How to Run a Sprint Retrospective
A sprint retrospective is the meeting where a team looks at how it worked — not what it built — and commits to one or two concrete changes for the next sprint. Here is how to run one that people don't dread, step by step.
Free · No credit card · Guests join by link
Why retrospectives work — the evidence
Retrospectives aren't an agile ritual to be endured — the mechanisms behind them are some of the best-studied in organizational research.
Reflection compounds learning
Harvard Business School research found that employees who spent even a few minutes deliberately reflecting on their work significantly outperformed those who spent the same time simply doing more work. A retro is that reflection, structured, for a whole team.
Psychological safety drives performance
Amy Edmondson's work at Harvard — synthesized across nearly two hundred studies — shows teams where it's safe to raise problems learn faster and perform better. Anonymous writing is that safety, built into the format.
Wellbeing shows up in output
A Harvard Business School review of 339 studies links meaningful improvements in team wellbeing to roughly a 10% uplift in productivity. The retro is where a team's friction gets named early — before it becomes attrition.
The six phases, step by step
Whatever format you pick, a well-run retro moves through the same phases. Each one exists to protect the meeting from a specific failure mode.
- 1
Write in silence
Everyone adds cards to the board at the same time, anonymously, with a timebox (5–10 minutes). Cards stay concealed while people write — nobody can see what anyone else wrote yet.
TipResist the urge to chat during writing. Silence isn't awkward here, it's the point: simultaneous, independent writing is what stops the first opinion in the room from anchoring everyone else's.
AvoidLetting people 'just discuss it live' instead of writing. The people who think out loud will dominate, and the quieter half of the team's honest input never makes it onto the board.
- 2
Vote with limited tokens
Each person gets a small, fixed budget of voting tokens (3–5 works well) and places them on the cards they think deserve discussion. Scarcity is what makes the vote meaningful.
TipConfigure fewer tokens than there are cards. If everyone can vote for everything, you've measured nothing — the constraint is what forces real prioritization.
AvoidUnlimited +1s. A vote that costs nothing produces a board where everything is 'important', which is the same as nothing being important.
- 3
Discuss the top cards
Walk through the highest-voted cards one by one, top vote-getter first. The facilitator steers which card the group is looking at; anyone can comment on any card.
TipTimebox each card (3–5 minutes) and do a quick 'keep going or move on?' check when time is up. The team's votes already told you what matters — trust them.
AvoidSpending 25 minutes on the first card. The facilitator's main job in this phase is protecting time for card two and three.
- 4
Reveal and group themes
Card authors are revealed, and the facilitator drags related cards together into named themes. Twelve scattered cards usually collapse into three or four real patterns.
TipName themes after problems, not people — 'deploys are stressful', not 'the backend team is slow'. The theme name will outlive the meeting; make it something you can fix.
AvoidSkipping grouping entirely. Ungrouped cards read as twelve disconnected complaints; grouped, they read as three problems with evidence.
- 5
Commit to action items
Convert the discussion into specific action items with an owner and, ideally, a due date. This is the phase most retrospectives skip — and the reason most retrospectives feel pointless.
TipOne to three actions, no more. A retro that produces eight actions produces zero completed actions. Small and owned beats ambitious and orphaned.
AvoidActions without owners. 'We should improve our testing' is a wish; 'Sarah spikes the flaky checkout test by Friday' is an action.
- 6
Close and measure
Wrap up with a summary everyone can export, and a quick anonymous pulse — one tap on 'how was this sprint?' — so team health becomes a trend you can see over time, not a feeling.
TipOpen your NEXT retro by reviewing this retro's action items. That thirty-second ritual is the single strongest signal that retro actions are real commitments.
AvoidEnding without closing the loop. If actions vanish into a document nobody reopens, the team learns that speaking up changes nothing — the opposite of what a retro is for.
Choosing a format
The phases stay the same; the columns change what the team reflects on. Every format below runs as a free board with the full six-phase flow built in.
All-Purpose
Balanced formats built for every regular sprint retro.
Strengths · Risks · Learnings
A balanced sprint review that looks backward and forward at once — Retromik's signature format.
