The best retrospectives end with one change
"That was a good retrospective," someone says as the call ends. The board is full of notes; everyone spoke, and the team has a list of nine action items. It feels productive. Two weeks later, most of those items haven't moved, and the same complaints return in slightly different words.
The common assumption is that a retrospective is measured by what it produces: honest talk, many ideas, a long list of improvements. In reality, it's measured by what changes afterwards. A retrospective that changes nothing is only a well-organised conversation about the past.
Long lists feel responsible. Every problem raised gets an action, so nobody feels ignored. But improvement work competes with the sprint for the same hours. When the next sprint starts, the tickets win, because they have owners, deadlines and someone waiting for them. The action items have none of that. "Improve test coverage" and "communicate better with the client" quietly fade. Nobody decides to drop them. They simply stop being mentioned.
The cost goes beyond the missed improvements. When developers raise the same issue three sprints in a row, and nothing happens, they stop raising it. The sessions get shorter, and people start to see the retrospective as one more meeting to sit through.
Good teams do something that looks less ambitious. They choose one change, sometimes two, and make it small enough to try in the next sprint. Instead of "review pull requests faster", they agree that no pull request waits more than one working day. The change has an owner. It lives where the team already works, often as a ticket in the sprint itself, so it competes for time openly instead of hoping for leftovers.
Then comes the step most teams skip. The next retrospective opens with a simple question: did the change help? If it did, it becomes part of how the team works. If it didn't, the team adjusts it or drops it, and that still counts as a useful result. A failed experiment teaches something. A forgotten action item teaches nothing.
Choosing one change means leaving real problems on the board, and that can feel uncomfortable. Some issues are also too large for one sprint or sit outside the team's control. Those deserve honesty: name them, pass them to whoever can act on them, and stop pretending the team will solve them alone. The rest can wait. If a problem truly matters, it will come up again.
Over a year, this adds up. A team that makes one real change each sprint slowly becomes a different team. Teams with long lists usually end up with longer lists.
A retrospective doesn't end when the call does. It ends when something works differently.


