Mad Sad Glad, 4Ls, Start Stop Continue: 12 Retro Formats Compared

Run Start-Stop-Continue for eight sprints in a row and you will watch the same thing happen every time: the board fills with the same six cards, the discussion takes fifteen minutes, and everyone leaves with a vague sense that the retro "went fine." The format did not break. It wore out. Formats are lenses, and any lens pointed at the same scene for long enough stops revealing anything new.

This guide compares 12 retrospective formats in practical terms: what each one surfaces, what it hides, the exact prompts to use, and how to run it with a remote team. It assumes you already run retros and want them to produce something. If you need the fundamentals first — the prime directive, facilitation basics, follow-through — start with our complete guide to retrospectives and come back.

One rule before the list: the format is maybe 30% of a retro's value. The other 70% is whether people feel safe saying true things and whether actions from the last retro actually got done. No format fixes a team that ignores its own action items. If that is your problem, read Action Items That Actually Get Done first.

How to read these comparisons

Each format below gets the same treatment:

• What it surfaces — the kind of insight this lens is good at extracting
• When to use it — sprint state, team mood, or situation where it shines
• Prompts — the exact column headers and the one-line explanation you give each
• Remote tips — what changes when the team is on a shared board instead of in a room
• Watch out for — the failure mode specific to this format

A quick orientation table before the detail:

Format	Emotional depth	Structure	Best for	Typical length
Start Stop Continue	Low	High	Action-focused teams, new retros	45 min
Mad Sad Glad	High	Medium	Tense sprints, post-conflict	60 min
4Ls	Medium	Medium	Learning-heavy periods	60 min
Sailboat	Medium	High	Goal-oriented teams, quarters	60 min
Starfish	Low	High	Calibrating effort, process tuning	60 min
KALM	Low	High	Mature process, small tweaks	45 min
DAKI	Low	High	Decisive teams, process pruning	45 min
Plus/Delta	Low	Low	Quick retros, workshops, events	20 min
Rose Bud Thorn	Medium	Medium	Balanced check, spotting opportunity	45 min
Lean Coffee	Medium	Low	Teams tired of columns	60 min
Timeline	Medium	Medium	Long periods, incidents, releases	75 min
Hopes and Fears	High	Low	Kickoffs, reorgs, new teams	45 min

The workhorses: Start Stop Continue, Mad Sad Glad, 4Ls

1. Start Stop Continue

What it surfaces: behaviors, framed as decisions. Every card is implicitly an action, which is why this is the most action-dense format on the list.

Prompts:

• Start — "What should we begin doing that we are not doing today?"
• Stop — "What is costing us more than it returns?"
• Continue — "What is working and should be protected?"

When to use it: new teams, new retro practices, or any sprint where you want concrete output more than emotional processing. It is the best default format precisely because it is boring — nobody needs the metaphor explained.

Remote tips: because cards are naturally action-shaped, resist the urge to skip discussion and just vote. The Stop column especially needs conversation: "stop having so many meetings" is a complaint, not an action, until someone names which meeting.

Watch out for: Continue becoming a dumping ground for pleasantries ("continue being awesome"). Cap it: ask for Continue cards that name a specific practice someone might accidentally drop. Also watch for the staleness described in the intro — this format wears out fastest because its outputs converge.

2. Mad Sad Glad

What it surfaces: emotion first, facts second. People file the same event under different feelings, and the filing is the data. A missed release might show up as Mad ("we ignored the warning signs") for one person and Sad ("I worked the weekend for nothing") for another. Those are different problems needing different fixes.

Prompts:

• Mad — "What frustrated you? What felt unfair or avoidable?"
• Sad — "What disappointed you? What do you wish had gone differently?"
• Glad — "What gave you energy? What went better than expected?"

When to use it: after a rough sprint, a conflict, a failed launch, or any period where you can feel tension in standup but nobody is naming it. Emotion-first formats give people permission to say the thing.

