Retrospective Formats

Thinking Tools6.4 Conducting Follow-Up and Taking FeedbackAlso known as Start, Stop, Continue, Sailboat Retrospective, 4Ls

From Norman L. Kerth (Project Retrospectives, 2001), Esther Derby and Diana Larsen (Agile Retrospectives, 2006), and Luke Hohmann's Speed Boat game (Innovation Games, 2006).

Gives a team a rotating set of reflection templates, such as Start, Stop, Continue or Sailboat, each surfacing a different side of recent work and ending in owned actions.

Three calls that decide whether a retrospective changes anything

At each of the facilitator's turns, pick a move: the format, how ideas are gathered, and what counts as an action. The team's response shows with a verdict, and the line under the panel counts the calls that kept the review useful. Illustrative.

Facilitator

Team

Facilitator

Team

Select a stage to read it, or play the sequence.The sprint. A team of eight at a payroll software company has closed a sprint with a late release and a tax-table bug that cost three engineers their weekend. The room is tired and a little sore.Pick the format. The last four retrospectives ran Start, Stop, Continue, and the last two produced the same three notes. Pick the format for this one.Pick a move to see the reply.Stop releasing on Fridays. Continue pairing. Start writing tests earlier. Nobody mentions the weekend.A habit pick: the answers arrive ready-made, and the sore point stays unsaid.Mad: we heard about the tax change from a customer. Sad: the weekend. Glad: how support handled the calls.A fit: a sore team names the feeling first, and the cause comes out with it.Anchors: one tester for eight developers, criteria written mid-sprint. Rocks: the year-end tax tables.Also a fit: it finds the drag and the risk ahead, though the weekend stays unspoken.The mood is named. Once the frustration has a place on the board, people stop defending and start describing what happened.Gather the items. The tech lead has strong views about the testing process and usually speaks first. Pick how the items get onto the board.Pick a move to see the reply.The tech lead talks for six minutes about testing. Two people say nothing at all.An open floor: the first and loudest voice sets the agenda.Thirty-one notes. Four, from four people, say the vendor's tax notice went to a shared inbox nobody owns.Writing alone first: independent items, and a theme nobody had said aloud.Five clusters. The themes: one tester for eight developers, criteria written mid-sprint, the unowned inbox, Friday releases, and how well support handled the calls.Decide what to do. The unowned inbox is the theme everyone agrees on. Pick what goes on the board as the action.Pick a move to see the reply.Everyone nods. The line goes into the wiki.A lesson with no owner: next sprint, nothing is different.Two names and two dates go on the board beside the theme.Actions with owners: the next retrospective can check them.Next time. The next retrospective starts with these two items, done or not and why. A team that sees its actions get done keeps writing honest notes.

Calls that kept it useful: 0 of 3Calls that kept it useful: 1 of 3Calls that kept it useful: 2 of 3Calls that kept it useful: 3 of 3

  1. The sprint (Team). A team of eight at a payroll software company has closed a sprint with a late release and a tax-table bug that cost three engineers their weekend. The room is tired and a little sore.
  2. Pick the format (Facilitator). The last four retrospectives ran Start, Stop, Continue, and the last two produced the same three notes. Pick the format for this one.
    • Start, Stop, Continue again: everyone knows it. Stop releasing on Fridays. Continue pairing. Start writing tests earlier. Nobody mentions the weekend. A habit pick: the answers arrive ready-made, and the sore point stays unsaid.
    • Mad, Sad, Glad first, then a short round on actions. Mad: we heard about the tax change from a customer. Sad: the weekend. Glad: how support handled the calls. A fit: a sore team names the feeling first, and the cause comes out with it.
    • Sailboat: what pushed us, what held us, the rocks ahead. Anchors: one tester for eight developers, criteria written mid-sprint. Rocks: the year-end tax tables. Also a fit: it finds the drag and the risk ahead, though the weekend stays unspoken.
  3. The mood is named (Team). Once the frustration has a place on the board, people stop defending and start describing what happened.
  4. Gather the items (Facilitator). The tech lead has strong views about the testing process and usually speaks first. Pick how the items get onto the board.
    • Ask the room what went wrong and take notes. The tech lead talks for six minutes about testing. Two people say nothing at all. An open floor: the first and loudest voice sets the agenda.
    • Ten minutes of silent writing, one item per note. Thirty-one notes. Four, from four people, say the vendor's tax notice went to a shared inbox nobody owns. Writing alone first: independent items, and a theme nobody had said aloud.
  5. Five clusters (Team). The themes: one tester for eight developers, criteria written mid-sprint, the unowned inbox, Friday releases, and how well support handled the calls.
  6. Decide what to do (Facilitator). The unowned inbox is the theme everyone agrees on. Pick what goes on the board as the action.
    • We agree to communicate better about vendor changes. Everyone nods. The line goes into the wiki. A lesson with no owner: next sprint, nothing is different.
    • Priya routes vendor notices to the team channel by Friday; Sam adds a tax-table test. Two names and two dates go on the board beside the theme. Actions with owners: the next retrospective can check them.
  7. Next time (Team). The next retrospective starts with these two items, done or not and why. A team that sees its actions get done keeps writing honest notes.

