Every planning layer needs its own feedback loop. Sprint retros catch process issues before they compound. Quarterly retros catch strategic drift between squads. Yearly retros catch business assumptions that have quietly gone stale. Same discipline at every scale: reflect, learn, adjust. This is the one-page reference.
The retro map
| Scale | Cadence | Who’s in the room | What it catches | Format | Worked example |
|---|---|---|---|---|---|
| Sprint | Fortnightly | The squad (5-8 people) | Process issues, team dynamics, technical debt, blocked work | Five-stage retro (set the stage, gather data, generate insights, decide actions, close) | A first retro that surfaces the wrong-kind-of-fast: Catching the Wrong Kind of Fast |
| Cross-team | After incidents or quarterly | Squad leads + affected team members | Coordination failures, duplicated work, API contract surprises | Incident-triggered: show the data, find patterns, agree actions | A cross-team retro after a contract change between two squads: Two Squads, One Direction |
| Quarterly | Quarterly (Block 1 of quarterly planning day) | Squad leads + engineering leadership + product | Strategic drift between squads, theme outcomes vs expectations, planning accuracy | Each squad lead narrates what worked / surprised / would change, then review incidents | A first quarterly planning day with multiple squads: Two Squads, One Direction |
| Yearly | Annual (part of strategy refresh) | Leadership team: founder, eng lead, mentor, squad leads | Business assumptions that turned out wrong, market shifts, what the org would do differently | Strategic reflection: “What did we believe a year ago that we now know is wrong?” | A yearly retro that names a market assumption that didn’t survive contact with reality: From Weekly Habits to Yearly Vision |
The five-stage format
The format below is the workhorse retro across every scale. Once a team learns it, they run it without prompting. The questions and the people change layer to layer; the structure does not.
- Set the stage — Read the Prime Directive. Check in: one word per person describing how they feel about the sprint. This takes two minutes and surfaces mood before content.
- Gather data — Silent writing on sticky notes. Each person writes as many observations as they have. Read aloud, place on the wall. No discussion yet.
- Generate insights — Cluster the notes. Find patterns. Name root causes, not symptoms. “Deployments are slow” is a symptom. “No automated pipeline” is a root cause.
- Decide what to do — Propose concrete actions. Dot vote. Pick one or two. An action without an owner and a deadline is a wish.
- Close — Each person appreciates something about a colleague. Takes sixty seconds. Ends the retro on connection rather than criticism.
The five stages remain constant. At sprint scale, “gather data” means sticky notes about the last two weeks. At quarterly scale, it means each squad lead narrating outcomes against expectations. At yearly scale, it means reviewing which assumptions from the previous strategy refresh turned out wrong. Same structure, different scope.
How retros evolve with team size
You don’t need all four retro layers on day one. Add each layer when the problems it catches start appearing.
| Team size | Retro layers needed | Why |
|---|---|---|
| 1-5 | Sprint retro only | One team, one context. Sprint retro catches everything. |
| 5-15, multiple teams | + Cross-team retro | Coordination failures are invisible to single-team retros. |
| 15-25, multiple teams | + Quarterly retro | Strategic drift between squads. Theme outcomes need review. |
| 25+ | + Yearly retro | Business assumptions compound silently. Need explicit check. |
For a worked example of layers being added at each transition, the Greenbox narrative covers it across The First Sprints (sprint retros), Two Squads, One Direction (cross-team and quarterly), and From Weekly Habits to Yearly Vision (yearly).
Common failure modes
Retros that produce no actions. “What went well / what didn’t” without “what will we do differently” is just venting. The retro must end with owned, concrete actions or it’s a waste of everyone’s time.
Actions that never get done. The retro produces actions but nobody tracks them. Next retro: same problems, same frustration, growing cynicism about the process.
Same problems every time. If the same issue appears three retros in a row, the retro isn’t failing — the team is failing to act on it. Escalate. Surface it at the next layer up. If sprint retros can’t fix it, it’s a cross-team or quarterly problem.
Skipping retros under pressure. “We don’t have time for a retro, we need to ship.” This is when you need a retro most — pressure creates the process problems retros are designed to catch. Skipping the retro under pressure is like skipping the brake check before driving faster.
Wrong people in the room. Sprint retros need the squad. Cross-team retros need the squad leads. Quarterly retros need leadership. If the people who can act on the problems aren’t present, the retro is theatre, it surfaces issues that nobody in the room has the authority to fix.
Building structured retros with Retromat
If your retros have fallen into a rut, the same format every fortnight, the same complaints, the same vague actions that never get done, Retromat is worth your time. It’s a free tool that generates structured retrospective plans by combining activities from a library of over a hundred facilitation exercises, one for each phase of the five-stage format.
The power isn’t the novelty, it’s the structure. A well-facilitated retro isn’t a venting session. It’s not forty-five minutes of complaints followed by a list of twelve things to change, none of which survive contact with the next sprint because the real work always wins. That spiral, complain, list, ignore, repeat, is what kills retro culture. People stop believing the process works, so they stop engaging, so the process stops working. It’s self-fulfilling.
Retromat breaks the cycle by giving the facilitator a different set of activities each time. Instead of “what went well / what didn’t” every fortnight until everyone’s eyes glaze over, you might run a speed-dating format where pairs discuss one topic for two minutes and rotate, or a sailboat exercise where the team maps winds (things pushing them forward) and anchors (things holding them back), or a “circles and soup” exercise that separates things the team controls, things they can influence, and things they can only respond to.
The variety matters because it forces different thinking. The same format surfaces the same observations. A new format surfaces new ones. And the five-stage structure stays constant underneath, set the stage, gather data, generate insights, decide actions, close. You’re not reinventing the retro. You’re varying the data-gathering technique while keeping the decision-making discipline.
The discipline part is non-negotiable. Every retro ends with one or two concrete actions, each with an owner and a deadline. Not twelve. Not “we should do better at X.” One thing, owned by a person, due by a date. If you can’t get it down to one or two, you haven’t generated enough insight, you’ve just collected complaints. The retro’s job is to find the one change that would make the biggest difference, commit to it, and follow through. Next retro, you check: did we do it? Did it help? What’s next?
That single-action discipline is what makes retros worth running. Choose ten things, do none. Choose one, do it, measure, repeat. The compounding effect over a year is enormous; the cost of running the retro is two hours a fortnight. The first retro in the Greenbox narrative is a worked example: it produces a single action, and that action changes the trajectory of the whole company.
The principle
Every planning layer needs its own feedback loop. The sprint retro asks “how is our process working?” The quarterly retro asks “are we pursuing the right themes?” The yearly retro asks “are we in the right business?” Same muscle. Different scope. Different cadence. All essential.
Related references
- Retromat — the facilitation library this post draws from
- Which Workshop When — every discovery and delivery technique in one place
- The Planning Onion — every planning layer in one place
- LLMs as Thinking Partners — discovery techniques as inputs to LLMs
- The Greenbox Story — narrative worked examples of every layer in this post