Use this when you have raw research data — interview transcripts, survey results, support tickets, usability test notes — and need to turn it into themes, patterns, key quotes, backlog connections, and a planning inbox entry.
Process
Step 1: Gather research inputs
Ask the user to provide:
- Raw research data — interview transcripts, survey responses, support tickets, usability test notes, analytics screenshots (paste or describe)
- Research questions — what were you trying to learn?
- Current assumptions — what does the team believe about users right now? What assumptions are you testing?
- Current backlog — (optional) so we can connect insights to existing work
The user can paste raw data in any format. If the data is too long for one message, process it in batches.
Step 2: Surface patterns
Analyze the research and output:
## Research Synthesis — (Topic, date)
### Top Themes
(3-5 themes, ordered by strength of evidence)
**Theme 1: (Theme name)**
- What it means: (One-sentence summary)
- Supporting evidence:
- "(Direct quote from participant)" — Participant N
- "(Direct quote)" — Participant N
- (Pattern observed across N of N participants)
**Theme 2: (Theme name)**
(Same format)
### Points of Agreement
- (What participants consistently said or did)
### Points of Disagreement
- (Where participants diverged — and what might explain the difference)
### Surprises
- (Things that came up that weren't directly asked about — unsolicited insights)
### Unanswered Questions
- (What we still don't know — gaps that need follow-up research)
### Confidence Level
(How confident are we? Based on sample size, consistency of findings, and data quality. Be honest — 3 interviews is exploratory, not conclusive.)
Step 3: Connect to backlog (if backlog was provided)
## Backlog Impact
### Validated Stories
- (Story title) — research confirms the need. Evidence: (brief citation)
### Challenged Stories
- (Story title) — research contradicts or complicates this. Evidence: (brief citation)
### New Stories to Consider
- (Suggested story) — based on (theme/finding). Evidence: (brief citation)
### Priority Implications
(How should this research affect the order of work?)
Step 4: Generate planning inbox entry
If the project uses the two-layer planning system, draft an inbox entry:
## Planning Inbox Entry
### Source
(Where the data came from, date, number of participants/responses)
### Key Findings
- (Most important insight, with supporting evidence)
- (Second insight)
- (Third insight)
### Impact on Current Plan
(What should change in the backlog or plan based on these findings)
### Recommended Actions
- (Specific next step — story to add, priority to shift, follow-up research to plan)
- (Next step)
### Confidence Level
(How confident are we? What would increase confidence?)
Step 5: Research readout (optional)
If the user asks for a stakeholder-facing version, produce a readout in addition to (not instead of) the working synthesis above. The readout is what you present to leadership or clients who need the "so what" without the raw evidence.
## Research Readout — (Topic, date)
### Executive summary
(3-4 sentences: what we studied, what we found, what it means for the product. No jargon. A VP should be able to read this and know what to do.)
### Key findings
**Finding 1: (Finding headline)**
- What we learned: (1-2 sentences)
- Evidence: "(Key quote)" + (supporting data point)
- Implication: (What this means for the product or business)
**Finding 2: (Finding headline)**
(Same format — 3-5 findings total)
### Recommendations
| # | Recommendation | Informed by | Urgency |
|---|---|---|---|
| 1 | (Specific action) | Finding N | Now / Next / Later |
| 2 | | | |
### What we still don't know
- (Open question 1 — and what research would answer it)
- (Open question 2)
### Methodology note
(Sample size, methods used, confidence level. One paragraph. Be honest about limitations.)
Readout rules:
- No more than 2 pages. Shorter is better.
- Lead with implications, not methodology. Stakeholders care about "so what," not "how we did it."
- Every finding must have at least one supporting quote or data point — but only one. Save the full evidence for the working synthesis.
- Recommendations must be specific enough to act on. "Do more research" is not a recommendation. "Run a usability test on the checkout flow to validate finding #2" is.
Step 6: Review and validate
Ask the user:
- Do the themes ring true based on your experience in the research sessions?
