Ken Schwaber and Jeff Sutherland, in the November 2020 Scrum Guide, give the Sprint Retrospective a narrow job: plan ways to increase quality and effectiveness. The team inspects how the last Sprint went for people, interactions, process, tools, and their Definition of Done. They look for assumptions that led them astray and where those assumptions came from. They talk about what went well, what problems showed up, and whether those problems actually got solved. Then they pick the most useful changes and address the most important ones as soon as possible, even putting them on the next Sprint Backlog. The event is timeboxed, at most three hours for a one-month Sprint, shorter when the Sprint is shorter. It closes the Sprint. It is not a feelings circle with no decision, and it is not a status report aimed at an audience outside the team.
You may be one person with a deadline and no Sprint. The Guide is explicit that using scraps of Scrum is not Scrum. Borrow the inspection anyway, and do not call the result Scrum. Call it a two-week lookback: a short, scheduled comparison of what you said would be done and what is done, early enough that the external date can still hear you.
Why two weeks, not the night before
A retrospective the night before delivery can only change your apology. Schwaber and Sutherland write that adaptation has to happen as soon as possible to limit further drift, and that adaptation gets harder when people are not in a position to change the plan. Two weeks is long enough for a real chunk of work to have happened and short enough that a slipped assumption is still a conversation, not a postmortem. If your whole project is shorter than two weeks, look back at the halfway mark instead. The number is a cadence, not a ritual from a tablet.
The Guide's theory section is the reason a lookback without notes is theater. Empiricism, in their words, says knowledge comes from experience and from decisions based on what is observed. Inspection without transparency is misleading. If your only record of the last two weeks is a mood, you will inspect the mood. Moods are loyal to the story you like. Notes are loyal to the page count.
Four questions, answered from artifacts
Set a 25-minute appointment with yourself. Bring the plan you wrote at the start of the window and the files as they are now. Answer in writing.
- What did I say would exist by today, in objects? "Progress" is not an object. A drafted section, a sent invoice, a merged change, a recorded decision: those are objects. If you cannot point at the object, it does not exist yet.
- Which assumption from day one is now false? The Guide asks teams to find assumptions that led them astray and to explore where those assumptions came from. Yours might be "the data would be clean" or "I would have three clear mornings." Name the source of the assumption. If the source was a guess you never checked, that is the finding.
- What problem appeared, and did I actually solve it or only describe it? Describing a blocker in a notebook is not the same as removing it. The retrospective's third move is to ask whether problems were solved. Unsolved problems get a next action with a date, or they get escalated. They do not get a more eloquent description.
- What single change goes into the next window's plan? One. The Guide says the most impactful improvements are addressed as soon as possible. A list of twelve improvements is a way to address none of them. Put the one change where the next plan cannot ignore it: on the calendar, or in the first line of the task list.
Stop at 25 minutes even if question 4 feels thin. A lookback that expands until it becomes the work is just procrastination with a virtuous name. The timebox is the part of the retrospective worth copying even when you copy nothing else.
What early slip looks like
Slip is not a moral event. It is a gap between the object you named and the object on disk, while the external date has not moved. Caught at two weeks, the gap has three legal moves. Move the date. Shrink the object. Add capacity you actually control, which usually means removing something else, not promising yourself a heroic Saturday. Caught at two days, the same gap has one move: tell the truth late, or cut in a panic. The lookback exists to make the first set of moves available.
Here is an illustrative example, invented, of a lookback that earns its keep. At the start of the window you wrote "interview notes summarized and outline approved." On day 14 the notes are summarized and the outline is a heading list nobody has seen. Question 2's false assumption: "approval takes a day." The source of that assumption: hope. Question 3: the reviewer has not been asked. Question 4: send the heading list today with a yes-or-no question, and do not start the draft until the answer arrives. That is a smaller story than "rethink my productivity." It is also a change the next fortnight can feel.
If the false assumption is that other people will deliver on your internal date, do not solve it inside your own notebook. The handoff has a date of its own. Read working with other people's deadlines before you promise the outline is "basically done." If the false assumption is that the work was six hours because you wanted it to be six hours, the log on estimating without the optimistic film is the change to schedule, not another retrospective about feelings.
Keep it from becoming a second job
Schwaber and Sutherland place the retrospective at the end of a Sprint, after a review of the product with stakeholders. You can split those jobs without dressing up. Once a fortnight, five minutes to compare objects (question 1) is the review. Twenty more minutes for questions 2 through 4 is the retrospective. Neither meeting requires slides. If you catch yourself preparing a document about the lookback, you have started a new project. Close the laptop and answer question 4 in one sentence.
Share the sentence when someone else is waiting on the date. A lookback you hide is a private worry. A lookback you send, "the outline is not approved; I asked today; the draft starts after the answer," is transparency in the Guide's sense: the people receiving the work can see the state you are inspecting. They may not like it. They can act on it. A surprise two days out gives them neither.
Limits
A personal lookback will not repair a team that punishes bad news, and it will not replace the Scrum events if you are actually accountable for running them. The Guide's timeboxes and roles are for Scrum Teams, typically ten or fewer people, with a Product Owner and a Scrum Master. Solo use of four questions is a habit, not a certification. It also will not fix a scope decision nobody is willing to make. If question 4 keeps coming out as "work more," the assumption you have not inspected is that the object and the date can both stay. They often cannot.
What object did you promise would exist fourteen days ago, and is it in the folder?
The guide being borrowed
- Schwaber, K., & Sutherland, J. (2020, November). The Scrum Guide. Sections on empiricism, adaptation, and the Sprint Retrospective. scrumguides.org/scrum-guide.html