translate Automatic translation · View original
Retrospectives that matter: when a retro is a ritual and when it is real improvement
When the Agile Manifesto was written in 2001, one of its principles stated something that seemed simple: "At regular intervals, the team reflects on how to become more effective, then tunes and adjusts its behavior accordingly." Only a few words, and the latent problem was already there: reflecting is not the same as changing.
The retrospective as a formal Scrum event was created to institutionalize that reflection. The Scrum Guide 2020 defines it as an opportunity for the team to "inspect itself and create a plan for improvements." Esther Derby and Diana Larsen, in their reference work Agile Retrospectives: Making Good Teams Great (2006), structured it into five phases with a logic that still holds today. The problem is not in the original design. The problem is in how that design gets implemented, wears down and, all too often, is emptied of content.
The question worth asking is not "Do we hold retrospectives?" but "What happens after them?"
How the ritual sets in without anyone noticing
The process by which a retrospective gets hollowed out is rarely dramatic. There is no visible breaking point. It is more like a silent erosion that Norman Kerth identified clearly when he warned about teams' tendency to seek quick conclusions and symbolic actions that relieve tension without committing to real change.
The most common pattern follows a recognizable sequence. In the first sprints, retrospectives generate enthusiasm. The team identifies genuine problems, makes decisions, and some of them work. There is a sense of control and agency. But over time, if the deeper impediments are not resolved, the team learns something dangerous: that improvement actions do not always lead anywhere. And when that happens repeatedly, the rational behavior is self-protection: keep taking part in the ritual, but without betting on it.
Patrick Lencioni (2002) describes how a lack of trust leads teams to avoid productive conflict. Applied to retrospectives, this translates into teams that talk about surface symptoms and systematically avoid root causes, which often involve naming interpersonal tensions, questioning management decisions, or structural problems the team cannot solve on its own.
Here lies the paradox: the more sophisticated the facilitated dynamic, the easier it is for it to serve as an emotional shield. A team can run a retrospective with the "six thinking hats" technique, background music and dot voting, and walk out without having touched anything that really mattered. Creativity in format does not guarantee depth of content.
Psychological safety: the condition that cannot be improvised
You cannot talk about effective retrospectives without talking about psychological safety, the concept Amy Edmondson popularized in The Fearless Organization (2018) and defines as the "shared belief held by members of a team that the team is safe for interpersonal risk-taking." In the context of a retrospective, that means any person can say "this isn't working" or "I think we made a mistake in this decision" without fear of retaliation, ridicule or being sidelined.
Edmondson's research showed that psychological safety was the most consistent predictor of team performance. Not the individual experience of its members, nor the clarity of goals, nor even the quality of leadership. The willingness to be vulnerable with one's colleagues.
What does this mean for retrospectives? It means that a retro in a team without psychological safety is, at best, an entertainment activity. At worst, it can be actively counterproductive: it creates an appearance of participation and openness that masks the fact that nobody is saying what they really think.
And this is where the facilitator faces their most honest challenge: psychological safety is not created in a single session. No retrospective dynamic, however well designed, can replace weeks or months of consistent behavior from the team's leaders. As Edmondson points out, psychological safety is built through everyday actions that happen outside the retrospective room.
The problem with improvement actions that improve nothing
One of the clearest indicators that a retrospective has become a ritual is the quality of its action items. And quality is not measured by quantity.
Lyssa Adkins (2010) identifies a pattern she calls "comfort actions": commitments the team can fulfill without real effort, which give a sense of progress without addressing underlying problems. "Improve communication between development and product" sounds good. But without specifying what that change means in terms of observable behavior, who is responsible, and how we will know it has happened, it is an intention, not an action.
Jeff Sutherland and J.J. Sutherland (2014) advocate limiting each retrospective to one or two improvement actions rather than producing long lists of improvements that are forgotten within three days. The constraint forces the team to truly prioritize, and that prioritization conversation is often, in itself, more valuable than the entire retrospective.
But there is an even more structural problem: some organizations lack the capacity or the will to resolve the impediments teams identify. The Scrum Guide 2020 states that the Scrum Master is accountable for "helping eliminate impediments to the Scrum Team's progress." However, when those impediments are organizational in nature, the Scrum Master finds themselves in a position of influence without authority. And teams know it.
When a team raises the same impediment in three consecutive retrospectives and sees no progress, it learns that raising that kind of problem is pointless. And it stops raising them. That is not cynicism: it is completely rational adaptive learning.
Distinguishing between the three types of retrospective that coexist
Not all retrospectives serve the same purpose, and treating them as if they do is another source of confusion. In practice, we can distinguish at least three types that coexist and require different approaches.