- Are the quotes representative, or cherry-picked? (You were in the room — the agent wasn't.)
- Any patterns the agent missed? (Tone, body language, context that doesn't show up in transcripts)
- Is the confidence level honest? (Small samples = low confidence, regardless of how clear the patterns seem.)
Adjust as needed.
Related skills
/opportunity-solution-tree— use after synthesis to map themes to a full opportunity solution tree. Themes become opportunities; supporting evidence carries over directly./interview-plan— if unanswered questions remain after synthesis, plan follow-up interviews to fill evidence gaps./experiment-design— if a theme suggests a testable hypothesis, hand off to experiment design for rigorous test planning.
Output location
Present the synthesis as formatted text in the conversation. If the project uses the two-layer planning system, the user saves the planning inbox entry to planning/inbox/.
Example Output
Input
- Raw research data: 6 user interview transcripts from Meridian Health app users (patients managing chronic conditions). Selected excerpts:
- "I never know if my doctor actually saw my message. I send it and then… nothing. I just wait and hope." — P1
- "I set up the medication reminders but they go off at the wrong times. I take my metformin with dinner, not at 6pm." — P2
- "Honestly I just call the office. The app feels like a black hole." — P3
- "I wish I could see my lab trends over time, not just the last result. Is my A1C getting better or worse? I have no idea." — P4
- "The reminders are useless for me. I turned them off after a week." — P5
- "Sometimes I use the app, sometimes my husband logs in for me. We share a login because setting up two accounts was confusing." — P6
- "I got a reply from my care team three days later. By then I'd already gone to urgent care." — P1
- "My doctor mentioned something in a visit and I wanted to look it up later but there's no notes section anywhere." — P4
- Research questions: Why are patient message response rates declining? Are patients getting value from medication reminders?
- Current assumptions: Patients find the messaging feature useful; low re-engagement is caused by notification fatigue from reminders
- Current backlog: (1) Redesign notification settings UI, (2) Add push notifications for unread messages, (3) Build lab results history view, (4) Caregiver/proxy access feature (deprioritized — low demand assumed)
Output
Research Synthesis — Meridian Health App: Messaging & Reminders, June 2025
Top Themes
Theme 1: Messaging feels like a void — no feedback, no trust
- What it means: Patients send messages but receive no acknowledgment of receipt or expected response time, eroding trust in the channel entirely.
- Supporting evidence:
- "I send it and then… nothing. I just wait and hope." — P1
- "Honestly I just call the office. The app feels like a black hole." — P3
- "I got a reply from my care team three days later. By then I'd already gone to urgent care." — P1
- Pattern: 4 of 6 participants described abandoning messaging in favor of phone calls after a negative wait experience.
Theme 2: Medication reminders are too rigid to fit real life
- What it means: Reminders fire at fixed times that don't match patients' actual medication schedules, making them easy to dismiss or disable entirely.
- Supporting evidence:
- "They go off at the wrong times. I take my metformin with dinner, not at 6pm." — P2
- "I turned them off after a week." — P5
- Pattern: Both participants who mentioned reminders reported disabling them. 0 of 6 participants described reminders as currently useful.
Theme 3: Lab results lack context for self-management
- What it means: Patients see only their most recent result with no trend data, leaving them unable to track whether their condition is improving.
- Supporting evidence:
- "Is my A1C getting better or worse? I have no idea." — P4
- "I wanted to look it up later but there's no notes section anywhere." — P4
- Pattern: Unprompted in 2 of 6 interviews; both from patients managing diabetes.
Theme 4: Shared and proxy access is a real workaround, not an edge case
- What it means: At least one participant is sharing credentials with a caregiver because multi-account setup is too difficult — creating security and usability risk.
- Supporting evidence:
- "We share a login because setting up two accounts was confusing." — P6
- Pattern: Surfaced in 1 of 6 interviews, but notably unprompted.
Points of Agreement
- All 6 participants use the app primarily for messaging and lab results — not for education or wellness content.
