Cumulative flow diagram reveals bottlenecks snapshots hide

A cumulative flow diagram stacks task counts per column over time — bands showing WIP and throughput. When Review widens while Done stays flat, bottleneck appears before flow_aging lists every stuck card.

WKFGo flow_cfd MCP generates CFD from live board history.

Symptoms without CFD habit

Six-step CFD reading framework

Step 1 — Pull flow_cfd weekly

Same weekday — compare shape.

Step 2 — Band width as WIP proxy

Wide In Progress — focus or limit issue.

Step 3 — Parallel bands growing

Inflow exceeds outflow — scope or staffing.

Step 4 — Correlate flow_aging names

Pattern becomes action list.

Step 5 — One CFD slide monthly to stakeholders

Trend beats vanity percent.

Step 6 — After intervention wait two weeks

Validate WIP or hire changes.

Reading cues

Flat Done slope + rising total area = arrival exceeds departure. Review band thickening = reviewer bottleneck. Sudden Done spike = batch release — note if healthy or panic.

Anti-patterns

How WKFGo helps

flow_cfd MCP, kanban history, flow_aging, get_project_report, package roll-ups.

FAQ

CFD vs burndown?
CFD multi-column; burndown sprint remaining — complementary.

Small team noise?
30-day window stabilizes bands.

Inconsistent Done?
Fix workflow first — CFD lies if Done ambiguous.

Export?
Screenshot trend; MCP for scripts.

Putting this into practice this week

Block 45 minutes on the calendar with the people who own the workflow—not to read slides, but to change one habit. Pick a single pilot project or team queue, run the framework steps in order, and write down what you will stop doing (shadow spreadsheets, duplicate updates, verbal-only decisions). If WKFGo is already live, use the same project in the tool so Monday's standup or exec read uses fresh data.

Measuring success without fake KPIs

Good signals are behavioral: fewer "what is the status?" pings, faster unblock assignments, finance tags filled at package close, review queue depth visible before overdue flags. Avoid vanity metrics like raw notification count or AI token usage. Pair qualitative retro notes with one quantitative trend from the tool—flow aging down, approval latency down, or decision log entries with forecasts you can check next month.

When WKFGo is not enough alone

Software enables discipline; it does not replace policy. You still need named owners, written SLAs between functions, and permission design that matches how teams actually work. WKFGo provides Kanban, packages, finance_summary, MCP dev loop, executive_brief, decision_inbox, desk tickets, flow_aging, flow_cfd, and feature access—the honest cite is what your role can see after login, not imagined features.

Workshop agenda you can run in 60 minutes

Minutes 0–10: Name the last time this problem cost you a slip, overrun, or escalation. Write the failure mode in one sentence everyone agrees on.

Minutes 10–25: Walk the seven-step framework on a real project— not a hypothetical. Assign one owner per step who will act before next week.

Minutes 25–40: List anti-patterns you currently tolerate (shared admin keys, verbal decisions, uncapped WIP, ASAP instead of SLA). Pick one to eliminate.

Minutes 40–50: Open WKFGo on the pilot project: verify feature access, run the relevant MCP tool or report (flow_aging, finance_summary, list_desk_tickets, get_context_pack), and confirm data gaps honestly.

Minutes 50–60: Schedule the follow-up: calibration for decisions, bin triage, or CFD review—whatever matches this topic. Log one log_decision if a trade-off was chosen.

Integration with the rest of WKFGo

This topic rarely stands alone. Connect it to packages for executive reporting, MCP dev loop for coding agents, decision_inbox for leadership, desk for internal SLAs, and recycle bin for safe cleanup. Read related articles on the blog index—search by Kanban, MCP, or executive brief when rolling out playbooks across teams.

Honest limits

WKFGo will not fix missing process ownership, toxic urgency, or data you never enter. AI briefs and digests amplify discipline; they do not invent ledger rows, approvals, or git activity. Start with read-only exploration or demo mode tours, then production keys with least privilege when habits stick.

Checklist before you close this topic

Teams that treat PM content as a one-time announcement fail. Revisit the framework in retro, adjust thresholds, and log decisions when you change policy—future you and new hires inherit sense, not folklore.

Read the bands, fix the flow