How to Read a Cumulative Flow Diagram (And What a Broken One Looks Like)

Introduction

There are a lot of guides to the cumulative flow diagram and most of them stop in the same place. They name the axes, explain that each coloured band is a workflow stage, note that the bands should stay roughly parallel, and finish. You close the tab knowing what the chart is and no better at reading your own.

The value of a CFD is not in the healthy case. A healthy CFD tells you nothing you did not already know from a calm week. The value is in recognising the four specific shapes that mean something is wrong, because each one has a different cause and a different fix.

What Is a Cumulative Flow Diagram?

A cumulative flow diagram is a stacked area chart showing the number of work items in each stage of your process over time. Time runs along the horizontal axis and item count up the vertical. Each coloured band is one stage, and because the counts are cumulative the bands stack upward from done at the bottom to the backlog at the top.

Three metrics are readable directly off it. The vertical thickness of the in-progress bands is your current work in progress. The horizontal distance between the top of the done band and the line above it approximates cycle time. The slope of the done band is throughput. One chart, three of the four numbers on the flow metrics guide.

Reading the Healthy Shape

On a healthy chart every band climbs at roughly the same angle. Arrival rate matches departure rate, so the bands stay parallel and each one keeps a roughly constant thickness. Total height grows steadily because you keep accepting work, and the done band grows at the same rate because you keep finishing it.

Parallel bands are the whole test. You do not need to read values off the axes to know a team is stable, which is why the CFD is worth a glance even when you have no specific question.

The Four Broken Shapes

1. The widening in-progress band

The in-progress band gets thicker week over week while the done band's slope stays flat or drops. Items are being started faster than they are being finished.

This is the most common broken shape and the easiest to misread as busyness. WIP is climbing, so by Little's Law cycle time is climbing with it even though nobody is working less hard. The instinct is to push for more output. The correct response is to stop starting things: finish what is open before pulling anything new. See WIP limits and Little's Law.

2. The flat done band

The done band stops climbing for a stretch while everything above it keeps growing. Throughput has gone to zero. Nothing shipped.

Sometimes this is real and known, a holiday or an incident week. When it is not known, the usual cause is a definition-of-done problem rather than an effort problem: work is finished by the people doing it and blocked on something else before it counts. A release that only goes out fortnightly, a QA pass nobody owns, a client sign-off. The chart cannot tell you which, but it tells you to go look at the last step rather than the first.

3. The band that never appears

One of your stages is a hairline on the chart, or missing entirely. Items pass through it in minutes or skip it altogether.

That stage is not a stage. It is a column someone added during a process discussion and nobody uses. Delete it. This shape is worth looking for specifically, because unused stages are the main way boards accumulate ceremony that costs everyone a decision on every item for no return. It is the same failure described in prioritising without kanban columns.

4. The staircase

The done band climbs in flat steps rather than a smooth line: nothing for eleven days, then a vertical jump, then nothing again.

You are batching. Work is finishing continuously but only counted when a release goes out, so the chart is measuring your release cadence rather than your delivery. Two things follow. Your measured cycle time is inflated by an average of half the batch interval for every item, and your real risk profile is worse than it looks because a failure now involves eleven days of changes at once. See release notes strategy for the shipping side of this.

What the CFD Cannot Tell You

It hides individual items completely. A CFD showing a healthy, stable in-progress band is perfectly consistent with one item sitting untouched for six weeks while everything around it flows, because the band thickness stays the same either way. Aggregate charts average away exactly the thing you most need to act on.

This is why the CFD is a weekly review chart rather than a daily one. For the daily question, use work item age, which looks at each unfinished item individually and is the only flow chart that reliably produces an action the same morning you read it.

Do You Need One?

Honestly: a team of three or four probably does not. The CFD earns its place when you have enough items in flight that you cannot hold the state of the board in your head, which in practice means somewhere north of twenty concurrent items. Below that, you can see the widening in-progress band by looking at the board.

FlowBoard does not render a cumulative flow diagram. It tracks cycle time, lead time, throughput and work item age directly, on the view that for teams of two to fifteen those four answer the questions a CFD would, with less to interpret. If you want a CFD specifically, export your status transition history and build it in a spreadsheet; the inputs are the same three timestamps described in cycle time vs lead time.

What to Do This Week

If you already have a CFD, do not read the values. Look at the shape for five seconds and decide which of the four it most resembles. That is the entire exercise, and it is more productive than any amount of staring at the axes.

Then go to the metric behind whichever shape you found: WIP limits for a widening band, throughput for a flat done band, flow efficiency for a staircase.