Identify Development Bottlenecks | Data-Driven Analysis
Find workflow bottlenecks using cycle time analysis, board observation, and team input. Common bottlenecks: code review, QA, deployment. Eliminate constraints with GitScrum.
6 min read
Bottlenecks are the hidden constraints that limit your entire team's output. A 10-person team with a 1-person code review bottleneck is effectively a 1-person team for review. Finding and eliminating bottlenecks is the highest-leverage improvement you can make.
Bottleneck Indicators
| Symptom | Likely Bottleneck |
|---|---|
| PRs waiting days | Code review |
| Bugs after release | Testing/QA |
| Deploy takes hours | Deployment process |
| Answers take days | Requirements/decisions |
| Work piles in column | That column's activity |
Finding Bottlenecks
Data Analysis
CYCLE TIME BREAKDOWN ANALYSIS
βββββββββββββββββββββββββββββ
MEASURE TIME IN EACH STAGE:
βββββββββββββββββββββββββββββββββββββ
Stage Avg Time % of Total
βββββββββββββββββββββββββββββββββββββ
Ready β Start 0.5 days 10%
Development 2.0 days 40%
Code Review 1.5 days 30% β Bottleneck?
QA Testing 0.5 days 10%
Deploy 0.5 days 10%
βββββββββββββββββββββββββββββββββββββ
Total 5.0 days 100%
ANALYSIS:
βββ Review is 30% of total cycle time
βββ Relative to development (40%), it's disproportionate
βββ Queue time in review = waste
βββ Focus improvement here
WEEKLY TREND:
Week 1: Review 1.2 days
Week 2: Review 1.5 days
Week 3: Review 1.8 days β Getting worse
Week 4: Review 2.1 days
Growing review time = capacity bottleneck
Board Observation
VISUAL BOTTLENECK DETECTION
βββββββββββββββββββββββββββ
LOOK AT YOUR BOARD:
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β Ready β In Progress β Review β QA β Done β
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β [T1] β [T5] β [T8] β [T14]β [Done] β
β [T2] β [T6] β [T9] β β [Done] β
β [T3] β [T7] β [T10] β β [Done] β
β [T4] β β [T11] β β β
β β β [T12] βββββββββββββ PILE β
β β β [T13] β β β
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
BOTTLENECK: Review column
βββ 6 items waiting
βββ More than other stages
βββ Age: Some 3+ days old
βββ QA is starved (only 1 item)
IMPACT:
βββ Developers finish but can't ship
βββ Waiting = inventory = waste
βββ Delays feedback to customer
βββ Frustration for team
Team Input
TEAM BOTTLENECK SURVEY
ββββββββββββββββββββββ
ASK THE TEAM:
βββββββββββββββββββββββββββββββββββββ
1. "Where do you wait most often?"
2. "What blocks you from shipping?"
3. "What process frustrates you?"
4. "If you could eliminate one wait, which?"