Remote tips: anonymity matters more here than in any other format. Run card-writing anonymously and reveal authorship only if the author chooses to speak to their card. On a shared retro board, write cards privately first, then reveal all at once — it prevents anchoring, where the first Mad card sets the tone for everyone else's.

Watch out for: the facilitator rushing past Mad because it is uncomfortable. If someone writes an angry card, the worst response is a quick nod and moving on. Ask "what would have needed to be true for this not to happen?" — it converts heat into a fixable condition.

3. The 4Ls: Liked, Learned, Lacked, Longed For

What it surfaces: a balance of appreciation and gap analysis, with an explicit learning column that no other classic format has.

Prompts:

• Liked — "What did you enjoy or appreciate?"
• Learned — "What do you know now that you did not know two weeks ago?"
• Lacked — "What was missing — skills, information, tools, time?"
• Longed For — "What do you wish existed, even if it feels unrealistic?"

When to use it: periods heavy on discovery — new tech, new domain, new teammates. The Learned column turns private lessons into team knowledge; it is worth running 4Ls once a quarter for that column alone.

Remote tips: Learned cards are the ones worth keeping after the retro ends. Export them or copy them into your wiki — they are documentation that wrote itself. Lacked and Longed For overlap in practice; if the team keeps filing the same card in both, merge them into one column and spend the time on discussion instead.

Watch out for: Longed For turning into a wishlist aimed at management ("longed for: double the headcount"). Legitimate, but not actionable in a team retro. Acknowledge it, park it, and note it for whoever owns that conversation.

The metaphor formats: Sailboat and Starfish

4. Sailboat

What it surfaces: a systems view — the team as a boat with wind (helping forces), anchors (drag), rocks (risks ahead), and an island (the goal). It is the only classic format with a built-in forward-looking risk column.

Prompts:

• Wind — "What is pushing us toward the goal?"
• Anchors — "What is slowing us down?"
• Rocks — "What risks could sink us before we arrive?"
• Island — "Do we agree on where we are going?" (Do this one first, out loud.)

When to use it: quarterly retros, project midpoints, or any moment when the team is executing fine but drifting. If the Island discussion produces three different answers, you have found your retro topic and can ignore the rest of the board.

Remote tips: the visual matters. Use a template with an actual boat drawn on it — on a plain four-column board the metaphor collapses into "good stuff / bad stuff" and you lose the risk framing. Most retro tools, Openbook's retro room included, ship Sailboat as a drawn template for exactly this reason.

Watch out for: rocks nobody owns. Risk cards feel virtuous to write and easy to forget. Every rock that survives voting needs either a named owner or an explicit "we accept this risk" — anything else is theater.

5. Starfish: Keep, Less, More, Stop, Start

What it surfaces: calibration. The genius of Starfish is the Less and More columns — most process problems are not "we do X and shouldn't" but "we do X too much" or "not enough." Binary formats miss that entirely.

Prompts:

• Keep doing — "Right amount, keep it as is."
• Less of — "Useful, but we overdo it."
• More of — "Useful, and we underdo it."
• Stop doing — "Not useful at all."
• Start doing — "Missing entirely."

When to use it: teams with established process that needs tuning, not replacing. A team that says "our code reviews are fine, mostly" will produce a revealing Less/More split: less nitpicking on formatting, more scrutiny on architecture.

Remote tips: five columns is a lot of board. Timebox card-writing to 7 minutes or people spread thin and every column gets two shallow cards. Better: tell the team in the invite which two or three columns you expect to matter this sprint.

Watch out for: the same practice appearing in Less and More from different people. That is not a contradiction to resolve by vote — it is a signal that the practice serves people unevenly. Discuss who needs it and who is taxed by it.

