The problem: dependency webs nobody sees
Excessive dependencies appear when tasks chain together but never show on the daily board. The PM says "we're waiting on the API"—that API lives on another board, owner unclear, blocker age ten days with no comment.
Dependencies are normal. The issue is unnamed count and depth: when every task waits on three predecessors, a small slip freezes the whole flow.
Symptoms and costs
| Symptom | Cost |
|---|---|
| "Blocked" with no task link | Ghost wait |
| Gantt separate from Kanban | Two truths |
| Critical path unclear | Wrong priority |
| Long Review/Test queues | Hidden bottleneck |
| Cross-team verbal only | Surprise slip |
Main cost is wait time—not work time. Aging in Waiting columns shows where flow dies.
Framework: Link → Age → Sequence
1. Link — every blocker is an edge
Rules:
- Predecessor/successor in the daily tool
- Blocker title names external owner: "Blocked by: Platform API (Team Atlas)"
- One coordinator per dependency edge
Soft dependencies ("maybe we need design") become tasks or leave scope.
2. Age — flow_aging for stuck work
Flow aging measures time in the current status:
- Top oldest in Blocked / Review
- SLA comments on blockers
- Escalate when age exceeds column median
Overdue is late; aging alarms earlier.
3. Sequence — critical path and Gantt
For fixed-date releases:
- Gantt for timeline and slack
- Critical path = longest chain—swarm first
- WIP limits upstream of Review/Test
Kanban for daily flow; Gantt for "is the date still possible?"
Practical hygiene
Weekly 15-minute dependency review — blocked list sorted by age, not round-robin updates.
Close the loop — closing a blocking task notifies the blocked side same day.
Reduce depth — chains beyond four hops need split or parallelize.
Portfolio lens — cross-project blockers in one view; local green, global red.
Where WKFGo helps
WKFGo supports task dependencies, flow_aging analytics, Gantt timelines, and portfolio overview. It does not auto-fix graphs—link discipline and review cadence still matter.
Wiki context on dependency tasks helps newcomers understand why the edge exists.
When the dependency graph is too dense
Signal: more than half of tasks blocked or waiting. Remediation:
- Parallelize — is sequential work truly required?
- Stub/mock — unblock downstream with a temporary contract
- Split team — dedicated swarm on the longest chain
Negative Gantt slack means sponsor conversation—not more dev pressure.
Pair aging with standup
Three oldest blocked items each standup—owner pings SLA. Same blocker three days running → escalate coordinator, not author.
Cross-team dependencies without portfolio view
When Team A waits on Team B and each PM sees a local green board, executives only feel shock at release week. Portfolio overview should roll up open cross-project blockers—even when daily standups look fine.
Coordinator role: one human per edge—not "both PMs assume the other will ping."
Weekly dependency checklist
Each week PMs ask five questions: (1) blocked without links? (2) oldest blocker comment within forty-eight hours? (3) cross-project coordinator named? (4) negative Gantt slack? (5) policy fix from retro deployed? Fifteen minutes often prevents next week's cascade. Open portfolio view in the same ritual—local green/global red mismatches surface early.
Anti-patterns
- Dependencies only in email
- Blockers without follow-up dates
- "Almost Done" columns without limits
- Ignoring Test queues
Getting started this week
Sort your blocked list by age. For each item without a link: link it today. For each linked item with no comment in three-plus days: add an SLA comment. Next sprint rule: "blocked without link = invalid status."
Frequently Asked Questions
How many dependencies are too many?
If more than roughly thirty percent of tasks stay blocked over three days, fix the graph or capacity—exact thresholds vary. Trends beat magic numbers.
Gantt vs Kanban?
Kanban for daily flow; Gantt for date constraints. One truth—two views.
Cross-project?
Explicit links plus one coordinator per edge. Weekly portfolio roll-up.
Automation dependencies?
n8n/BPMN orchestration does not replace task links—human blockers still need edges.
Summary
Dependency graphs must live in the daily tool. Link → age → sequence catches cascade delays before release. Portfolio views are essential for cross-team blockers. PMs who enforce weekly dependency hygiene often avoid weekend hero firefights—invest fifteen minutes in ritual.
Start now
Export your five oldest blocked items. Each: linked? owner? SLA? Ship one policy fix this sprint.