Purpose
This skill governs how raw research inputs — user feedback, qualitative interviews, market signals, and product observations — are transformed into structured, decision-ready artifacts. It prevents common synthesis failures: insight dilution, false clustering, missing evidence links, and opportunity statements without business grounding.
Apply this skill whenever research data needs to be organized, themed, prioritized, or written up for product, design, strategy, or stakeholder audiences.
Source Types & How to Handle Them
Each research input type requires a different synthesis lens. Identify which type(s) are present before starting.
User Feedback (written, in-app, survey responses)
Extract verbatim quotes that carry signal (frustration, confusion, delight, workarounds).
Tag each quote with: source channel, sentiment (negative / neutral / positive), and affected feature area.
Cluster by recurring theme — not by feature or product area, but by underlying user need or failure mode.
Minimum cluster threshold: 3 independent data points before calling it a pattern.
Qualitative Interviews (UX research, sales calls, customer success, onboarding)
Capture: stated goal, actual behavior, friction point, workaround used, emotional signal.
Distinguish between what users say they want vs. what their behavior reveals they need.
Flag contradictions across participants — these are often the most valuable signals.
Attribute insights to participant role/segment, not individual names (privacy-safe labeling: e.g., "Enterprise user, logistics sector, 3 years tenure").
Market Observations (competitive intel, analyst reports, trend signals, social listening)
Summarize the observation and its source credibility (primary vs. secondary research).
Assess relevance: is this an emerging pattern, a validated trend, or noise?
Connect to internal product context: does this validate, challenge, or expand current direction?
Avoid presenting market observations as user needs — they require separate triangulation.
Product Opportunity Inputs (internal hypotheses, stakeholder requests, roadmap gaps)
Reframe each input as a problem statement, not a solution: "Users cannot X, which leads to Y" — not "We should build Z."
Validate against research evidence before elevating to opportunity status.
Flag if an opportunity lacks corroborating user or market evidence.
Synthesis Process
Follow this sequence. Skip steps only if the research scope explicitly doesn't include that input type.
Step 1 — Intake & Source Inventory
List all available inputs: source type, volume (e.g., 12 interview transcripts, 340 feedback tickets), timeframe, and population covered. State what's missing or unknown.
Step 2 — Atomic Insight Extraction
Break each source into its smallest meaningful units — one observation per insight card. Each card contains:
Raw evidence (quote, paraphrase, or observed behavior)
Source label (anonymized)
Input type
Preliminary tag (theme candidate)
Step 3 — Affinity Clustering
Group insight cards by underlying need, failure mode, or job-to-be-done — not by UI element or feature. Name each cluster with a verb-led insight statement: "Users struggle to reconcile data from multiple sources without leaving the platform." Avoid cluster names like "Dashboard issues" or "Onboarding feedback."
Cluster quality checks:
Does the cluster have evidence from at least 2 different source types? (Stronger signal)
Are there outlier insights that don't fit any cluster? (Preserve them — don't force-fit)
Are any clusters so large they should be split? (Target: 3–7 insights per cluster)
Step 4 — Pattern Validation
For each cluster, assess:
Frequency: how many data points support it?
Severity: how much does this problem cost users (time, error, churn risk)?
Breadth: which user segments are affected?
Strategic fit: does this align with current product direction or reveal a blind spot?
Step 5 — Opportunity Framing
Convert validated clusters into opportunity statements using this structure:
Opportunity: [User segment] needs a way to [achieve goal] without [current friction/limitation], because [underlying reason this matters].
Include: supporting evidence count, affected segments, severity score (High / Medium / Low), and open questions requiring further research.
Step 6 — Synthesis Narrative
Write a short narrative (150–300 words) that connects the top 3–5 findings into a coherent story. Answer: What is the user most struggling with? Where does the market validate or diverge? What should product teams prioritize investigating next?
Output Formats
Choose the format based on the audience and use case.
A — Insight Summary (for team sharing, Slack, async reads)
## Research Snapshot — [Topic / Sprint / Date]
**Sources:** [List input types and counts]
**Period:** [Data collection timeframe]
**Population:** [Segments covered]
### Top Findings
1. [Insight statement] — supported by [N] data points across [source types]
2. ...
3. ...
### Key Quote
> "[Verbatim quote]" — [Anonymized participant label]
### Open Questions
- [Question requiring further investigation]
B — Opportunity Brief (for product/strategy decisions)
## Opportunity Brief — [Opportunity Title]
**Problem Statement:** [Verb-led, user-centered problem framing]
**Affected Segments:** [Roles, tiers, or contexts]
**Evidence Base:** [N interviews, N feedback items, N market signals]
**Severity:** High / Medium / Low
**Strategic Fit:** [Aligned with / challenges / expands current direction]
### Supporting Evidence
- [Quote or behavioral observation] — [Source label]
- ...
### Market Context
[1–2 sentences on relevant market signal, if available]
### Recommended Next Steps
- [Research action or validation experiment]
- [Stakeholder to involve]
### Open Questions
- [Unresolved uncertainty]
C — Full Synthesis Report (for quarterly reviews, leadership readouts, major discovery cycles)
Structure:
Executive Summary (150 words max — findings, implications, recommended actions)
Research Scope & Methodology
Key Themes (one section per cluster, with evidence table)
Product Opportunity Landscape (opportunity briefs, prioritized)
Market & Competitive Context
Limitations & Confidence Level
Recommended Next Steps
Use create_text_document for Format A and B. Use create_html_document for Format C when a designed, stakeholder-ready presentation is needed.
Clarification — When to Ask
Ask before starting only when:
No research inputs have been provided → "What research materials are you working with? (e.g., interview transcripts, feedback exports, survey data, competitive notes)"
Audience is completely unknown → "Who will consume this synthesis — product team, leadership, external stakeholders?"
Output format is ambiguous for a full-cycle synthesis → "Should this be a quick insight summary, an opportunity brief, or a full synthesis report?"
Never ask about:
How many clusters to create — determine this from the data; explain your reasoning
Whether to include quotes — always include at least one verbatim quote per major finding
Whether to flag gaps — always flag what's missing; incomplete evidence is a synthesis finding
Quality Rules
Apply these checks before finalizing any synthesis output:
No unsupported claims: every insight must trace back to at least one named evidence source.
No solutionism: synthesis surfaces problems and opportunities, not features or roadmap items.
No false precision: do not assign percentages or quantitative weights to qualitative data unless the underlying sample size justifies it.
Preserve dissenting signals: if evidence contradicts the dominant pattern, include it — do not suppress minority findings.
Segment specificity: state which user segment each finding applies to; avoid "users generally feel…" without qualification.
Confidence calibration: label each finding as Confirmed (3+ corroborating sources), Indicative (2 sources), or Preliminary (1 source, needs validation).
Tone & Language
Write for decision-makers who will act on findings — be direct, specific, and evidence-led.
Use active constructions: "Participants abandoned the flow when..." not "It was observed that flows were abandoned..."
Avoid hedging that dilutes signal: not "users might possibly struggle" — use "users consistently failed to complete X (confirmed across 7 interviews)."
German-language outputs follow the same structure — apply Tone of Voice Skill rules for register and phrasing when writing for German-speaking stakeholders.