Running a Retrospective: A How-to Guide.

Ian Carroll
11 min readNov 2, 2020


Pillows arranged for group meditation

Let’s assume you know what a retrospective is (or well-enough),
and you know what Agile Development is (or well-enough).

And let’s assume that you’re sold on the idea of conducting a retrospective,
Or that you’ve agreed to conduct a retrospective for your team
without really knowing what you’re getting into.

Your mission: Get the team to self-organize and solve its own problems.
How? Cultivate Trust!

Your job as facilitator of the retrospective is to frame the team’s discussion so that there can be trust between the members.

As that is being achieved, you’ll want to focus their attention on the events of their last creative cycle (sprint, show, iteration, what-have-you), and get them to figure out what the root cause of their strengths and weaknesses are.

As that is being achieved, you’ll want to help them create experiments based on their realizations that enhance their strengths and control for their weaknesses. (Sometimes strengths are weaknesses and vice-versa, but the team will discover that over time.)

The Philosophy of How: It’s all in how you frame it.

“[The right] Structure equals freedom” — Stephen Nachmanovich
“The battle is won or lost in its preparation” — Sun Tsu
“As an archer, once your arrow is in flight, it is out of your control.”
— Some Zen Guy?

Most of your work as the facilitator will be in how you set the event up for success, how you communicate the retrospective’s objectives, games, and techniques to the team. Once the team starts going, it’s in their hands. You can’t correct them easily at that point, nor should you.

If you discover the team has not left the retrospective empowered and energized to make their work more effective. You are empowered to adjust the retrospective!

I have a potted orchid. Much like how my potted orchid wants to grow, your team wants to self-organize.

The reason my orchid doesn’t grow is because it’s not getting the right balance of sunlight. Or it’s potted wrong. I could be fertilizing it too much, or there’s something in the water, or some other environmental factor. Orchids can be fussy. Teams aren’t as fussy as orchids, but they still need the right conditions to flourish.

If you find your team isn’t getting the most out of their retrospective, that means the meeting’s conditions need to be adjusted. Maybe it’s how you framed the rules to them, or maybe it’s some outside disruption that needs to be controlled for. Give it an educated guess, adjust your execution, and try your adjusted retrospective next session.

Don’t look too closely, but if you watch how the meeting is going and then tweak the retrospective to work better for the team, then you’re retrospecting the retrospective! It’s turtles-all-the-way-down!

Some questions to ask yourself:

  • “How can I put people at ease to share their experiences with vulnerability and candor?”
  • “How can I put people at ease to listen to others’ experiences without judgement?”
  • “How can I make it okay for everyone to speak candidly even when they have wildly different positive and negative experiences?”
  • “How can I encourage the timid, or those frequently denied a voice in meetings, to speak up?”
  • “How can I get people excited to actively participate?”
  • “How can I protect the session from well-meaning over-involved authority figures who might unintentionally rob the team of its opportunity to learn about and improve itself?”

Some Caveats: The first well-meaning over-involved authority figure to protect the group from is yourself.

“You’re the guide on the side; not the sage on the stage.” — Jane Sandmeir, Tour Manager at The Huntington Library, Art Museum & Botanical Gardens

It doesn’t matter if you know what the group needs to improve. The group needs to know. Years of experience has taught me no amount of telling the group will let them know. They can only know if they discover it themselves. If telling people worked, the project director would hand out notes and we wouldn’t need a retrospective. Besides, maybe you don’t know!

Instead, read the room. Observe the team. Is someone talking a lot? Is it because they’re nervous or because they have authority over the team? Is someone else quiet? Is it because they’re afraid to speak up or because they can’t get a word in edgewise? Those are things to bring to the group’s attention if they’re in a trusting place.

If they’re a new team, be less direct. New teams haven’t had the opportunity to establish trust yet. Even with experienced professionals, trust needs to be developed. Never assume trust exists. If your team is new or hasn’t yet been given the opportunity to trust each other, instead of coaching directly and publicly, give notes privately afterwards to share your observations, or rethink how you frame the session next time to give the team a chance to establish trust.

