Synthesize a body of user research — interviews, surveys, usability tests, NPS responses, support tickets, or mixed sources — into structured insights, JTBD themes, segment patterns, and recommended actions. Use when someone has completed N interviews or studies and needs to make sense of them, asks to synthesize research findings, summarize user interviews, identify themes from feedback, or turn raw research into product decisions. Different from summarize-interview (single session) — this skill handles a corpus of research and produces insights at the strategic level. Commands: /synthesize-research, /research-synthesis
From pm-market-researchnpx claudepluginhub jupitermoney/pm-superic-skills --plugin pm-market-researchThis skill uses the workspace's default tool permissions.
Guides Next.js Cache Components and Partial Prerendering (PPR) with cacheComponents enabled. Implements 'use cache', cacheLife(), cacheTag(), revalidateTag(), static/dynamic optimization, and cache debugging.
Migrates code, prompts, and API calls from Claude Sonnet 4.0/4.5 or Opus 4.1 to Opus 4.5, updating model strings on Anthropic, AWS, GCP, Azure platforms.
Details PluginEval's skill quality evaluation: 3 layers (static, LLM judge), 10 dimensions, rubrics, formulas, anti-patterns, badges. Use to interpret scores, improve triggering, calibrate thresholds.
This skill transforms a body of raw research into structured product insight. It handles multiple sources, surfaces patterns across them, and produces outputs that drive decisions — not just documentation.
The key distinction: summarize-interview handles one session.
This skill handles the whole study.
Read the full file before producing anything.
Before synthesising anything, gather:
If the user pastes transcripts, notes, or data directly — process them. If they describe the research — ask for the raw material before synthesising.
Before pattern-finding, extract atomic observations across all sources. Do not cluster yet.
For each source, note:
Present this as a raw signal table only if the user asks to see it. Otherwise use it internally to power the synthesis.
Group signals into Jobs to Be Done patterns. A JTBD theme is:
"When [situation], users want to [motivation], so they can [outcome]."
Surface 3–6 themes maximum. More than 6 means you haven't clustered tightly enough.
For each theme:
Theme [N]: [Short label]
JTBD: "When [situation], users want to [motivation], so they can [outcome]."
Evidence:
Frequency: High / Medium / Low (across participants) Intensity: High / Medium / Low (how strongly felt) Satisfaction gap: [Are users getting this job done? If not, what's failing?]
If the research spans multiple user types, identify whether the themes hold uniformly or diverge by segment.
Useful segments to check:
Present segment findings as a matrix only where meaningful divergence exists. Don't force segmentation if the patterns are consistent.
| Theme | New Users | Power Users | Churned Users |
|---|---|---|---|
| [Theme] | Strong signal | Absent | Dominant |
Interpretation: [One sentence on what the divergence means for product decisions]
Separate the signals that indicate what's working from what's failing.
What users celebrate (keep and amplify):
What users tolerate (fix before it becomes a churn driver):
What users abandon or avoid (urgent to address):
What users ask for (feature or experience gaps):
Be transparent about the quality of the conclusions.
| Theme | Confidence | Basis |
|---|---|---|
| [Theme 1] | High — 8/10 participants, consistent across segments | ... |
| [Theme 2] | Medium — 4/10, diverges by user type | ... |
| [Theme 3] | Low — 2/10, worth monitoring but not acting on yet | ... |
Low-confidence themes are still worth surfacing — they may become high- confidence with more data. Flag what additional research would confirm or invalidate them.
Translate themes into product decisions. Each recommendation must be traceable to at least one theme.
Format:
Recommendation [N]: [Action label] Driven by: [Theme(s) it addresses] Priority: High / Medium / Low Suggested next step: [Experiment, PRD, user test, deeper research] Success metric: [How you'd know this worked]
Keep recommendations specific and actionable. "Improve onboarding" is not a recommendation. "Reduce time to first value in onboarding by adding a personalisation step at step 2 based on stated job role — target 30% reduction in drop-off at step 3" is a recommendation.
State clearly what this research cannot answer.
What we still don't know:
Recommended next research:
Decision readiness: [Can the decision the user named in Step 1 be made now, or does it need more data? State clearly.]
Ask the user which format they need if not specified:
Anchor every claim in evidence. No unattributed insights. Every theme needs at least 2 supporting signals.
Separate observation from interpretation. Be clear about what the data says vs what you're inferring. Label inferences as such.
Name the confidence level. Honest research synthesis acknowledges what's directional vs conclusive.
Prioritise by frequency AND intensity. A theme felt strongly by 3 users may outweigh one mentioned casually by 10.
Connect to decisions. Every synthesis output should map to a product or business decision. If a theme doesn't connect to any decision, flag it as "informational only — revisit in future planning."
After completing synthesis:
/write-prd — to turn a top recommendation into a full PRD/prioritize-features — to stack-rank the feature gaps surfaced/interview — to design follow-up research for low-confidence themes/stakeholder-update — to communicate the research findings to leadershipTrigger with data: "Here are 8 interview transcripts from churned users — can you synthesise the key themes and tell me what we should do?"
Trigger with described study: "We ran 12 usability tests on our onboarding flow and 200 NPS survey responses last month. I need to pull out the key insights for our Q2 planning session."
Trigger needing clarification: "Can you help me make sense of my user research?" → Respond: "Yes — can you share the raw material? Transcripts, notes, survey responses, or any format works. Also, what decision are you trying to inform with this research?"