1. Define the project and the review boundary
Name the project, delivery period, and part of the process you want to understand. A tutorial launch may involve research, recording, editing, client approval, and publication; choose whether this review covers the whole sequence or a specific recurring difficulty. State who will read the notes before collecting feedback.
Atlassian’s retrospective playbook describes reviewing successes, problems, and possible improvements. That is a documented company practice, not independent evidence that a particular meeting will increase revenue or productivity. Use the idea as a starting point and keep the scope proportionate to your actual project.
Sources: [1]
2. Gather a small set of relevant evidence
Bring the agreed brief, delivery dates, revision log, and any production-time notes that were actually recorded. Identify gaps instead of recreating exact hours from vague memory. Keep private customer details and unrelated personal information out of the discussion materials.
Describe observations in terms someone else could check: two approval rounds were added after the original brief, or the final export used the wrong filename. Keep interpretations in a separate sentence. “The brief was unclear” may be a useful hypothesis, but it should not replace the concrete evidence needed to decide what to change.
3. Compare perspectives without assigning motives
Ask what helped the work progress, where a handoff became difficult, and what information was missing at the time. Let participants correct the timeline and explain constraints before agreeing on a cause. A delayed response does not establish why someone was unavailable.
For a solo project, write down the same distinctions for yourself. Separate a one-time interruption from a repeatable process problem. Preserve successful steps as deliberately as you identify friction, so a new workflow does not accidentally remove the part that already worked. Record disagreements as open questions rather than forcing a tidy narrative.
4. Choose one bounded experiment
Select a change that addresses the evidence and is within the team’s authority. For example, test a brief field that names the person responsible for final approval, or require the delivery filename to be checked during handoff. Avoid an action as broad as “communicate better,” which leaves the next step undefined.
Name the owner, the project or date when the change will be tried, and what observation will help assess it. Set a realistic boundary: the next two deliveries, one client workflow, or one template. Keep additional ideas in a short backlog rather than turning every comment into an immediate new obligation.
5. Review the result before making a larger claim
At the agreed review point, check whether the action happened and what was observed. Compare like situations where possible, and note other changes that could explain the result. One smoother handoff is useful feedback, but it does not establish a universal productivity gain.
Decide whether to retain, revise, or stop the experiment. Update the relevant project checklist only when the decision is clear, and keep a short rationale so the next collaborator understands it. This guide proposes an editorial workflow; it does not measure your team, require a paid planning product, or replace a formal workplace process.
Your quick checklist
- The project and readers of the notes are defined.
- Recorded evidence is separated from interpretation.
- Participants can correct the timeline and explain constraints.
- One bounded action has an owner and review point.
- Results are described without invented productivity claims.
Sources & scope
Atlassian provides an example of a company retrospective practice. ATN’s creator-specific evidence and experiment workflow is editorial judgment. No independent performance study, personal testing, or product endorsement is claimed.
- Atlassian Team Playbook: Retrospective (opens in a new tab)Source checked October 9, 2026
Published October 10, 2026. No affiliate links or paid placements in this guide.