The decision formats: KALM and DAKI

6. KALM: Keep, Add, Less, More

Starfish minus the Stop column. That sounds trivial; it is not. Removing Stop changes the emotional register — KALM assumes the current process is basically sound and invites adjustment rather than rejection. Use it with a team that gets defensive when their process is attacked, or shortly after a big process change when relitigating the change would be destructive. Prompts mirror Starfish. Watch out for: using KALM to avoid a Stop conversation the team genuinely needs. If people keep writing "less of X" when they mean "stop X," the format is muzzling them — switch.

7. DAKI: Drop, Add, Keep, Improve

What it surfaces: decisions, stated bluntly. Drop is stronger than Stop — it targets artifacts and practices by name: drop this meeting, drop this report, drop this tool.

Prompts:

• Drop — "What should we delete outright?"
• Add — "What is worth introducing?"
• Keep — "What earns its place?"
• Improve — "What stays, but needs work?"

When to use it: process cleanup. Ideal once or twice a year as a deliberate pruning ritual, or when a team inherits process from a previous manager or a larger org and nobody remembers why half of it exists.

Remote tips: for Drop cards, run a cost check before the vote: "who would notice if this disappeared, and what would they lose?" Answering it in the card comments before discussion keeps the session honest and fast.

Watch out for: Improve becoming the coward's Drop. "Improve the weekly sync" often means "I want this gone but won't say so." Ask Improve card authors to name the specific improvement; if they cannot, ask whether the card belongs in Drop.

The lightweight formats: Plus/Delta and Rose Bud Thorn

8. Plus/Delta

Two columns: Plus ("what worked?") and Delta ("what would you change?"). Delta is deliberately not "Minus" — it asks for a change, not a complaint, which keeps the tone constructive with almost no facilitation effort.

When to use it: anything short — a workshop, a launch day, a single meeting, an event. This is the format for retro-ing things that are not sprints. Ten minutes, two columns, done. It is also the right training-wheels format for a team that has never retro'd and is suspicious of the whole idea.

Watch out for: using it as your standing sprint format. It is too shallow to carry a team's only reflection ritual. It is a snack, not a meal.

9. Rose, Bud, Thorn

Prompts:

• Rose — "What bloomed? A win, large or small."
• Bud — "What shows promise but has not paid off yet?"
• Thorn — "What hurt?"

The Bud column is the reason to pick this format: it is the only lens on this list that explicitly hunts for early positive signals — the half-finished tooling that is already saving time, the new hire's untapped skill, the experiment worth doubling down on. Teams that only discuss problems and wins miss the middle category where compounding investments live.

When to use it: stable teams in steady periods, where the interesting question is not "what broke?" but "what should we invest in?"

Remote tips: ask people to write at least one Bud before anything else. Left to instinct, the Thorn column fills first and Buds get token effort.

The structure-breakers: Lean Coffee, Timeline, Hopes and Fears

10. Lean Coffee

No columns at all. Everyone writes topics they want to discuss, the team votes, and you discuss in vote order with strict timeboxes: 8 minutes per topic, then thumbs up/down — continue for 4 more or move on.

When to use it: teams with strong opinions and retro fatigue. Lean Coffee hands the agenda entirely to the room, which makes it the highest-signal format when the team has things to say and the emptiest when it does not. If topic-writing produces three weak cards, end early and take the hint: shorten your retro cadence or the sprint genuinely had no friction.

Remote tips: the timebox votes need to be visible and fast — use board reactions or a quick poll rather than asking around the room. The facilitator's whole job is enforcing the clock; everything else runs itself.

Watch out for: the loudest person's topic winning by charisma rather than votes. Vote on written topics before any topic is pitched aloud.

11. Timeline Retrospective

Draw the period as a horizontal line with real dates. Everyone places events on it — releases, incidents, decisions, departures, wins — then marks how they felt at each point (a simple color or up/down works). Discussion follows the line left to right.

When to use it: anything longer than a sprint. Quarter retros, project post-mortems, release retros, incident reviews. Memory over long periods is unreliable and recency-biased; the timeline rebuilds the actual sequence before anyone draws conclusions from it. It routinely produces the best "wait, THAT is when things went sideways" moments of any format — often at an event nobody would have written on a column board.

