Most retrospectives die by sprint 6. Same format every time, same surface complaints, no follow-through. The fix isn't a fancier format — it's an action-tracker, named owners, and a diagnosis when retros stall.
Three structural failures kill retros before sprint 10: actions don't get owners, the same format dulls the signal, and complaints recur because nothing changed last time. Fix any one of those and the retro starts pulling its weight again.
The five structural patterns that kill retro engagement, and the structural fixes for each. Not "rotate the format" — the actual problem.
Pick a format (start/stop/continue, 4Ls, sailboat, mad-sad-glad, KALM). Get a structured template with prompts and timeboxes. Copy as markdown.
The .md template. Owner per action, due-by sprint, status, and a review at the next retro. The single artefact that keeps retros from dying.
Reviews drift into demo theatre. Here's the format that keeps them honest — what shipped, what didn't, what the data says, what's next.
Disappearing retros are one of the six. Five other patterns to watch and the structural fix for each.
ReadRetro action follow-through is one of them. The four metrics worth tracking and the trap to avoid (Goodhart's law).
ReadIf standups are eating context, retro fatigue compounds. Async written standups + a focused retro is a higher-signal pairing.
ReadMost retros surface scope-creep complaints. Fix the upstream cause — three questions, four scenarios, the no-script that doesn't burn the relationship.
ReadSprintFlint has retros built in — templates per sprint, action-tracker with owners, carry-forward to next retro, and surfaces unresolved actions. The toolkit, but in your sprint workflow.