A cumulative flow diagram (CFD) is a stacked area chart that shows how many work items sat in each stage of a workflow on each day. Read well, it shows work in progress, throughput, bottlenecks and scope changes on one chart, which is why kanban and Scrum teams use it to spot trouble while there is still time to act.
This guide explains how a CFD is built, the three signals to read, how to read WIP and cycle time straight off the chart, the common patterns and what to do about them, an active sprint example, and how a CFD compares with burndown and burnup charts. Most flow metrics tools and kanban tools draw one from your board history.
Key takeaways
Each band is a workflow state. Its thickness is how many items are in that state.
The vertical gap between the arrival and done lines is work in progress. The horizontal gap is roughly the average cycle time.
A steadily rising done line means steady throughput. A flat one means nothing is finishing.
A widening middle band points at a bottleneck. A rising top line means scope is growing.
The chart is only as good as the board data behind it.
What is a cumulative flow diagram?
The horizontal axis is time. The vertical axis is the cumulative count of items. Each colored band is one workflow state, such as To do, In progress, Review and Done, stacked with Done at the bottom.
The chart never removes history. Once an item reaches Done, it stays counted, so the done band only grows. That cumulative record is what makes patterns visible over weeks.
A CFD is different from a process flowchart. A flowchart shows how work is meant to move. A CFD shows how it actually moved.
Three signals to read
Most of what a CFD tells you comes from three things:
Band width
How many items are in that state. A widening In progress band means more is starting than finishing.
Slope
The rate of movement. The slope of the done line is your throughput.
Shape changes
Bulges, flat stretches and sudden jumps point at changes in scope, capacity, blockers or habits.
Read the top line first, then the done line, then the bands in between.
Reading WIP, cycle time and throughput off the chart
Daniel Vacanti's book Actionable Agile Metrics for Predictability shows how three flow measures sit on every CFD:
Work in progress
The vertical distance between the line where items arrive (enter the first active state) and the done line, on any given day.
Approximate average cycle time
The horizontal distance between those two lines. It shows roughly how long items take to cross the system.
Average throughput
The slope of the done line over a period, in items per day or week.
These are the three parts of Little's Law: average cycle time = average WIP ÷ average throughput. When the vertical gap grows and throughput stays flat, the horizontal gap, and the time items take, grows with it.
Common CFD patterns
Patterns are prompts for a conversation, not verdicts:
| Pattern | Likely meaning | What to try |
|---|---|---|
| A middle band keeps widening | Work is piling up in that state | Look at capacity, blockers, handoffs and the entry and exit rules for that column |
| The done line is flat for days | Little or nothing is finishing | Split work smaller, cut batching, swarm on nearly finished items |
| The top line rises fast | Work is being added faster than it finishes | Review intake rules and how the sprint or queue accepts new work |
| Bands are smooth and parallel | Work is flowing predictably | Keep monitoring and protect the current working agreements |
| Bands jump around | The workflow or the data is unstable | Check that statuses are defined and cards are moved consistently |
| Done climbs in steps at the end | Work finishes in a batch | Move review and testing earlier and finish items one at a time |
A middle band keeps widening
Work is piling up in that state
What to try: Look at capacity, blockers, handoffs and the entry and exit rules for that column
The done line is flat for days
Little or nothing is finishing
What to try: Split work smaller, cut batching, swarm on nearly finished items
The top line rises fast
Work is being added faster than it finishes
What to try: Review intake rules and how the sprint or queue accepts new work
Bands are smooth and parallel
Work is flowing predictably
What to try: Keep monitoring and protect the current working agreements
Bands jump around
The workflow or the data is unstable
What to try: Check that statuses are defined and cards are moved consistently
Done climbs in steps at the end
Work finishes in a batch
What to try: Move review and testing earlier and finish items one at a time
An active sprint example
This example is illustrative. A Scrum team uses five states: To do, In progress, Peer review, Testing and Done.
At the start of the sprint most items are in To do. By the middle, In progress is wide and only a few items are done. Near the end, Peer review and Testing swell, and Done jumps sharply in the last two days.
The team felt busy, and the CFD shows why that was misleading. Work started faster than it finished, and almost everything finished in a batch. One surprise in the last two days and several items would miss the sprint goal. A healthier chart would show items reaching Done throughout the sprint.
What the team might try next
Each of these focuses on finishing rather than starting:
- Cap work in progress: Limit how many stories can be in progress at once.
- Swarm near the finish: Pair on items that are close to done.
- Review and test earlier: Not only at the end of the sprint.
- Split large stories: Before they enter the sprint.
- Clarify acceptance criteria: Before development starts.
- Pause new starts: Until blocked or aging work moves.
Too much parallel work slows everything. Psychologist David Meyer, whose task-switching research the American Psychological Association summarizes, has said that even brief mental blocks from shifting between tasks can cost as much as 40% of someone's productive time.