COMMON ANSWERS:
βββ "Waiting for code review"
βββ "Waiting for requirements clarification"
βββ "Waiting for design approval"
βββ "Waiting for environment access"
βββ "Waiting for third-party API"
βββ "Waiting for deployment window"
RETROSPECTIVE EXERCISE:
1. Each person writes top 3 waits
2. Group similar items
3. Vote on biggest impact
4. Focus on top 1-2
5. Create action item
Common Bottlenecks
Code Review
CODE REVIEW BOTTLENECK
ββββββββββββββββββββββ
SYMPTOMS:
βββ PRs open for days
βββ Developers waiting
βββ Context lost by review time
βββ Review backlog grows
ROOT CAUSES:
βββ Too few reviewers
βββ Large PRs (take long to review)
βββ No review SLA
βββ Reviewers are senior devs (busy)
βββ Review not prioritized
SOLUTIONS:
βββββββββββββββββββββββββββββββββββββ
1. SMALLER PRs
βββ Target: <200 lines changed
βββ Faster to review
βββ Faster feedback
βββ Less context loss
2. REVIEW SLA
βββ First review: 4 hours
βββ Track and report
βββ Make visible
βββ Hold accountable
3. MORE REVIEWERS
βββ Train junior devs
βββ Peer review (not just seniors)
βββ Review rotation schedule
βββ Automated review for style
4. PAIR PROGRAMMING
βββ Review built-in
βββ No separate step
βββ Ship faster
QA/Testing
QA BOTTLENECK
βββββββββββββ
SYMPTOMS:
βββ Items pile up waiting for QA
βββ Bugs found late
βββ QA rushed before release
βββ Manual testing takes too long
ROOT CAUSES:
βββ Manual testing only
βββ Too few QA resources
βββ Testing at end (not continuous)
βββ Unclear test criteria
SOLUTIONS:
βββββββββββββββββββββββββββββββββββββ
1. SHIFT LEFT
βββ Developers write tests
βββ Testing during development
βββ Not a separate phase
βββ Continuous integration
2. AUTOMATE
βββ Automated test suite
βββ Run on every PR
βββ Fast feedback
βββ Manual for exploratory only
3. ACCEPTANCE CRITERIA
βββ Clear before development
βββ Testable criteria
βββ Developers can verify
βββ QA confirms, not discovers
4. ADD CAPACITY
βββ More QA if needed
βββ Or developer testing time
βββ Balance team composition
Deployment
DEPLOYMENT BOTTLENECK
βββββββββββββββββββββ
SYMPTOMS:
βββ Deploy takes hours/days
βββ Manual steps required
βββ Limited deploy windows
βββ Waiting for "deploy person"
ROOT CAUSES:
βββ Manual deployment process
βββ Limited deploy access
βββ Infrequent deploys (scary)
βββ No rollback capability
SOLUTIONS:
βββββββββββββββββββββββββββββββββββββ
1. AUTOMATE DEPLOYMENT
βββ CI/CD pipeline
βββ Merge = deploy
βββ No manual steps
βββ Repeatable, reliable
2. DEPLOY MORE OFTEN
βββ Small changes = low risk
βββ Multiple times per day
βββ Deploy is normal
βββ Not an event
3. ENABLE ROLLBACK
βββ Easy rollback process
βββ Reduces fear
βββ Faster recovery
βββ More confidence
4. DEMOCRATIZE ACCESS
βββ Developers can deploy
βββ Not just ops team
βββ Self-service
βββ Remove bottleneck person
GitScrum for Bottlenecks
Cycle Time Analysis
GITSCRUM BOTTLENECK DETECTION
βββββββββββββββββββββββββββββ
CYCLE TIME REPORT:
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β Cycle Time Analysis - Last 30 Days β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β β
β AVERAGE: 4.5 days β
β β
β BY STAGE: β
β ββββββββββ Ready: 0.5 days (11%) β
β ββββββββββ Dev: 1.8 days (40%) β
β ββββββββββ Review: 1.4 days (31%) β FOCUS HERE β
β ββββββββββ QA: 0.4 days (9%) β
β ββββββββββ Deploy: 0.4 days (9%) β
β β
β INSIGHT: Review stage is disproportionately long β
β relative to QA and Deploy β
β β
β RECOMMENDATION: Implement review SLA, reduce PR size β
β β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
WIP BY COLUMN:
βββ Ready: 4 items (OK)
βββ In Progress: 5 items (OK)
βββ Review: 8 items (HIGH - investigate)
βββ QA: 2 items (OK)
βββ Done: --
ACTION: Add WIP limit to Review column
Best Practices
For Bottleneck Elimination
Anti-Patterns
BOTTLENECK MISTAKES:
β Optimizing non-bottleneck stages
β Adding capacity everywhere equally
β Ignoring team input
β Not measuring before/after
β Treating symptoms, not causes
β Multiple changes at once
β Not addressing cultural factors
β Expecting immediate results