The key thing is to make sure everyone gets a chance to freely participate and contribute their point of view.

An Example One-Hour Program:


  • Make sure whatever scheduling software is being used has the information the team needs to prepare in it.
  • In the schedule notes, make sure the video conferencing link and retro app link is included.
  • There are many retro apps, but here are 2 of my favorites:,
    Google it. I’m always hunting for better tools.
  • Post in your team’s chat app the day before to get people excited, get them thinking, and start setting their expectations. This is part of “winning the battle before you start fighting”, as Sun Tsu might say.
  • Feel free to post the links for the conferencing app and any retro app links in your team’s chat app 15 minutes prior to the event. With computer code, repeating ourselves adds confusion. But with human dynamics, repeating ourselves frequently adds clarity. Make it easy for the team. Repeat yourself in more than one channel.


5 minutes — Warm-up/Ice-breaker
5 minutes — Recite the Prime Directive and Reiterate the Rules
5 minutes — Idea Generation
5 minutes — Grouping and Clarification
5 minutes — Dot Voting
25 minutes — Root Cause Analysis and Action Item Generation
5 minutes — Appreciations

But that’s only 50 minutes! You’re spot on. At least 10 minutes will get eaten by team members arriving late, or segments running too long. Over time as people realize that a lot gets done in retros, and they’re enjoyable to do, they’ll start arriving earlier. When that happens, put the extra time in Root Cause Analysis.

5 minutes — Warm-up/Ice-breaker: Post GIFs in the group chat representing the week, play a quick round of Exploding Kittens, play a fast-paced improv game, group-draw a picture where each person in turn adds one stroke of the mouse. The objective is to get people out of their routine, thinking about something new, and allow them time to drop their emotional guard and be candid with each other without fear of recrimination.

5 minutes — Recite the Prime Directive and Reiterate the Rules: This is the crucial step. If you get this right, the rest of the session will go well. To encourage participation, ask for a volunteer to recite the prime directive out loud. This is about improving processes. Finding blame does not improve processes.

This is also the time to reiterate the objective:
Generate experiments to enhance the team’s strengths and control for its weaknesses.

and the time to mention helpful guidelines:
Keep the Cameras On: They don’t have to be on all the time. Camera fatigue is a real thing. But in general, encourage participants to show themselves. It helps develop trust, helps with non-verbal queues, and allows the team to connect.
Yes-And: Encourage participants to support each other’s ideas and experiences. That can be challenging when talking about technical things. Sometimes things said will just be incorrect. But in order to proceed efficiently when generating ideas, sharing experiences, and especially when committing to action items, a lot of time is saved if the team doesn’t negate each other. Instead, try to build on others’ ideas or make adjustments as opposed to outright denials by thinking or saying “Yes, and…” at the beginning of your statement. Sometimes saying “No-but” is necessary, but it should be understood it’s a destructive and wasteful operation. Use “Yes-and” where possible; use “No-but” sparingly.
The Match is Burning: Imagine there’s a match that you just lit and you are holding it. What you have to say should be done before the flame burns your fingers. The reason behind this is to encourage people to say just what they need to say and not repeat themselves. This saves time and allows others a chance to speak or respond.
You’ll find your own pithy sayings and rules in time. Make up your own too.

This is also an excellent time to ask for a volunteer to use their phone to keep track of time here on in (this adds one more engaged participant).

5 minutes — Idea Generation: The app you’re using will probably help you from this point forward. If you don’t see anyone putting any ideas in, you’ll have to encourage them, but I’ve rarely encountered this situation. Usually you end up with more ideas than can be talked about, which is fine. Ideas that aren’t mentioned are still seen, and will still sit in the back of the minds of the team. Some teams like to be able to see other people’s ideas as they come because it sparks other ideas, other teams prefer to keep it secret so that each person’s responses are not influenced by others. Try it each way, see what you like. If you’re behind on time, you can try limiting this section to 2 or 3 minutes.

