Retrospective Question Wheel: 8 Prompts for Agile Teams

When should a retrospective use a random question?

Use a random retrospective question when the team has already collected observations and several reviewed prompts could help it interpret the same evidence. Do not randomize the purpose, attendance, safety boundary, or final improvement decision. Those remain facilitator and team responsibilities.

The official Scrum Guide describes the Sprint Retrospective as a place to inspect people, interactions, processes, tools, and the Definition of Done, then identify helpful changes that improve effectiveness. A Wheel can introduce one neutral lens inside that work. It cannot make a vague meeting useful or guarantee that people feel able to speak honestly.

Skip the spin when a known incident, urgent risk, policy requirement, or unresolved harm needs direct attention. Use the appropriate incident, people, legal, security, or management process instead. A random prompt is for ordinary team learning after a low-stakes work cycle, not for deciding accountability or evaluating an employee.

Choose the phase before the prompt

A retrospective needs different questions at different moments. First gather observations, then generate insight, then decide what to change. Randomly’s quick-start Wheel contains only insight prompts, so use it after the team has described what happened and before it selects an action.

Do not put all four phases on one Wheel. A closing question selected before the team has gathered facts will feel premature, while another broad observation prompt at the end can reopen the whole meeting. Keep each list narrow enough that any result fits the work the group is doing now.

Review the eight quick-start prompts

The quick-start list contains eight process-focused prompts about outcomes, delays, assumptions, recovery, handoffs, useful practices, recurring patterns, and learning. They are deliberately general enough to use across many adult teams, but the facilitator should still remove any question that does not fit the completed cycle.

Read the complete list aloud before spinning. Remove prompts that duplicate the agenda, assume a problem that did not occur, or would expose a confidential matter in the room. Add a team-specific question only when everyone can understand it and answering it will not require names, case details, customer identifiers, security information, health information, or personnel notes.

Run one low-pressure discussion round

Show the reviewed prompt list, make one Selection, read the Winner aloud, then give everyone two or three quiet minutes to write. Invite observations rather than forcing a turn order. Participants should be able to contribute in speech, chat, a shared board, or another accessible route, and they may pass without explanation.

GOV.UK’s Service Manual recommends broad questions, an open atmosphere, and a concrete list of actions from a retrospective. The random step should support that structure, not turn the meeting into entertainment. If the selected question lands badly, pause and ask whether the list was wrong; correct the setup openly rather than quietly spinning until the facilitator gets a preferred prompt.

1. State the scope: name the completed work cycle and what this part of the retrospective should produce.
2. Review the eight prompts, remove unsuitable Entries, and explain that the Wheel chooses one shared lens rather than one person.
3. Spin once, read the selected prompt, and allow quiet individual writing before discussion begins.
4. Invite examples, group similar observations, and check whether the proposed pattern fits more than one piece of evidence.
5. Stop spinning when the group has enough insight to design one practical improvement experiment.

Turn the discussion into one owned experiment

A useful retrospective ends with a change the team can try, not a pile of interesting answers. Choose one small experiment that addresses the observed pattern, name an owner who will make the experiment possible, define a signal of improvement, and set a date to review what happened.

For example, if work repeatedly waited at a review handoff, the experiment might be a 15-minute review window at a fixed time for one cycle. The owner confirms the window, the signal is median waiting time or the count of items carried over, and the team reviews the result at the next retrospective. The Wheel does not select the owner or decide whether the intervention is appropriate.

Keep the action in the team’s normal backlog, board, or meeting record. Randomly stores Wheel Entries in the browser by default and is not a task tracker. Reset the prompt Wheel only when you intentionally begin a new insight round; do not treat repeated spins as progress when the existing discussion has not produced a decision.

Protect privacy and psychological safety

Use neutral process language and keep personal, confidential, regulated, and identifying information out of Wheel Entries. Do not record employee names, customer cases, incident details, credentials, health information, performance concerns, or private feedback in a Shared Setup or projected list.

Entries stay in the browser by default. A Shared Setup stores configuration in the URL fragment, which the Randomly server does not receive with the page request, but anyone with the complete link can inspect it. For a retrospective, the safest default is to load the public prompt set, edit it locally, and keep sensitive observations in the team’s approved system rather than in the Wheel.

A pass is not evidence of disengagement. Someone may need more context, a private channel, or time before responding. Do not use the Wheel to select who must disclose an experience, defend a mistake, receive feedback, or own a corrective action. Facilitation and organizational policy must govern those moments directly.