Use this when a retrospective is finished and you have raw retro output (sticky notes, discussion notes, or a board export) that needs to be synthesized into grouped themes, patterns, and specific action items with owners and deadlines.
Process
Step 1: Gather inputs
Ask the user to provide:
- Raw retro output — sticky notes, board export, discussion notes, or a summary. Any format works.
- Retro format used — Start/Stop/Continue, 4Ls, Rose/Thorn/Bud, Sailboat, or other.
- Previous retro action items — what the team committed to last time (optional, but helps track follow-through).
- Team members — who was in the retro (for assigning action item owners).
Step 2: Group and theme
- Read all retro items and cluster them into themes (e.g., "communication gaps," "testing bottlenecks," "pairing quality")
- For each theme, count how many items relate to it
- Identify which themes appeared in previous retros (recurring patterns)
- Note the emotional tone — are items mostly positive, mostly frustrated, or mixed?
Step 3: Generate the synthesis
Output in this format:
Retro Synthesis: Iteration (number/date)
Format used: (Start/Stop/Continue, 4Ls, etc.) Attendees: (names)
Themes (ranked by frequency)
1. (Theme name) — (X items)
- (Representative item)
- (Representative item)
- (Pattern or root cause if visible)
2. (Theme name) — (X items)
- (Representative item)
- (Representative item)
(Continue for each theme)
Recurring patterns
- (Themes that have appeared in 2+ retros, if previous retro data was provided)
- (Trends — is sentiment improving or declining on specific topics?)
Previous action items: follow-through check
| Action item | Owner | Status |
|---|---|---|
| (item from last retro) | (name) | Done / In progress / Not started |
New action items
| # | Action | Owner | Due by | Theme |
|---|---|---|---|---|
| 1 | (Specific, measurable action) | (name) | (date) | (theme) |
| 2 | (Specific, measurable action) | (name) | (date) | (theme) |
| 3 | (Specific, measurable action) | (name) | (date) | (theme) |
(Limit to 2-3 action items — more than that won't get done)
Team sentiment
(Brief read on overall mood — energized, cautious, frustrated, optimistic)
Step 4: Review
Ask the user:
- Do these themes feel right? Anything miscategorized?
- Are the action items specific enough to follow through on?
- Are the owners right?
- Anything important that was discussed but isn't captured here?
Remember: this draft multiplies your output, not your judgment. You own the final narrative.
Output location
Present the synthesis as formatted text in the conversation for the facilitator to share with the team.
Example Output
Input
- Raw retro output: Start/Stop/Continue board export from Sprint 34 at Meridian Health Tech. START: "Dedicated QA environment per squad," "Daily async standup summaries in Slack," "Mob programming on tricky integrations." STOP: "Merging to main without a passing CI run," "Pinging engineers directly in DMs for prod issues — use #incidents," "Holding 90-min refinements when backlog is already groomed," "Skip the Friday 4pm deploy window." CONTINUE: "Pairing on new feature work," "Design reviews before dev starts," "Weekly 1:1s between EM and TLs," "Blameless post-mortems," "Async standup format (it's working!)."
- Retro format used: Start/Stop/Continue
- Previous retro action items: (1) Priya to set up a shared QA environment by Sprint 33 end — Owner: Priya Nair. (2) Team to document on-call escalation path in Confluence — Owner: Marcus Webb. (3) Move refinement to Tuesday so designs are ready — Owner: Sasha Kim.
- Team members: Priya Nair (EM), Marcus Webb (TL), Sasha Kim (PM), Dev Okonkwo (senior eng), Lena Strauss (QA lead), Tobias Reyes (eng), Chloe Park (eng)
Output (abbreviated)
Retro Synthesis: Sprint 34 — Meridian Health Tech
Format used: Start/Stop/Continue Attendees: Priya Nair, Marcus Webb, Sasha Kim, Dev Okonkwo, Lena Strauss, Tobias Reyes, Chloe Park
Themes (ranked by frequency)
1. CI/CD discipline and release safety — 4 items
- Merging to main without passing CI is creating downstream breakage
- Friday 4pm deploys are a recurring risk
- Root cause: no hard gate on the merge button; norms exist but aren't enforced by tooling
2. Testing infrastructure — 3 items
- No dedicated QA environment per squad means Lena is context-switching constantly
- Mob programming on integrations is helping catch issues earlier (positive signal)
- Pattern: QA bottlenecks have appeared in 3 of the last 4 retros
3. Async communication norms — 3 items
- Direct DMs for prod issues are bypassing #incidents, slowing response coordination
- Async standup is explicitly praised — the team wants to protect it
- Tension between "stop DM-ing me" and "async is working" suggests the format is fine; routing is the problem
4. Meeting efficiency — 2 items
- 90-min refinements when backlog is already groomed feel wasteful
- Tuesday refinement move (Sprint 33 action) seems to have helped — design readiness improved
Recurring patterns
- QA environment / testing bottlenecks — appeared in Sprint 31, 32, and 34. Priya's Sprint 33 action item (below) was not completed; this theme will keep surfacing until infra is provisioned.
- Incident routing via DMs — second consecutive retro. Escalation path doc (Marcus, Sprint 33) was delivered, but behavior hasn't changed — awareness alone isn't fixing it.
Previous action items: follow-through check
| Action item | Owner | Status |
|---|---|---|
| Set up shared QA environment by Sprint 33 end | Priya Nair | Not started — blocked on DevOps capacity |
| Document on-call escalation path in Confluence | Marcus Webb | Done |
| Move refinement to Tuesday so designs are ready | Sasha Kim | Done — team noted improvement |
New action items
| # | Action | Owner | Due by | Theme |
|---|---|---|---|---|
| 1 | Add required CI status check to main branch in GitHub — no merge without green build | Marcus Webb | Feb 7 | CI/CD discipline |
| 2 | Priya to unblock QA environment request with DevOps lead; timebox to 1 working session this sprint | Priya Nair | Feb 5 | Testing infrastructure |
| 3 | Post a 1-sentence norm in #incidents channel pinned message: "All prod issues here first, no DMs" — Dev to model and reinforce for one sprint | Dev Okonkwo | Feb 3 | Async communication norms |
Team sentiment
Cautiously optimistic. The async standup praise and blameless post-mortem shoutout suggest psychological safety is solid. The frustration is concentrated in tooling and process gaps — things the team believes are fixable — rather than interpersonal friction. The QA theme carrying over a third time may start to erode trust in follow-through if it slips again.
Draft for Priya to review before sharing with the team. Check: Are the CI gate and QA owners right given current DevOps bandwidth?