- All participants expressed some version of uncertainty about whether their care team had received or read their messages.
- Medication reminders were either disabled or described as low-value by every participant who mentioned them.
Points of Disagreement
- Comfort with digital communication varied: P3 defaults to phone calls regardless of app improvements; P4 and P2 would stay in-app if response visibility improved. Likely explained by age or prior negative experience — worth segmenting in follow-up research.
Surprises
- Caregiver credential-sharing (P6) was entirely unprompted — patients are solving for a missing feature themselves rather than requesting it.
- Patients don't cite notification volume as a problem; they cite notification accuracy. The team's assumption about notification fatigue appears to be off-target.
- Visit note access came up organically (P4) — patients want continuity between in-person visits and the app, which no current backlog item addresses.
Unanswered Questions
- What is the actual average response time for care team messages — and does it vary by clinic or condition?
- How many accounts show signs of credential sharing (same IP, overlapping sessions)?
- Do patients with caregivers have meaningfully different app engagement patterns than solo users?
- What do patients do between sending a message and giving up — do they send follow-ups, or just abandon?
Confidence Level
Moderate-low. 6 interviews with chronic condition patients is exploratory. The messaging and reminder themes are consistent enough to act on directionally. The caregiver finding is a single data point — worth validating before investing in a full proxy-access feature. Lab trends appeared in 2 interviews but both were diabetes patients; may not generalize.
Backlog Impact
Validated Stories
- Lab results history view — research confirms unmet need. P4 explicitly described wanting trend data; the finding emerged unsolicited in both diabetes patient interviews.
Challenged Stories
- Redesign notification settings UI — partially challenged. Patients aren't asking for more control over notification frequency; they need reminders to fire at times they choose, not times the system defaults to. A scheduling model (not just a settings redesign) may be the right solution.
- Add push notifications for unread messages — insufficient on its own. Patients have lost trust in the channel; a push notification won't fix the underlying problem of invisible response status. Delivery confirmation or expected-response-time display should accompany this.
New Stories to Consider
- Message delivery confirmation + expected response time — addresses root cause of messaging abandonment. Evidence: P1, P3 (Theme 1).
- User-defined reminder scheduling (time + meal/activity anchor) — replaces rigid fixed-time reminders. Evidence: P2, P5 (Theme 2).
- Caregiver/proxy access with separate login — credential-sharing is already happening; formalizing it reduces security risk. Evidence: P6 (Theme 4). Note: validate demand before building.
- Visit note or care summary view — patients want continuity between in-person and digital touchpoints. Evidence: P4 (Surprise finding).
Priority Implications
Push notification work should be scoped to include response-status visibility — shipping push notifications alone risks reinforcing the "black hole" perception. Lab history view is validated and should move up. Caregiver access should be pulled back into active discovery before committing to build.
Planning Inbox Entry
Source
User interviews, Meridian Health app — June 2025. 6 participants, chronic condition patients (diabetes, hypertension). Conducted by product team; sessions 45–60 min.
Key Findings
- Patients have abandoned in-app messaging because there is no delivery confirmation or response timeline — they default to phone calls instead, increasing call center load.
- Medication reminders are firing at system-default times that don't match patient schedules; both participants who mentioned them had disabled them entirely.
- Lab trend data is the most-requested missing feature among diabetes patients; currently only the most recent value is displayed.
Impact on Current Plan
The assumption that notification fatigue drives low re-engagement is not supported. The actual driver is lack of trust in the messaging channel. Backlog priorities should shift accordingly: notification settings redesign is lower value than message status transparency. Caregiver access — currently deprioritized due to assumed low demand — needs a demand-validation spike before the next planning cycle.
Recommended Actions
- Add message delivery confirmation and estimated response time display to the push notification story before it moves to sprint.
- Replace fixed-time reminder defaults with user-defined scheduling tied to meals or activities.
- Move lab results history view into the next sprint — validated need, low disagreement.