A retrospective is a regular team meeting for examining recent work and deciding what to improve. It can make planning or standups better, but those conversations may miss where delivery actually stalls. A team might meet every sprint goal while pull requests wait for review, staging deploys remain manual, and work queues at a QA handoff. Look beyond team rituals to the value stream: the path from request to production. section 10.3
Table 10.3 Retrospective versus value stream focus
|
Typical retro focus
|
Value stream focus
|
|
Sprint planning took too long
|
Tickets wait 5 days in the backlog before work starts
|
|
We didn't finish the sprint goal
|
Lead time increased from 8 to 14 days this quarter
|
|
Stories need better acceptance criteria
|
40% of activity time is spent reworking after review
|
|
Standups feel pointless
|
The review queue averages 5 items—more than any other
column
|
|
We should estimate better
|
Cycle time P90 is 3x the median. What's trapping the slow
items?
|
Compare a complaint about a sprint event with a question about waiting, rework, lead time, or a review queue. Both may matter, but only the latter directly tests where work stalls. section 10.3
Keep the improvement loop small enough to finish:
- Check the previous improvement first: was the change made, measured, and useful? section 10.3.1
- Choose one high-impact source of waiting, rework, or delay that the team can address next sprint. section 10.3.1
- Record the problem, baseline, specific change, owner, measure, and target. Compare the result afterward rather than assuming the change helped. section 10.3.1
Shorten the review queue
If 14 pull requests show a median review time of 2.8 days, try automatic reviewer assignment through CODEOWNERS and a four-business-hour review target. Assign an owner and measure median review time next sprint against a target below 1.5 days. This is more testable than “improve reviews.” section 10.3.1
Do not turn the meeting into a long list of unowned wishes. Keep a record of changes and results so improvements can be retained. When a bottleneck crosses team boundaries—such as incompatible message contracts—ask why the divergence went unnoticed and escalate the fix to whoever owns the affected value stream. Distributed teams can collect observations asynchronously before a facilitated discussion. section 10.3.1 section 11.2.2 section 13.7.2






