crosstabs with minimum cell sizes.
  4. Qual: code open ends into themes with counts.
  5. Insights + recommendations with confidence.
  6. Limitations: bias, wording, seasonality.

Output format

## Survey readout: <name>
**Sample:** n=…

### Key findings
1. …

### Quant summary
tables

### Themes from open ends
| Theme | Mentions | Example |

### Recommendations
…

### Limitations
…

Rules

  1. Always show denominators.
  2. Avoid percent of tiny bases without warning.
  3. Do not invent responses.
  4. Separate correlation from "users want X because".
  5. Protect respondent privacy in examples.
  6. NPS: document calculation if used.

Edge cases

  • Low n: directional only.
  • Multi-select: clarify base.
  • Translated responses: caution on nuance.