The most useful information in a retrospective is rarely the feedback itself. It is who speaks, who stays silent, and how people talk to each other when something uncomfortable is on the table.
New leads pour their energy into the mechanical half of the job. Planning done, board tidy, every merge request reviewed. I did the same. What nobody put on my checklist was the other half, the one that decides whether any of the mechanics matter: how the people in the team are actually doing.
The good news is that you can observe team health with the same discipline you apply to a system.
Retrospectives are my first instrument, and they are not optional. Beyond the actions raised, they tell me whether people are willing to give feedback at all. Because if someone won't talk in a meeting designed for it, what else is going unsaid?
Spotify squad health checks are the second. Less deterministic, more honest, and they give you something factual to track over time rather than a vague sense that morale is fine.
The third is simply listening. How people speak to each other in meetings, who gets interrupted, whose ideas get built on.
None of these work as one-off exercises. Done regularly, they give you a baseline, and the baseline is what lets you spot drift early. Team health decays exactly like technical debt does. Quietly, gradually, and expensively if you only notice once something breaks.
Happy teams do better work. But happiness is not luck. It is something you monitor and maintain.