Purpose
Merge quantitative signals (analytics, survey stats, funnel metrics) with qualitative evidence (interview quotes, usability observations, open-text feedback) into a single triangulated set of findings — surfacing where the data converges, diverges, or explains each other — and translate every validated pattern into a specific, prioritized design implication a team can act on.
When to Use
Invoke this skill when the user asks to:
- Combine analytics data with user interview or usability findings
- Triangulate survey results against qualitative feedback
- Make sense of a mixed pile of quant and qual research
- Figure out why a metric is moving using qualitative evidence
- Turn research findings into concrete design or product decisions
- Resolve a contradiction between what users say and what the data shows
Inputs Required
Before starting, collect:
- Research question: The primary question or decision this synthesis needs to inform
- Quantitative data: Metrics, funnel numbers, survey stats, feature usage, drop-off rates, satisfaction scores — with sample sizes where available
- Qualitative data: Interview quotes, usability observations, open-text responses, support tickets, session notes
- Segments (if any): User segments, personas, plans, or cohorts the data can be split by
- Study context: How and when the data was collected (recency, method, any known bias)
- Decision at stake: What the team plans to do with these findings (roadmap, redesign, prioritization)
- Confidence tolerance: How much certainty the decision needs (exploratory vs. high-stakes)
If sample sizes or study context are missing, flag confidence limits explicitly in the findings rather than overstating certainty.
Instructions
Step 1: Normalize and Inventory the Evidence
Before finding patterns, organize the raw inputs.
- Tag each data point as QUANT or QUAL, and by the topic/theme it touches (e.g., onboarding, pricing, navigation)
- Note the strength of each quant metric: sample size, statistical reliability, and whether it's a leading or lagging indicator
- Cluster qual observations into themes — group quotes and observations that describe the same underlying behavior or friction
- Produce a short inventory: how many quant metrics, how many qual observations, and how many distinct themes emerged
Step 2: Triangulate — Map Quant Against Qual per Theme
For each theme, classify the relationship between the numbers and the stories into one of four types:
- Convergence — Quant and qual point the same direction. High confidence. Example: 42% drop-off at step 3 (quant) + five users saying "I didn't know what to enter here" (qual).
- Explanation — Qual explains the why behind a quant signal. The metric tells you what; the observation tells you the cause.
- Contradiction — Quant and qual disagree. This is the most valuable and most dangerous zone. Resolve by checking: Is one sample biased? Is there a segment split? Are they measuring different moments in the journey?
- Gap — A signal exists in one dataset with no counterpart in the other. Flag as needing follow-up research; do not treat as validated.
For each theme, state the relationship type explicitly and cite the specific evidence on both sides.
Step 3: Resolve Contradictions and Segment Splits
When quant and qual disagree:
- Check for a hidden segment: A metric averaged across all users often masks two opposing behaviors. Split by the available segments and re-examine.
- Check for the say/do gap: Stated preference (qual) often diverges from actual behavior (quant). Weight behavioral data higher for what users do; weight qual higher for why.
- Check sample validity: A vocal 5 interviewees can contradict a statistically significant 5,000-user metric — note which carries more weight for this specific question.
- Document the resolution and the residual uncertainty.
Step 4: Rank Findings by Confidence and Impact
Score each validated finding on two axes:
- Confidence (High / Medium / Low): based on convergence, sample size, and method quality
- Impact (High / Medium / Low): based on how many users affected × severity × relevance to the decision at stake
Rank findings High-confidence + High-impact first. Findings that are Low-confidence but High-impact become research recommendations, not decisions.
Step 5: Translate Each Finding into a Design Implication
For every ranked finding, produce an actionable implication with this structure:
- Finding: One-sentence statement of the validated pattern
- Evidence: The specific quant + qual data supporting it (with numbers and a representative quote)
- Relationship type: Convergence / Explanation / Contradiction resolved / Gap
- Design implication: The concrete change or direction this suggests — a specific hypothesis, not "improve onboarding"
- Confidence: High / Medium / Low, with the reason
- Suggested next step: Ship-and-measure, prototype-and-test, or investigate-further
Step 6: Identify Open Questions and Research Gaps
List what remains unresolved — the gaps from Step 2 and the low-confidence findings — and recommend the specific follow-up method (analytics query, targeted interview, A/B test) that would close each gap.
Output Format
- Structure: Evidence inventory → triangulation matrix by theme → ranked findings with design implications → open questions
- Length: 1,000–1,400 words
- Format: Structured document; triangulation summary as a table (Theme | Quant Signal | Qual Signal | Relationship | Confidence); findings as structured blocks
- Sections: Evidence Inventory → Triangulation Matrix → Ranked Findings & Design Implications → Open Questions & Recommended Research
- Tone: Rigorous and evidence-first — a senior researcher who cites data, resists overclaiming, and never turns a single quote into a mandate
Quality Checklist
Before delivering output, verify:
- [ ] Every finding cites both a quantitative and qualitative data point, or explicitly notes when only one exists
- [ ] Each theme in the triangulation matrix is labeled with one of the four relationship types (Convergence / Explanation / Contradiction / Gap)
- [ ] Every contradiction includes an attempted resolution (segment split, say/do gap, or sample weighting) — not left unexplained
- [ ] Findings are ranked by confidence and impact — not listed in input order
- [ ] Each design implication is specific and testable — no "improve X" placeholders
- [ ] Confidence levels reflect actual sample sizes and method quality, not enthusiasm for the finding
Notes
- The say/do gap is the most common source of quant-qual contradiction: users report what they wish were true; behavioral metrics report what actually happened. When in doubt about behavior, trust the numbers; when in doubt about motivation, trust the interviews.
- A statistically significant metric with no qualitative explanation is a signal to go interview — a strong number you can't explain is a risk, not a decision.
- Resist the pull to let one memorable quote override the data. A vivid anecdote and a representative pattern are not the same thing; label which is which.
- Contradictions are the highest-value output of this synthesis — they almost always reveal a hidden segment or a measurement mismatch worth surfacing to the team.
How to use this skill
This file is a Claude Code skill in Anthropic's standard format — markdown with YAML frontmatter at the top.
Installing in Claude Code:
~/.claude/skills/<folder-name>/SKILL.md— auto-loads in every session anywhere on your machine (recommended for skills you'll use across multiple projects)<your-project>/.claude/skills/<folder-name>/SKILL.md— auto-loads only when working in that project (recommended for project-specific skills)
Restart Claude Code or open a new session to register.
Using in Claude.ai, ChatGPT, Gemini, or other AI tools:
- Drag this file into a new chat, or copy and paste its contents
- Send your task or request as your next message — the AI will apply the skill where it fits
For Claude Code skill details: https://docs.anthropic.com/en/docs/claude-code/skills