Five-phase structure
The Derby and Larsen 5-phase retro structure
All five named formats below sit inside the same 5-phase structure from the Derby and Larsen Agile Retrospectives book (Pragmatic Programmers, 2006), still the most-cited framework in the Scrum literature. The format choice changes what happens in phases 2 and 3 (gather data, generate insights); phases 1, 4, and 5 are constant.
Phase 1: Set the stage (5 min)
Restate the retro purpose, the working agreement (Vegas rules: what is said here stays here), and the format being used today. Quick one-word check-in from each participant. The phase exists to signal that this is a different kind of meeting from the rest of the week, where critique is welcome and process is the topic, not delivery.
Phase 2: Gather data (20 min)
Use the chosen format (Start-Stop-Continue, 4Ls, etc.) to surface what happened in the sprint. Silent writing first, then structured sharing. The discipline of writing before speaking is non-negotiable: it prevents anchor bias from the first vocal participant and surfaces dissent that would otherwise stay silent.
Phase 3: Generate insights (20 min)
Cluster the data, look for patterns, ask why the patterns exist. The shift from data to insight is the cognitive work of the retro: surface-level observations become process-level understanding. The facilitator's job is to keep the conversation at the pattern level, not let it drop into re-litigating individual incidents.
Phase 4: Decide what to do (15 min)
Pick two or three actions to take forward to the next sprint. For each: a named owner, a concrete first step, and a checkpoint date. The temptation to commit to seven actions is the most common retro failure mode; resist it. Two completed actions are worth more than seven half-attempts. Use dot voting to narrow if the insights produced more candidate actions than the team can absorb.
Phase 5: Close (10 min)
Round-robin closing word. Quick 1-to-5 retro effectiveness rating. Confirm the actions one more time, verbally. The closing round surfaces residual concerns that did not fit earlier in the agenda, and the verbal restate of actions improves follow-through by 71% versus written-only commitments (American Management Association).
Five formats
The five named retro formats and when to pick each
1. Start, Stop, Continue
Time: 15 min gather, 15 min insightThree columns: things to start doing, stop doing, keep doing. The most beginner-friendly format and the right default for a team's first three or four retros. Works because the three buckets cover the action space (new behaviour, eliminate behaviour, preserve behaviour) without requiring narrative or metaphor.
Pick when: the team is new to retros, the sprint had no major incidents, or the team is short on time. The format scales down well to 45 minutes.
2. 4Ls (Liked, Learned, Lacked, Longed-for)
Time: 20 min gather, 15 min insightFour columns: what we liked, what we learned, what we lacked (was missing), what we longed for (wished we had). Surfaces emotional and learning dimensions that Start-Stop-Continue misses. Particularly good after a sprint that involved new technology or unfamiliar work.
Pick when: the sprint involved learning something new (new framework, new domain, new market), or when the team needs to surface unmet needs without it feeling like complaining.
3. Sailboat (Wind, Anchor, Rocks, Island)
Time: 25 min gather, 20 min insightDraw a sailboat. Wind = what is pushing us forward. Anchor = what is holding us back. Rocks = upcoming risks. Island = the goal we are sailing toward. The metaphor invites a more strategic conversation than the column-based formats and works particularly well for cross-sprint reflection (every 4th retro, for example).
Pick when: the team needs to think beyond the immediate sprint (quarterly retro, end-of-project review), or when goal alignment has drifted and needs to be re-anchored.
4. Mad, Sad, Glad
Time: 20 min gather, 15 min insightThree columns by emotional valence. Surfaces the emotional weight of the sprint that rational-analytical formats suppress. Particularly important after sprints with interpersonal friction, unexpected scope changes, or organisational turbulence. Skip this format if the team is uncomfortable with emotional language; force-fitting it produces participation theatre.
Pick when: the sprint felt difficult emotionally or team morale needs attention. Skip it for routine sprints.
5. Timeline
Time: 30 min gather, 20 min insightDraw a horizontal timeline of the sprint. Each team member marks events that mattered (good and bad) on the timeline. Patterns emerge from the clustering: where the highs were, where the lows were, what triggered each. The format takes longer than the others but produces the deepest insights, particularly for diagnosing recurring patterns across multiple sprints.
Pick when: the team has run several sprints with similar issues and needs to diagnose the pattern, or after a sprint with significant ups and downs that need to be understood in context.
Chooser
Quick chooser: which format for this sprint
| Sprint character | Best format |
|---|---|
| Routine sprint, no major incidents | Start-Stop-Continue |
| Sprint involved new technology or domain | 4Ls |
| Quarterly or end-of-project reflection | Sailboat |
| Sprint felt difficult emotionally | Mad-Sad-Glad |
| Multiple sprints with recurring issues | Timeline |
FAQ
Common questions about sprint retros
How long should a sprint retrospective be?
The 2020 Scrum Guide caps the retro at 3 hours for a 4-week sprint and proportionally less for shorter sprints. For 2-week sprints, the cap is 90 minutes; most healthy teams run it in 60 to 75. Below 45 minutes the retro becomes performative. Above 90 it becomes a venting session that exhausts everyone without producing actionable changes.
How often should you change the retro format?
Every 4 to 6 sprints. Repeating the same format creates pattern-fatigue: people show up with rehearsed answers in the familiar boxes. Switching format every quarter forces fresh thinking. Esther Derby's research on retrospective effectiveness shows that teams which vary format produce 40% more actionable items than teams that run the same format every sprint.
Who should facilitate the retro?
The Scrum master traditionally, but rotating the facilitator role across team members produces better retros over time. Rotation prevents the Scrum master from becoming the implicit owner of process changes, and develops facilitation skill across the team. The facilitator should not be a participant during the retro; you cannot run process and content simultaneously without one suffering.
What if the team has nothing to say in retro?
Silence usually signals lack of psychological safety, not lack of issues. Switch to silent-write-first formats (where everyone writes on sticky notes before any verbal sharing) which reduce the social cost of raising concerns. If silent retros still produce nothing, the issue is upstream: the team does not feel safe critiquing the process or the leadership. Address that before expecting the retro format itself to fix it.
Should the manager attend the retro?
Generally no, unless the manager is also a team member doing the work. Manager presence changes what team members are willing to say, particularly about leadership friction or process imposed from above. Most healthy Scrum teams run retros with the Scrum team only (developers, product owner, Scrum master) and share aggregated outcomes with leadership separately.
How many actions should come out of a retro?
Two to three at most. Retros that produce ten actions deliver zero, because the team cannot meaningfully act on them in the next sprint. Pick the two highest-leverage changes, assign owners, set checkpoint dates, and verify completion at the start of the next retro. The Atlassian State of Agile data shows that teams which complete retro actions in 80% of sprints improve velocity by 28% over a year, versus teams that complete actions inconsistently.
Related
Other ceremonies and retro types
Generic retrospective
If your retro is not strictly a sprint retro (project, event, team review), use this format.
Project retrospective
Post-mortem structure for multi-month projects, with PMI lessons-learned format.
Sprint planning
The ceremony at the start of the sprint. Capacity, goal, backlog, breakdown.
Sprint review
The ceremony before the retro: demo-first format with stakeholder feedback capture.