View templateStart · Stop · Continue
The most actionable retro format — every column maps directly to a behaviour change.
View templateDrop · Add · Keep · Improve
Four dials instead of three — the only format that separates a brand-new experiment from tuning what already exists.
View templateMore of · Less of · Keep
Calibrate instead of overhaul — dial behaviours up and down rather than starting from zero.
View templateGoals · Fouls · MVP Moves
A football-flavored retro for teams who'd rather talk goals and fouls than strengths and risks — same substance, more fun to say out loud.
View templateTeam Health
Surface how the team actually feels, safely.
Dig Deeper
Root-cause and learning-focused formats for periodic deep dives.
Change The Rhythm
Structurally different formats for when the usual columns don't fit.
Early · Mid · Late Sprint
Sorts by when it happened, not what kind — the only format built to catch narrative arcs other retros flatten into a single pile.
View templateOpen Agenda
The only single-column format here — no categories, just an open, priority-ordered agenda the group builds together.
View templateWorth It · Not Worth It · Change
The only format here that retros the retro — evaluate the ritual itself instead of the sprint.
View templateCommon facilitation mistakes
Running the same format until it's autopilot
After months of identical columns, people pattern-match instead of reflecting. Rotate formats — a chronological or open-agenda retro asks genuinely different questions than a categorized one.
No follow-up on action items
The fastest way to kill retros is for their outputs to vanish. Review last retro's actions at the start of every retro, and keep them visible in a shared action list between sprints.
Letting the loudest voice set the agenda
Anonymous simultaneous writing plus token voting exists precisely to prevent this — the board's priorities should come from the whole team's votes, not from whoever speaks first or most.
Blaming people instead of examining systems
If retros feel unsafe, honest input stops. Metaphor-based formats depersonalize criticism ('the anchor is our release process' lands differently than 'you slow down releases'), and lean formats aim the discussion at process causes.
Skipping the retro when the sprint was rough
The sprints teams most want to skip reflecting on are the ones with the most to learn from. A rough sprint is a reason to run the retro, not a reason to cancel it.
Remote and async retrospectives
A remote retro follows the exact same phases — the board just replaces the wall of sticky notes. Share the board link, teammates join as guests without creating accounts, and the facilitator advances the phases while everyone follows along live.
For distributed teams across time zones, split the retro: open the board a day early and let people add cards asynchronously during the writing phase, then meet for thirty minutes to vote, discuss, and commit to actions together. Writing async loses nothing — it's silent and independent by design anyway. The discussion is the part that deserves everyone live.
Frequently asked questions
How long should a sprint retrospective take?
For a team of 5–8 people, 45–60 minutes works well: about 10 minutes of silent writing, 5 minutes of voting, 25 minutes discussing the top-voted cards, and 10 minutes committing to action items. Larger teams need firmer timeboxes, not longer meetings.
How often should we run retrospectives?
Once per sprint, at the end of the sprint, while it's still fresh. Teams that batch retrospectives monthly or quarterly lose the detail that makes actions specific — by then nobody remembers why week two felt chaotic.
Who should facilitate the retrospective?
Anyone on the team can — and rotating the facilitator keeps the ritual from calcifying. It deliberately doesn't need to be the manager: when the manager facilitates, anonymous writing matters even more, because it lets the team raise issues the manager is part of.
Does a retrospective really need anonymity?
During writing, yes — the research on psychological safety is clear that people self-censor when input is attributable, especially about problems involving people senior to them. Authors are revealed later, at the grouping phase, once every card is already on the board on its own merits.
What if the team has nothing to say?
That's almost always a format problem, not a team problem. Switch to a format that asks different questions — a chronological retro (what happened early vs. late in the sprint) or an open-agenda format surfaces things a strengths/risks board never will.
Estimating work too?
The same free workspace includes Scrum Poker — Fibonacci estimation with simultaneous voting and a live agreement score.
Run your next retro on Retromik
Anonymous cards, token voting, theme grouping, and tracked action items — the whole six-phase flow in one free board. Teammates join by link, no accounts needed.
Start free