- The process retrospective focuses on how the team works: its technical practices, its workflow, its team agreements. It is the easiest to do well because the topics are relatively objective and the improvement actions are tangible. Are the Definition of Ready and Definition of Done up to date? Does the board reflect the actual flow? These are conversations the team can manage with reasonable autonomy.
- The relationship retrospective explores the team's human dynamics: how members communicate, how they handle conflict, how they make decisions together. It is considerably harder, requires greater psychological safety, and its facilitation cannot be improvised. Esther Derby and Diana Larsen devote a significant part of Agile Retrospectives to techniques for working at this level, recognizing that an unprepared facilitator can do more harm than good.
- The context retrospective attempts to address factors outside the team that affect its work: shifting organizational priorities, unresolved dependencies, governance problems. This type of retrospective often fails not because it is poorly facilitated, but because its conclusions require someone outside the team to make decisions. Without a clear channel for escalating those conclusions, the conversation is left hanging in the air.
Confusing these three types, or trying to address all of them in 90 minutes, is a recipe for superficial conversations on every front.
Signs that a retrospective has stopped improving things
How can you tell whether your team's retrospectives have crossed the line into ritual? There are observable signs that do not require deep analysis.
The first is the repetition of topics without resolution. If a problem appears in more than two consecutive retrospectives without any movement, something is failing. Either the improvement action was not specific enough, or the impediment has roots the team cannot resolve alone, or simply nobody truly took ownership.
The second is uniform, frictionless participation. This may sound counterintuitive, but a completely comfortable retrospective is often a sign that difficult topics are not on the table. Productive conflict, as Lencioni describes, is an indicator of health, not dysfunction.
The third is the disconnect between the retro's actions and the sprint's work. If, by the time of the next Sprint Review, nobody remembers what the previous sprint's improvement commitment was, that commitment was not real. Continuous improvement is not a separate activity from the work; it is built into it.
The fourth, perhaps the most revealing, is format fatigue. When the team starts developing strong preferences about the dynamic ("not this week, the six hats one is too long") and those preferences are disconnected from the topics that need addressing, the format is dominating the content. The dynamic has become the goal.
Recovering the purpose without destroying the team in the process
If you recognize something in the description above that resembles what your team is going through, the temptation is to make a dramatic reset: "From now on, our retros are going to be different." Experience suggests this rarely works.
Henrik Kniberg, whose work at Spotify and whose reflections on agile teams are widely cited in the community, has insisted that systemic improvements require small, persistent changes rather than grand declarations of intent. A retrospective is not rehabilitated with a special session: it is rehabilitated with a concrete action that the team sees actually happen before the next sprint, and then another, and another.
There are three interventions that tend to work for redirecting retrospectives that have become ritual, without requiring a traumatic change.
- Review previous actions at the start of each retrospective. Not as a formality, but honestly: what did we make progress on? What did we not? Why? This practice alone, if done candidly, changes the nature of the conversation. Teams that know their commitments will be reviewed tend to make more realistic and more specific commitments.
- Separate the impediments the team can resolve from those that need to be escalated. Diana Larsen and Ainsley Nies (2011) propose diagnostic tools that help teams distinguish between what is within their circle of influence and what is outside it. That distinction is liberating: it allows the team to act autonomously on what it can change and not feel defeated by what it cannot.
- Make the conversation about the retrospective itself explicit. What do we think of our retros? Are they serving any purpose? This meta-conversation makes many teams uncomfortable, but it opens the way to redesigning the format and purpose based on team consensus, not on the Scrum Master's unilateral initiative.
The role of the Scrum Master and the coach: facilitator or guardian of the format
There is a tension in the role of the person facilitating the retrospective that deserves attention. Lyssa Adkins describes in Coaching Agile Teams the difference between the coach as "mirror", who reflects what they see so the team can make decisions, and the coach as "expert", who knows what the team needs and steers it there. Neither extreme works well in retrospectives.
A facilitator who merely runs the dynamic without intervening when the team avoids difficult topics is not helping. They are, quite literally, facilitating the ritual. But a facilitator who steers the conversation toward where they believe it should go is not doing well either: they are using the retrospective to impose their own diagnosis, which undermines precisely the team autonomy the tool is meant to strengthen.
The balance is subtle: asking questions that open up space without dictating the answers. "What aren't we saying that would be important to say?" is a question many teams never hear in their retrospectives. "Why has this problem been on our board for three sprints without moving?" is another. These are uncomfortable questions. And sometimes the uncomfortable is exactly what is needed.
When the problem is not in the retrospective
It is worth being honest about something that is often left out of articles on improving retrospectives: sometimes the problem is not in the retrospective. It is in the organization.