Getting a useful CFD from Jira and other tools
Jira and most kanban tools build a CFD from status changes, so the chart reflects how carefully the board is kept:
- Meaningful statuses: Each one is a real state of the work.
- Cards moved when work moves: Late updates hide the real queue.
- Few columns: Too many statuses make the chart hard to read.
- Clear entry and exit rules: Everyone knows when an item is ready for review or done.
- No hidden work: Tasks tracked elsewhere never appear on the chart.
- Reviewed during the sprint: Waiting for the retrospective is too late to act.
CFDs in kanban, Scrum and at scale
The chart reads the same way, but the questions differ:
Kanban
The top line rises as new work arrives, which is normal. The question is whether arrivals and departures stay balanced. Use it with WIP limits and cycle time.
Scrum
The sprint backlog is planned, so a rising top line mid-sprint signals scope change worth discussing.
Programs and portfolios
A CFD of features moving through discovery, build and release shows delays between teams, where most large-scale waiting happens.
Keep the unit consistent. A chart mixing stories and features answers neither question well.
CFD vs burndown, burnup and flowchart
Each visual answers a different question, and teams often use two together:
| Chart | Best for | Misses |
|---|---|---|
| Burndown chart | Remaining work in a sprint or project | Can hide where work is stuck inside the workflow |
| Burnup chart | Completed work against total scope | Shows scope change, but not which stage is clogged |
| Process flowchart | How work is meant to move | Static unless paired with real data |
| Cumulative flow diagram | WIP, throughput, queues and stability | Needs accurate status data and careful reading |
Burndown chart
Remaining work in a sprint or project
Misses: Can hide where work is stuck inside the workflow
Burnup chart
Completed work against total scope
Misses: Shows scope change, but not which stage is clogged
Process flowchart
How work is meant to move
Misses: Static unless paired with real data
Cumulative flow diagram
WIP, throughput, queues and stability
Misses: Needs accurate status data and careful reading
A burndown can tell you a sprint is at risk. The CFD tells you where.
Common CFD mistakes
Most problems come from how the chart is used, not how it is drawn:
- 1
A report nobody discusses. A chart that never leads to a decision changes nothing.
- 2
Overreacting to one day. Look for repeated patterns, not daily wobbles.
- 3
Too many states. Every minor handoff as a column makes the chart noisy.
- 4
Comparing teams. Different work types, item sizes and policies make cross-team comparisons unfair.
- 5
Blaming people for system signals. A bottleneck usually points at a process constraint.
- 6
Stale data. If the board is not updated, the chart cannot be trusted.
A CFD improvement routine
- 1
Define the workflow. Choose states that represent real movement from request to done.
- 2
Clean up the board. Remove duplicate statuses and agree the rules for each column.
- 3
Read the current chart. Look for widening bands, a flat done line and a rising top line.
- 4
Pick one experiment. Lower a WIP limit, split stories smaller or review earlier.
- 5
Run it for a set period. Change one thing at a time.
- 6
Compare the pattern. Steadier finishing and thinner middle bands mean it helped.
- 7
Keep or adjust the agreement. Make useful experiments the normal way of working.
Frequently asked questions
What does a cumulative flow diagram show?
How do you read cycle time on a cumulative flow diagram?
What does a widening band mean on a CFD?
What is the difference between a burndown chart and a cumulative flow diagram?
Choosing Flow Metrics Tools?
Read the buying guide: what to look for, mistakes to avoid and questions to ask vendors.
Related guides
- WIP limits: why limiting work in progress makes teams fasterWhat WIP limits are, why limiting work in progress shortens lead time under Little's Law, how to set a first WIP limit and how to handle urgent work.Read the guide
- Actionable metrics vs vanity metrics, with examplesActionable metrics explained: how they differ from vanity metrics, the three A's, examples by team, flow metrics for predictability and a checklist.Read the guide
- Cycle time: definition, formula and how to reduce itCycle time explained: start and end points, the formula, cycle time vs lead time and takt time, percentiles, what makes it long and how to bring it down.Read the guide
- Cohort analysis: how to build and read a retention tableCohort analysis explained: cohort types, starting events, retention and churn metrics, how to read a retention table, a spreadsheet method and mistakes.Read the guide
Sources
Books and papers show their publication year. Web pages show the month we last checked them.
- 1.Daniel S. Vacanti, Actionable Agile Metrics for Predictability (ActionableAgile Press, 2015)
- 2.John D. C. Little, A Proof for the Queuing Formula: L = λW (Operations Research 9(3), 1961)
- 3.David J. Anderson, Kanban: Successful Evolutionary Change for Your Technology Business (Blue Hole Press, 2010)
- 4.Donald G. Reinertsen, The Principles of Product Development Flow (Celeritas Publishing, 2009)
- 5.American Psychological Association, Multitasking: Switching costs (apa.org, March 2006, checked October 2026)