| name | survey-analysis |
| description | Analyzes survey results with statistical summaries, cross-tabulations, trend identification, and actionable insight recommendations. Use when interpreting survey data for decisions. |
| allowed-tools | Read Write Glob |
| metadata | {"author":"matthewhitcham","version":"1.0"} |
Survey Analysis
When to Use This Skill
Use this skill when you need to:
- Analyze results from a customer, employee, or market research survey
- Generate statistical summaries and identify meaningful patterns
- Create cross-tabulations to find segment differences
- Turn raw survey data into actionable business recommendations
DO NOT use this skill for designing surveys, running statistical software, or academic research analysis. This is for practical business survey interpretation.
Core Principle
SURVEY DATA IS ONLY VALUABLE WHEN IT LEADS TO A DECISION — EVERY ANALYSIS MUST END WITH "HERE IS WHAT TO DO ABOUT IT."
Phase 1: Brief
Required Inputs
| Input | What to Ask | Default |
|---|
| Survey type | "What kind of survey? (customer satisfaction, NPS, market research, employee engagement)" | Customer satisfaction |
| Response count | "How many responses did you receive?" | Must be provided |
| Key questions | "What business questions should the analysis answer?" | Must be provided |
| Data format | "How is the data stored? (CSV, Google Sheets, raw platform export)" | Spreadsheet |
| Segments | "Any segments to compare? (customer type, plan tier, demographics)" | None — overall analysis |
| Prior benchmarks | "Do you have previous survey results or industry benchmarks to compare against?" | No prior data |
GATE: Confirm brief and review the data before proceeding.
Phase 2: Analyze
Analysis Framework
- Response overview — total responses, response rate, completion rate
- Summary statistics — means, medians, distributions for quantitative questions
- Frequency analysis — response counts and percentages for categorical questions
- Cross-tabulations — compare responses across key segments
- Trend identification — patterns, outliers, and notable clusters
- Open-ended theming — categorize free-text responses into themes
Statistical Guidelines
- Report response rate and note if it is below 20% (potential non-response bias)
- Use median for skewed distributions, mean for normal distributions
- Note sample sizes per segment — do not draw conclusions from groups under 30
- Flag statistically small differences — a 2% gap with 50 responses is noise
GATE: Present preliminary findings and confirm focus areas before building the full report.
Phase 3: Build
Deliverables
1. Executive Summary (1 page)
- Top 3-5 findings in plain language
- Key metric headline numbers
- Recommended actions
2. Detailed Analysis Report
- Section for each survey question or question group
- Visualizations: bar charts for categorical, distribution plots for scale questions
- Cross-tabulation tables for segment comparisons
- Open-ended response themes with representative quotes
3. Insight-Action Map
| Finding | Implication | Recommended Action | Priority |
|---|
| NPS dropped 15 points | Customer sentiment declining | Investigate top complaints, prioritize fixes | High |
4. Raw Data Summary Tables
- Clean tables with all response data aggregated
- Ready for copy-paste into presentations or reports
Phase 4: Polish
Presentation Package
Format the key findings for sharing:
- 5-slide summary deck structure (overview, top findings, segment insights, recommendations, next steps)
- Talking points for each slide
- Anticipated questions and answers
Follow-Up Recommendations
- What to investigate further based on findings
- Suggested follow-up survey questions for the next cycle
- Recommended timeline for the next survey
Example 1: NPS Survey (200 responses, SaaS product)
Key outputs: NPS score with promoter/passive/detractor breakdown, NPS by customer segment, top 5 themes from detractor comments, 3 priority actions to improve score.
Example 2: Post-Purchase Survey (500 responses, e-commerce)
Key outputs: Satisfaction score, top reasons for purchase, product quality ratings, delivery experience ratings, repurchase likelihood, open-ended improvement suggestions themed and ranked.
Anti-Patterns
- Reporting without interpreting — "42% said yes" is data, not analysis. Always add "which means..." and "therefore we should..."
- Cherry-picking results — reporting only favorable findings undermines trust. Include the bad news with recommendations.
- Over-interpreting small samples — 10 responses from enterprise customers is not enough to draw segment conclusions. State limitations.
- Ignoring open-ended responses — free-text answers often contain the most actionable insights. Always theme and analyze them.
- Analysis paralysis — a 50-page report nobody reads is worse than a 2-page summary with clear actions.
Recovery
- Low response rate: Acknowledge the limitation upfront. Focus on directional insights rather than precise percentages.
- Contradictory results: Segment the data — contradictions often resolve when you split by customer type or demographics.
- No actionable findings: Reframe the analysis around the original business questions. If the survey did not answer them, recommend better questions for next time.
- Data quality issues: Flag invalid or suspicious responses (all same answers, impossible combinations). Clean before analyzing.