5 minutes — Grouping and Clarification: Most apps have a way to do this. If they don’t you might have to find some way to do it manually. It is an important step because it’s the beginning of finding root causes. It gives the team a chance to find links between what might be very different observations. It’s also a good time for the team to ask non-debatable clarifying questions. Just make sure the only questions being asked and answered are just to clarify meaning rather than diving into a discussion. This is also a good time to remind the team about the burning match rule to keep the explanations from getting too involved and taking up time.

5 minutes — Dot Voting: This is a simple way to let the team decide what are the most pressing issues. Each person gets 3 “dots” to vote with. Some variants give each person 5 or 7 dots, so balance it the way that makes sense for your team. Each person can choose to spread out their dots as they see fit. They can put all 3 on one issue because it’s really pressing to them, 2 on one; 1 on a second, or 1 on three different issues. This style of voting gives participants the chance to express intensity in what issues matter. That way one member’s pain can’t be ignored by the rest of the group if it’s not pressing to anyone else. If you’re behind on time, you can try limiting this section to 2 or 3 minutes.

25 minutes — Root Cause Analysis and Action Item Generation: You’ll want to give the team the maximum amount of time possible for this. The whole point of all of the framing in the previous segments was to set this discussion up for success. Here is the point where you’ll want to step back and let the team chart its own course. Note who is participating, who is remaining silent, read the room and occasionally make small comments to side-coach the team toward a healthy dialogue. The less comments you make the better, so only step in if it’s necessary. In order for the team to self-organize, they need to be the ones taking the action, not you.

To help keep them on track, a little more framing at the top here might be necessary. Now that you have your topics sorted, understood, and dot-voted, take whatever got the most dots and focus the group’s attention on that topic. Remind the team that their objective is to generate tweaks and experiments in the way they work to try out for the following work cycle. In order to do that, their discussion should revolve around finding a root cause for the topic, which may require repeatedly asking “why”, never stopping on a reason that puts blame on any of the team members. We’re looking for the root cause in the team’s process. The company should have another mechanism for performance reviews, hiring and firing.

It is also helpful to remind the team of the concept of “Yes-and” here, and the burning match rule if you see people denying others’ suggestions or repeating points that have already been made.

Ask your appointed time-keeper to set an alarm for 8 minutes. Once 8 minutes is up, have the group vote by a quick thumbs-up or thumbs-down whether they need to keep talking about this topic, or move on to the next. This also helps prevent discussions with a long tail where nothing new is discussed.

By the end, there should be 1 to 3 new experiments or action items for the team to try. Have these enacted immediately, and remind the team that all experiments are fair game next retrospective for refinement or pivoting.

5 minutes — Appreciations: This can often be thought of as a cool-down, and if you’re feeling creative, you may find the team will feel more energized by ending in a game rather than warm affirmations. It depends on the team, so try different variations. The basic form of a cool-down is having team members share appreciation for their fellow team members when those team members had a positive impact on them. Or to celebrate team victories. There’s usually one or two of these that end up in the idea generation that will not get much discussion time. Now, at the end of the meeting, is a good time for the team to leave on an up-beat feeling ready, and capable to tackle the next cycle’s creative work.


  • Follow up with anyone if they participated a lot or a little, or had something to share that didn’t get resolved.
  • Solicit feedback, and use that along with your own self-reflections to adjust your framing for the next retro.
  • Apply the changes, and repeat.

Over a few iterations of this retrospective, the team should start to take more and more ownership of the process, your coaching will be needed less and less. As that happens, you’ll want to give each team member the opportunity to run the retrospective. Your long-term goal is to remove the team’s dependence on you for their process improvements. As you step back, the team should be charting their own course, optimizing their own workflows, and truly self-organizing. If you are also a team member, once the team reaches this stage, step down and enjoy being a part of a high-functioning team. You’ve earned it.

If you’re not an official part of the team, you’re probably on to a new team that needs some help self-organizing. Still, give yourself a treat. It’s not every day a high-functioning team is born.



Ian Carroll

Software Crafter at 8th Light Consultancy, Organizer for Fullstack LA meetup, Eagle Scout, Theatre Person, Taoist Philosopher among other passions.