Retros fail in a specific way: same three questions, same three answers, week after week, until the ritual means nothing. The fix isn’t a better meeting. It’s a better artifact — short, rotating, and obligated to produce exactly one change.
Here’s the async retro template we use. It takes each person about ten minutes, lives in the workspace next to the board, and has survived four months without going stale. Steal it.
The template
Posted Monday morning, answered by end of day, read by everyone before Wednesday.
The three fixed questions (every week)
1. What slowed you down this week? Two sentences max. Not “what did you work on” — what slowed you. Waiting on review, unclear ticket, flaky test, a decision nobody made. This is the only question that never rotates, because friction is the thing retros exist to find.
2. What’s one thing you’re unusually sure of right now? Borrowed from the observation that teams drown in uncertainty and never inventory their confidence. “The new onboarding flow is right.” “That dependency is going to bite us in Q3.” Confidence stated in writing becomes checkable later — the best early-warning system we’ve found.
3. The experiment. One thing the team will do differently next week. One. Not a list. “All PRs under 300 lines.” “No meetings Wednesday.” If last week’s experiment is still running, it gets one line: kept, killed, or still judging.
The two rotating questions (pick a pair each week)
Rotate these so the muscle doesn’t atrophy:
- What did we ship that nobody will notice, but matters?
- What conversation are we avoiding?
- What would a new hire find weird about how we work?
- What did we measure this week that changed a decision?
- What’s the oldest open ticket, and why is it still alive?
- If this week were a postmortem, what would it be for?
The rules that make it work
Two-sentence cap. This is the load-bearing rule. Paragraph answers don’t get read; two sentences do. Anyone who needs more room starts a thread.
The author rotates. Each week a different person picks the rotating questions and writes the one-paragraph summary at week’s end: friction found, experiments running, confidence checks. Rotating authorship keeps it from becoming management’s report.
The experiment is sacred. A retro that produces no change is a diary. One experiment, reviewed next week, is the entire improvement engine. Small and real beats ambitious and abandoned.
A filled-in example
1. Slowed me down: Waited 26 hours on a review for the billing PR; the spec for the retry logic was ambiguous and I guessed wrong once.
2. Unusually sure of: The queue-based approach for webhooks is right — it survived the traffic spike without drama.
3. Experiment: Review response under 4 working hours, measured by the board, judged next Monday.
Rotating — oldest open ticket: K-411, the CSV export edge case. Alive because it’s ugly and small. Proposal: timebox it Friday or close it.
That’s it. Ten minutes, readable in ninety seconds, and it produced a concrete change plus a decision about a zombie ticket.
Getting started
Post the three fixed questions plus a rotating pair in whatever tool your team already lives in. Cap answers at two sentences. End with one experiment. If you want the artifact next to the work itself, that’s the shape of workspace we build — but honestly, the template works in a shared doc. The discipline is the tool.
👉 Board, whiteboard, and live GitHub activity in one workspace — Kahoona is free to start: kahoona.app
Related: An Async Standup Template: Five Prompts That Work · The Weekly Planning Ritual for Small Dev Teams