Working Start something skill
Run a team retrospective
Facilitate a psychologically safer retrospective that turns shared evidence into owned improvement experiments.
Study the system and its conditions without erasing individual responsibility. The output is a small set of testable changes.
Procedure
- Define scope, participants, facilitator, privacy, timebox, and the decisions the group may make.
- Gather delivery evidence, outcomes, delays, quality signals, workload, customer impact, and prior action results.
- State participation and attribution rules, including how sensitive issues will leave the group forum.
- Build a shared timeline of observable events before discussing causes or judgments.
- Invite what helped, hindered, surprised, and remains unclear using multiple participation modes.
- Cluster themes and examine conditions, incentives, interfaces, controls, and decisions.
- Select a few changes by expected value, controllability, and learning potential.
- Write each experiment with hypothesis, owner, due date, signal, guardrail, and review date.
- Publish an appropriately scoped record and begin the next retrospective with prior results.
Failure plan
- Do not use the session for performance ranking, forced confession, retaliation, or public complaint investigation.
- Never promise anonymity the collection method cannot provide.
- If safety, harassment, discrimination, or misconduct emerges, stop group analysis and route it properly.
- Avoid action lists that add work without stopping or changing anything.
--- name: run-a-team-retrospective category: start description: Facilitate a psychologically safer retrospective that turns shared evidence into owned improvement experiments. Use when reviewing a delivery cycle, event, incident, or repeated workflow. --- # run-a-team-retrospective Study the system and its conditions without erasing individual responsibility. The output is a small set of testable changes. ## Procedure 1. Define scope, participants, facilitator, privacy, timebox, and the decisions the group may make. 2. Gather delivery evidence, outcomes, delays, quality signals, workload, customer impact, and prior action results. 3. State participation and attribution rules, including how sensitive issues will leave the group forum. 4. Build a shared timeline of observable events before discussing causes or judgments. 5. Invite what helped, hindered, surprised, and remains unclear using multiple participation modes. 6. Cluster themes and examine conditions, incentives, interfaces, controls, and decisions. 7. Select a few changes by expected value, controllability, and learning potential. 8. Write each experiment with hypothesis, owner, due date, signal, guardrail, and review date. 9. Publish an appropriately scoped record and begin the next retrospective with prior results. ## Failure plan - Do not use the session for performance ranking, forced confession, retaliation, or public complaint investigation. - Never promise anonymity the collection method cannot provide. - If safety, harassment, discrimination, or misconduct emerges, stop group analysis and route it properly. - Avoid action lists that add work without stopping or changing anything. ## Done - The timeline and themes are grounded in shared evidence - A bounded set of experiments has owners, measures, and review dates - Sensitive routing, published notes, and prior-action follow-up are confirmed