Starfish retrospective

Not every change is a choice between doing something and not doing it. The Starfish adds More of and Less of to Start, Stop and Keep, so your team can tune what it already does instead of only switching habits on and off.

Start

New habits or experiments we should begin—what’s missing?

Stop

Things that drain value or energy—what do we cut?

Keep

Proven practices to preserve—what still works?

More of

Wins we want to amplify—what should we double down on?

Less of

Useful but overdone items—what should we dial back?

Use this template — it's free

What a Starfish retrospective is

The Starfish retrospective sorts the team's practices into five groups, one for each arm of a star: Start, Stop, Keep, More of and Less of. The first three are the familiar Start Stop Continue. The other two are what make the format useful: they hold practices that are right in principle but wrong in amount. The format is often attributed to Patrick Kua.

Think of the five columns as a dial rather than a switch. Start and Stop are the two ends: turn something on, turn it off. Keep leaves the dial where it is. More of and Less of turn it up or down. Most of what an established team does — code review, meetings, documentation, testing, estimation — sits somewhere in the middle of that dial. A three-column format forces people to argue whether to stop or continue a practice when the honest answer is "continue, but half as much".

What the extra two columns buy you:

• It matches how teams really change. Few practices deserve to be dropped outright; many deserve adjusting. Five columns let people say so without exaggerating.
• It makes disagreement visible. When one person writes "more of" for pairing and another writes "less of", you have found a real question about how the team works.
• It takes the drama out of Stop. "Less of" is a softer and often more accurate way to challenge a habit someone cares about.
• Its actions can be measured. "More" and "less" invite numbers: how much more, how much less, and how will we know the dial moved?

When to use it, and when not to

The Starfish works best:

• when Start Stop Continue feels too coarse and most cards land in Continue with a "but";
• for a team that has worked together for a while and has settled habits worth fine-tuning;
• after the team has grown, when rituals designed for five people start to strain with ten;
• when people feel overloaded — the Less of column is where the relief comes from;
• every few months as a calibration session between lighter formats.

Pick something else when:

• The team is new. Without established habits, More of and Less of stay empty. Start with What Went Well or Start Stop Continue.
• You have half an hour. Five columns need a full hour. For a quick directional check, SaMoLo keeps only Same, More and Less.
• The process needs a clear-out, not a tune-up. After a period of tool churn or process bloat, DAKI (Drop, Add, Keep, Improve) is built for cleaning house.
• A specific event went wrong. An outage or a missed launch needs the facts laid out first. Tuning practices comes after the team agrees on what happened.

The five columns

In ProstoRetro the board opens with Start, Stop, Keep, More of and Less of, in that order. The hard part of this format is the border between neighboring columns, so each section gives you the test that settles it.

Start: what's missing

New habits or experiments the team doesn't do at all yet. The test against More of: is the dial at zero today? If you already do it, even occasionally, the card belongs in More of.

• A ten-minute design critique before tickets go into refinement
• Recording architecture decisions in a short note next to the code
• A rotating "shield" who takes all interruptions for the day
• One sprint of trunk-based development on the mobile app, as an experiment

Stop: what drains value

Practices that cost time or energy and give little back. The test against Less of: if we stopped completely, would anyone miss it? If nobody would, it goes here. If someone would, it is a Less of card.

• The Monday metrics meeting that never leads to a decision
• Sending every UI text change through the full review process
• Keeping two task boards in sync by hand
• Promising dates to sales before the team has looked at the work

Keep: what still works

Proven practices worth protecting at their current level. The test against More of: would doing more of this actually help? If more would be better, it is More of. If the amount is right, it is Keep.

• Two reviewers on every database migration
• Product and the tech lead refine the backlog together on Tuesdays
• Release notes written by whoever shipped the feature
• No meetings on Friday afternoons

More of: what to double down on

Things that work and would work better if the team did them more often, more widely or more seriously. A good More of card says how much more, or at least where.

• Pairing — at least one paired ticket per person each sprint
• Talking to users: developers join one customer call a month, not only product
• Small pull requests; the ones under 200 lines got reviewed the same day
• Demos of work in progress, not only finished features

Less of: what to dial back

Useful things the team overdoes. This column is often the most valuable on the board, because overdone practices are hard to criticize — each one was a good idea once. The test against Stop: is there a smaller version we would keep?

• Build notifications in the team channel, including the green ones
• Planning poker on every ticket; rough sizes are enough for most
• Detailed specs for internal tools only we use
• Stand-up updates about work nobody else depends on

Use this template — it's free

A 60-minute Starfish agenda

Planned for five to nine people. The first time you run it, allow an extra 15 minutes for writing and grouping.

