The problem: project bottlenecks
Project bottlenecks appear when dev marks Done but Testing queues 50 items—the bottleneck is Testing, not dev. Without CFD you miss the thick Review band. Managers hire more devs; queues worsen.
Bottleneck hunt = aging + CFD + WIP experiments.
Leaders who only see project bottlenecks in crisis meetings pay reactive costs—weekly leading metrics beat monthly firefighting. Tools alone do not fix it; policy plus cadence plus owner are required.
Symptoms and cost of project bottlenecks
project bottlenecks shows these signals:
| Symptom | Cost |
|---|---|
| Done up, cycle time up | Downstream choke |
| One column huge WIP | Named queue |
| Aging in one step only | Bottleneck |
| Hire upstream only | Worse queues |
| Local optimize | Global slips |
Unnamed bottlenecks drive hires in the wrong column—CFD before every capacity decision.
Bottleneck Hunt Framework (7 steps)
1. CFD baseline. flow_cfd bands. Execute this week—named owner on a task. One-line success criterion in wiki.
2. Aging by column. flow_aging SLA. Execute this week—named owner on a task. One-line success criterion in wiki.
3. Name top 10 aging. Owners. Execute this week—named owner on a task. One-line success criterion in wiki.
4. WIP limit experiment. Bottleneck column. Execute this week—named owner on a task. One-line success criterion in wiki.
5. Measure cycle time. Before/after 2 weeks. Execute this week—named owner on a task. One-line success criterion in wiki.
6. Escalate if policy. Reviewers, tools. Execute this week—named owner on a task. One-line success criterion in wiki.
7. Retro which band moved. Document fix. Execute this week—named owner on a task. One-line success criterion in wiki.
Leading vs lagging for project bottlenecks
| Leading | Lagging |
|---|---|
| Table symptom—weekly trend | Stakeholder surprise |
| Experiment with owner | Blame and firefighting |
| Metric from system | “Which number?” debates |
| Updated wiki policy | Repeat mistake next project |
Anti-patterns
- Blind hires at bottleneck
- Ignore CFD
- Optimize dev only
- No WIP limits
- Heroics to drain queues
How WKFGo helps with project bottlenecks
MCP flow_cfd.
MCP flow_aging.
Kanban WIP limits.
Reports—cycle time.
Sample project bottlenecks workflow
Monday: MCP brief before sync—same metric as the symptom table. Wednesday: aging or heatmap check if flow-related. Friday: if the experiment changed, one-line wiki update. project bottlenecks with steady cadence beats monthly workshops.
Scenarios
SaaS: QA choke. Hardware: Compliance review. Agency: Client approval wait.
In each case, project bottlenecks improves with the framework above.
Bottleneck—experiment
Print the CFD—thickest band = bottleneck. WIP limit only that column for two weeks. Measure cycle time before/after. If no improvement, fix policy blockers (approvals, tools) not headcount. Faster dev without QA capacity only grows Testing queues to 60 items—name the bottleneck in weekly review.
Project bottlenecks: hiring upstream without WIP limits only grows queues—CFD before HR requests. Virtual bottlenecks (legal sign-off) count too.
Summary on project bottlenecks
CFD before hires.
For product managers
WIP limit experiments. Check the metric in the next review.
For engineering
Weekly aging reviews. Check the metric in the next review.
For PMO
Document bottleneck fixes. Check the metric in the next review.
Teams that treat project bottlenecks seriously for two sprints have evidence before scaling to a second squad. Leadership must accept trade-offs—add without drop or local fixes repeats the same failure mode. Fifteen-minute weekly reviews on the same metric beat monthly workshops. Wiki the playbook that worked. New tools without policy repeat project bottlenecks with another dashboard. Revisit the symptom table each quarter—markets and teams change. Start small; evidence before mandates.
FAQ — project bottlenecks
Theory of Constraints? Yes—focus worst band. For project bottlenecks, without weekly cadence this question repeats every month.
Scrum bottleneck? DoD includes downstream. For project bottlenecks, without weekly cadence this question repeats every month.
Virtual bottleneck? Approval policy counts. For project bottlenecks, without weekly cadence this question repeats every month.
MCP? flow_cfd, flow_aging. For project bottlenecks, without weekly cadence this question repeats every month.
Project bottlenecks—start now
flow_cfd plus top ten flow_aging. One WIP limit experiment. Owner on bottleneck column.
Name your bottleneck
Run a small pilot this week.
Practical reminder — project bottlenecks
The most durable teams manage project bottlenecks with named owners, steady metrics, and fifteen-minute weekly reviews—not monthly workshops. When symptoms return, check policy and definitions first—not individuals. Pilot in one squad before PMO mandates. Wiki playbooks scale. MCP briefs before steering remove number debates. Leadership accepts one explicit trade-off each quarter—add without drop repeats the same cycle. Quarterly retro: did leading metrics improve? If not, change the experiment—not the tool. Starting small today beats big planning tomorrow.