Remote tips: build the event layer async before the meeting. Placing 40 events live eats half the session; annotating and discussing them is the valuable part. A whiteboard or retro room where people can drop events during the week beforehand turns a 2-hour session into 75 minutes.

Watch out for: litigating what happened instead of learning from it. The timeline is a shared memory aid, not a court exhibit. If two people remember an event differently, note both versions and move on.

12. Hopes and Fears

Two columns, asked before the work instead of after: "What do you hope happens?" and "What are you afraid will happen?" Technically a futurespective, and the single best opening ritual for a new team, a new quarter, a reorg, or a scary project. Fears written in week one become the checklist you revisit in week six — half will have been dodged, and the other half were early warnings you can still act on. Pair it with a team health check at the midpoint to see which fears are materializing.

Watch out for: collecting fears and never revisiting them. The format's entire value is the revisit. Schedule it when you run the session, not after.

Choosing a format: a decision guide

Do not rotate formats for novelty. Choose the lens that matches the question you need answered:

The situation	Use
New team or new retro practice	Start Stop Continue or Plus/Delta
Tension you can feel but nobody names	Mad Sad Glad
Heavy learning period, new domain	4Ls
Executing fine but possibly drifting	Sailboat
Process needs tuning, not replacing	Starfish or KALM
Process needs pruning	DAKI
Steady state, looking for investments	Rose Bud Thorn
Retro fatigue, opinionated team	Lean Coffee
Quarter end, post-incident, post-release	Timeline
Kickoff, reorg, new manager	Hopes and Fears

A sustainable rotation for a two-week-sprint team: Start Stop Continue as the default, Mad Sad Glad or 4Ls every third retro, Timeline at quarter boundaries, and DAKI twice a year. That is variety with a reason behind each switch.

Remote facilitation: what applies to every format

Five practices that matter regardless of which columns you draw:

1. Write privately, reveal together. Anchoring is the biggest quality killer in remote retros. Every decent retro tool supports hiding cards until reveal; use it every time.
2. Timebox writing to 8-10 minutes. Longer does not produce more insight, it produces longer cards. Two short cards beat one paragraph.
3. Group before voting, vote before discussing. Duplicate cards split votes and bury the real priorities. Give the facilitator two minutes to cluster, then dot-vote (3 votes per person), then discuss top-down.
4. Cap actions at three. A retro that produces eight actions produces zero. Three actions, each with an owner and a date, reviewed at the start of the next retro before any new cards are written.
5. Keep the artifact. A retro that lives only in the meeting evaporates. Export the actions, keep the board.

This is also where tooling stops being a nice-to-have. A shared doc can host a Plus/Delta; it cannot do hidden-until-reveal cards, dot voting, card grouping, or an exportable action list. Openbook's Retrospective room ships all 12 of these formats with those mechanics built in — private writing, voting, grouping, an AI summary of the session, and CSV export of actions so they land in your tracker instead of dying on the board. If your retros currently live in a doc, that gap is worth an hour of trial: see /features.

Next steps

• This week: identify which format you have been running and how many sessions in a row. If the answer is "the same one, more than five times," schedule a switch using the decision table above.
• Next retro: open by reviewing last retro's actions before writing any cards. If fewer than half got done, make follow-through the retro topic — no format will out-perform that fix.
• Next quarter boundary: run a Timeline retro with async event collection the week before.
• Once: run Hopes and Fears at your next kickoff and put the revisit date on the calendar the same day.

If you want the full facilitation playbook around these formats — prime directive, handling dominant voices, retro cadence — the complete retrospective guide covers it end to end.

Want to try these formats with the mechanics handled for you? Openbook's Retrospective room includes all 12 formats, real-time voting and grouping, and action export — free to start at openbook.work.