The State of Agile Report, published annually by Digital.ai, has consistently documented that among the main obstacles to organizational agility are resistance to change, an organizational culture that is incompatible with agile values, and a lack of leadership support. These are systemic obstacles that no retrospective technique can resolve from within the team.
This does not mean giving up. It means calibrating expectations honestly and focusing energy where it can have a real effect. A team that learns to distinguish between "what we can change today" and "what requires a broader organizational change" is not a resigned team: it is a mature one.
Artificial intelligence enters the room: tool or new layer of ritual?
It would be difficult to write in 2026 about retrospectives without mentioning the arrival of artificial intelligence in teams' work processes. Tools that automatically transcribe sessions, identify patterns in language, generate summaries of the most recurring topics, or even propose improvement actions based on the history of previous sprints are moving from experimental to everyday use in some teams. The relevant question is not whether this is useful, but what happens to the underlying problems when we add a technological layer on top of them.
Let us start with what AI can honestly contribute. Analyzing patterns across past retrospectives is genuinely valuable: a team that has been running retros for twelve months has a volume of information about its own impediments that no human facilitator can systematically process between sprints. Tools that aggregate that information, identify recurring themes, or visualize how actions evolve over time can help the conversation start from real data rather than from the selective memory of participants. That is not a small thing.
The problem appears when AI starts to replace reflection instead of supporting it. If the team delegates to the model the identification of problems, the formulation of questions or the proposal of improvement actions, what happens is that the retrospective takes on an appearance of analytical sophistication that can be, paradoxically, more opaque than a board of sticky notes. Nobody questioned the results because "AI generated them." Nobody felt uncomfortable because the dynamic flowed smoothly and the summary looked great. The ritual has become more presentable, not more useful.
There is also a specific risk that deserves attention: language models tend to produce outputs that are coherent, well written and reasonable, regardless of whether they capture what actually happened in the room. A retrospective in which the most important tension was never explicitly verbalized will leave no trace in the transcript. The model will analyze what was said, not what was avoided. And what was avoided is, very often, the only thing that mattered.
AI is a tool, and like any tool, it amplifies what the team already has: if there is genuine reflection, it can strengthen it; if there is ritual, it can make it more sophisticated and harder to question. The question a Scrum Master or agile coach should ask before introducing any AI tool into their team's retrospectives is not "What can this do for us?" but "What are we avoiding doing ourselves that we expect this to solve?"
Conclusions
The retrospective remains one of the most powerful tools in agile teamwork. Precisely for that reason, it deserves more than becoming a well-executed ritual that changes nothing.
The difference between a retrospective as ritual and a retrospective as real improvement does not lie in the format, the chosen dynamic, or the facilitator's skill. It lies in whether the team has enough trust to talk about what matters, enough clarity to commit to specific actions, and an organizational context receptive enough for those actions to thrive.
The good news is that any of these three conditions can be improved. None of them requires a perfect transformation before you start. They require honesty about your starting point.
Where are your team's retrospectives right now? Are they becoming the space where difficult things get said, or the space where it is comfortable to say what is comfortable?
Bibliography
- Adkins, L. (2010). Coaching Agile Teams: A Companion for ScrumMasters, Agile Coaches, and Project Managers in Transition. Addison-Wesley Professional.
- Derby, E. and Larsen, D. (2006). Agile Retrospectives: Making Good Teams Great. Pragmatic Bookshelf.
- Digital.ai. (2023). 17th Annual State of Agile Report. Digital.ai. Available at: https://digital.ai/resource-center/analyst-reports/state-of-agile-report/
- Edmondson, A. C. (2018). The Fearless Organization: Creating Psychological Safety in the Workplace for Learning, Innovation, and Growth. Wiley.
- Kerth, N. L. (2001). Project Retrospectives: A Handbook for Team Reviews. Dorset House Publishing.
- Larsen, D. and Nies, A. (2011). Liftoff: Launching Agile Teams & Projects. Onyx Neon Press.
- Lencioni, P. (2002). The Five Dysfunctions of a Team: A Leadership Fable. Jossey-Bass.
- Rozovsky, J. (2015). The five keys to a successful Google team. Google re:Work. Available at: https://rework.withgoogle.com/blog/five-keys-to-a-successful-google-team/
- Schwaber, K. and Sutherland, J. (2020). The Scrum Guide: The Definitive Guide to Scrum: The Rules of the Game. Scrum.org. Available at: https://scrumguides.org
- Snowden, D. J. and Boone, M. E. (2007). "A Leader's Framework for Decision Making". Harvard Business Review, 85(11), 68–76.
- Sutherland, J. and Sutherland, J. J. (2014). Scrum: The Art of Doing Twice the Work in Half the Time. Crown Business.
Take part in the poll
This article includes an active poll in the Scrum Manager community.
poll View the poll in the community