A retrospective can identify a recurring problem without changing anything. A kaizen card turns one friction point into a testable commitment: record the problem and its baseline, specify a change and owner, then name the measure and target. Keep the change small enough to try in the next sprint. section 10.3.1 section 10.7
Shorten pull-request reviews
Problem: reviews took a median of 2.8 days across 14 pull requests. Change: auto-assign reviewers with CODEOWNERS and add a four-business-hour review SLA. Owner: Priya. Measure: median review time next sprint. Target: under 1.5 days. The next measurement was 1.1 days, so the team made the change part of its standard work. section 10.3.1
Run the card as a short observe–change–measure loop:
- Start the retrospective by checking the previous card. Was its change made, and what happened? If it was not made, reconsider its size or ownership. section 10.3.1
- Choose one high-impact friction point, assign an owner, and implement the change early enough to observe it during the sprint. section 10.3.1
- Compare the result with the baseline at the next retrospective. Standardize a useful change or revert it; log the outcome either way. section 10.3.1 section 10.7
The same format works beyond sprint retrospectives. After an incident, a card can assign an alert improvement to an owner and set a diagnosis-time target. For a personal workflow, it can test whether a pre-commit hook eliminates lint-related CI failures while noting any friction the hook adds. A card captures a change one team can make; an A3 carries a broader proposal across teams. section 12.6.1 section 10.6
Table 10.4 A Kaizen log tracking sprint over sprint improvements
|
Sprint
|
Problem
|
Change
|
Result
|
Status
|
|
14
|
Reviews: 2.8 day median
|
Auto-assign + 4hr SLA
|
1.1 day median
|
Standardized
|
|
15
|
Flaky login test
|
Quarantined to the non-blocking stage
|
0 false failures in main pipeline
|
Standardized
|
|
16
|
Manual staging deploy
|
Automated via pipeline stage
|
Merge
to staging: 14hr → 22min
|
Standardized
|
|
17
|
Large PRs slow reviews
|
400-line soft cap with CI warning
|
Avg
PR size: 580 → 340 lines
|
Standardized
|
|
18
|
Unclear PR descriptions
|
PR template with checklist
|
Review start time: -40min avg
|
Standardized
|
|
19
|
Too many in-progress items
|
WIP limit of 2 per developer
|
Cycle
time: 4.2d → 3.1d
|
Standardized
|
|
20
|
Deploy rollback was manual
|
One-command rollback script
|
MTTR:
45min → 12min
|
Standardized
|
The kaizen log shows how cards become institutional memory: each entry pairs a problem and change with its measured result and status. section 10.3.1