1. Warm-up, 5 minutes. A quick question from our icebreaker questions, or a round of Scribbles, ProstoRetro's draw-and-guess game, if the team needs waking up.
2. Review open action items, 4 minutes. More of and Less of actions fade fastest. Check whether the dial actually moved since the last retro.
3. Explain the borders, 2 minutes. One sentence each: Start versus More of, Stop versus Less of, Keep versus More of. It saves ten minutes of moving cards later.
4. Silent writing, 12 minutes. Cards can go on any of the five arms. Suggest people think about practices — meetings, reviews, testing, documentation, how work is planned — rather than the sprint's events.
5. Read and group, 9 minutes. Group by practice across columns, so every card about code review ends up in one place regardless of which arm it was written on.
6. Vote, 4 minutes. Vote for the practices you most want to change.
7. Discuss, 18 minutes. Take three or four topics. For Start and Stop, ask "what is the smallest experiment?"; for More of and Less of, "how much, and how will we know?"; for Keep, "what is most likely to break this?"
8. Action items, 4 minutes. Each gets an owner, a due date and, wherever it is about more or less, an amount.
9. Close, 2 minutes. Read the actions back. Ask everyone which single change they expect to notice first.

Running it with a remote team

• Write in two passes. Five columns on a laptop screen is a lot. Spend the first half of the writing time on Start, Stop and Keep and the second half on More of and Less of. Set the phase timer and announce the switch at the halfway mark.
• Hide cards until grouping. In ProstoRetro cards stay hidden from others while people write. That matters here because the More of and Less of columns are where opinions differ; you want each person's view before they see anyone else's.
• Check the AI's groups for opposites. ProstoRetro can suggest groups of similar cards, which you keep or undo. Look closely at any group that combines More of and Less of cards — those are the disagreements worth talking through.
• Settle a split with a poll. When the team is divided on a practice, an anonymous single-choice poll — "Code review: more, the same, or less?" — shows where the room really stands in under a minute, without the loudest voice deciding.
• Put numbers into actions. "Cap sprint planning at 60 minutes for the next two sprints — Priya" can be checked; "less planning" cannot. ProstoRetro emails the owner a week later and again before the due date, a natural moment to check whether the dial has moved.

Variations

A Starfish for the people you work with

Invite the people who depend on your team — support, a partner team, the product owner's stakeholders — and ask them to fill in the five arms from their side: what should your team start, stop, keep, do more of and do less of when working with them? When the room is open to anyone, guests join from the link with just a name. Treat it as a listening session: the team asks clarifying questions, and debates the answers later in its own retro.

The middle three arms

If five columns feel heavy, keep only the dial: Keep, More of and Less of. That is essentially SaMoLo, which is built in. It is a good fit for a quick check-in between full retros.

A quarterly Starfish

Run the full Starfish once a quarter and open the summary of the previous one before you start. Practices that moved from More of to Keep show that a change stuck. Practices that land in Less of again and again need a firmer decision, possibly a move to Stop.

Common mistakes

• Using More of and Less of as polite Start and Stop. "Less of the Monday meeting" when nobody wants any of it avoids a decision. Ask: would we keep any of it?
• Actions without amounts. "More testing" can't be checked. Agree how much more, and when you will look again.
• Ignoring Keep. Keep cards are the habits you lose first under pressure. Read them out even if they get no votes.
• Turning every dial at once. Five columns tempt the team to act on each one. Pick three changes; if you change everything, you won't know which change helped.
• Running it with a brand-new team. There is nothing to tune yet, and the session ends up as a slower Start Stop Continue.

FAQ

What are the five parts of a Starfish retrospective?

Start (new practices), Stop (practices to drop), Keep (practices that work at their current level), More of (practices to increase) and Less of (practices to reduce).

What is the difference between the Starfish and Start Stop Continue?

Start Stop Continue treats every practice as on or off. The Starfish adds More of and Less of, so the team can adjust how much it does something. Use the Starfish once most of your practices work and need tuning rather than replacing.

What is the difference between Stop and Less of?

Stop means nobody would miss the practice if it disappeared. Less of means it is useful but overdone, and a smaller version is worth keeping.

How long does a Starfish retrospective take?

An hour suits a team of five to nine. With a bigger team, or the first time you run it, plan 75 minutes: writing and grouping in five columns take longer than in three.

Can I run a Starfish retrospective online for free?

Yes. The Starfish is one of ProstoRetro's built-in templates, and the app is free. If five columns feel like too many for your team, build a smaller custom template with just the arms you need. For a full walkthrough, read how to run a great retro, or compare it with the other retrospective templates.