An example

A team of eight at a payroll software company closed a two-week sprint with a release two days late and a tax-table bug that cost three engineers their weekend. The scrum master had run Start, Stop, Continue for four sprints running, and the last two had produced the same three notes. This time she opened with Mad, Sad, Glad, because the room was sore and she wanted that said before anyone wrote an action.

After ten minutes of silent writing the board held thirty-one notes. Four of them, written by four different people, said the same thing: the vendor had announced the tax change in an email to a shared inbox that nobody owned. Nobody had said it aloud in the standups. The team clustered the notes into five themes and kept two actions: the product owner would route the vendor's change notices to the team channel by Friday, and one engineer would add a tax-table test to the release checklist. Both were done when the next retrospective opened by checking them.

Why it works

The questions a team is asked decide what it says. Start, Stop, Continue pulls for actions, Mad, Sad, Glad for feelings, Sailboat for the forces pushing and holding the work, and the four Ls for what was learned and what was missing. Asked the same questions every sprint, people learn the answers that close the meeting fastest, so changing the format is a way of asking about the part of the work nobody has looked at lately.

The rest of the structure protects the honesty of the answers. Writing alone before anyone speaks keeps the first and loudest voice from setting the agenda, which is how four people who had each noticed the shared inbox found each other. Ending on owners and dates, and opening the next session by checking them, shows the team that what it writes changes something.

The thinking behind it

Norman L. Kerth Norman L. Kerth, a consultant who facilitated reviews for software teams, wrote Project Retrospectives: A Handbook for Team Reviews (2001), the book that gave the practice its name and its structure. His prime directive opens a retrospective by stating that everyone did the best job they could with what they knew at the time, so the review looks at the work and the system around it instead of at who failed.

Esther Derby Esther Derby and Diana Larsen wrote Agile Retrospectives: Making Good Teams Great (2006), the standard guide for teams that review their work every iteration. They set out five stages: set the stage, gather data, generate insights, decide what to do, and close the retrospective. Their catalog of activities gave facilitators several ways to run each stage, and it made the retrospective a regular, facilitated meeting held throughout a project as well as at its end.

Luke Hohmann Luke Hohmann, a product management consultant, described a dozen serious games for learning what customers need in Innovation Games: Creating Breakthrough Products Through Collaborative Play (2006). One of them, Speed Boat, asks customers to draw the anchors that hold a product back. Teams later turned the game inward and adapted it into the sailboat retrospective, which asks the same question of the team's own work: what is holding it back.

How to run it

  1. Check the last actions first. Open by asking which of last time's actions were done, and why the others were not.
  2. Read the room and pick the format. Use a feelings format for a sore team, Sailboat for a stuck one, and Start, Stop, Continue for a routine sprint.
  3. Set the ground rule. Say aloud that the review looks at the work and the system, and that nobody is on trial.
  4. Write alone first. Give everyone ten minutes to write one item per note before anyone speaks.
  5. Cluster and name the themes. Group the notes on the board and give each cluster a short name the team agrees with.
  6. Pick two or three to act on. Choose the themes worth changing this cycle and leave the rest on the board for later.
  7. Give each action an owner and a date. Write a name and a day beside every action, so the next session can check it.

Where it goes wrong

The meeting produces lessons and no actions. Communicate better, plan more carefully, and test earlier go into the notes and nothing changes. Spot it when an action has no name beside it, and refuse to close until each one has an owner and a date.

The same format runs for months and the answers turn ritual: the notes could have been written before the sprint started. Change the format, or ask the question the old one never asks.

Actions are written and never checked. The team learns that the meeting is theater and stops writing honest notes. Open every retrospective with the previous actions, and if most were not done, make that the first theme.

When to use something else

  • After-Action Review. When one event or operation needs review against what was planned, with a team that will do something like it again.
  • Blameless Postmortem. When an incident hurt customers and needs a written timeline, causes, and follow-up items others can learn from.
  • Double-Loop Learning. When the same problems keep coming back and the team's own goals and assumptions need questioning.

Where it sits in SOLVE eX

Go to the source

  • Project Retrospectives: A Handbook for Team Reviews, 2001
  • Agile Retrospectives: Making Good Teams Great, 2006
  • Innovation Games, 2006

Back to the catalog