Written by the AgileFever editorial team. Facts checked on November 10, 2026 against the official pages of the organizations named in this post. AgileFever sells training for some of the topics covered here, and we say so wherever it applies. The data set in this post is invented and labelled as such. Definitions come from the Kanban Guide, version 2025.5.
Your tool draws kanban reports for you, but a chart you cannot read changes nothing. This post shows how to read each report, what to do when it looks wrong, and which one to open for which question. One invented data set runs through all of it, so the charts and the numbers always agree.
Kanban reports and what the Kanban Guide says about them
The Kanban Guide from Kanban University (version 2025.5, updated May 1, 2025) defines four flow metrics: work in progress, cycle time, work item age and throughput. It leaves the team to decide how to use them, for example to visualize them in charts and assess variation. It does not name a cumulative flow diagram, a control chart or a scatterplot, and it does not say how they overlap. Those chart types come from Kanban practice, so tool labels differ.
The overlap is clear from the definitions. Every metric needs only two dates per item, start and finish, and the cumulative flow diagram adds the date each item entered each column. All the kanban reports below are views of those dates.
| Report | Question it answers | Built from | Main metric |
|---|---|---|---|
| Cumulative flow diagram | Where does work pile up? | Daily count of items that reached each column | WIP, arrival and departure rates |
| Cycle time scatterplot | What can we promise? | One dot per finished item | Cycle time, as percentiles |
| Control chart | Did something change, or is this noise? | The same dots, plus a mean and limits from a stable baseline | Cycle time over time |
| Aging work in progress | Which open items are in danger? | Start date of each unfinished item | Work item age |
The example data: 46 items over 12 weeks
A team pulls items through three states: in progress, in review and done. In twelve weeks it started 46 items. 40 are finished and 6 are open on the last Friday. Cycle time runs from the start day to the finish day in calendar days, and the average is 6.1.
| Week | Started | Finished | Open at Friday (WIP) | Of those, in review |
|---|---|---|---|---|
| 1 | 4 | 3 | 3 | 3 |
| 2 | 2 | 1 | 2 | 0 |
| 3 | 3 | 2 | 3 | 0 |
| 4 | 7 | 5 | 5 | 2 |
| 5 | 2 | 7 | 1 | 0 |
| 6 | 3 | 2 | 3 | 2 |
| 7 | 2 | 3 | 1 | 1 |
| 8 | 2 | 0 | 2 | 0 |
| 9 | 5 | 2 | 5 | 3 |
| 10 | 6 | 4 | 9 | 4 |
| 11 | 9 | 7 | 12 | 6 |
| 12 | 1 | 9 | 6 | 6 |
Illustrative example, not real team data. Every chart and percentile here is computed from the same 46 start and finish dates.
Watch the last column. Through week 7 no more than 3 items sat in review on any Friday. In weeks 11 and 12 it was 6.
Cumulative flow diagram: how to read the bands
A cumulative flow diagram (CFD) counts, for each day, how many items have ever reached each column. The counts only grow, so no line goes down, and the colored bands are the gaps between lines.
Items that have reached each state by each day. The top edge is the number started, the top of the blue band is the number finished.
Illustrative example, not real team data. Weeks run from Monday, August 3 to Friday, October 23, 2026.
- Band height is WIP. The vertical gap between two lines is the number of items in that state on that day.
- Slope is rate. The top line’s slope is how fast work arrives. The finished line’s slope is throughput. If the top climbs faster, WIP grows.
- Horizontal distance is time. Measure sideways from a height on the top line to the finished line for roughly how long that item took. It is exact only when items finish in the order they started.
- Flat means stuck, a step means batching. A flat finished line says nothing was delivered. A jump says many items moved on one day, often a bulk edit in the tool.
Here the lines run nearly parallel until about week 8, the picture of a stable system. Then the yellow review band thickens. Items spent 1.9 days in review on average for the first 20 finished, and 5.0 days for the last 20. Work entered review faster than it left. That is a queue, not slow coding.
The CFD and Little’s Law
Little’s Law says average WIP equals average throughput times average cycle time, and a CFD shows all three: band height, finished-line slope and sideways distance. See our post on Little’s Law for the conditions. It works on this data too.
| Quantity | Value | How it was computed |
|---|---|---|
| Average WIP | 3.6 | 295 item-days in progress divided by 82 days |
| Throughput x cycle time | 3.0 | 0.488 items a day x 6.12 days, finished items only |
| Difference | 0.6 | The 6 open items: 50 item-days divided by 82 |
The sides match once unfinished items are counted. That is why aging matters: it is WIP the cycle time average cannot see yet.
Cycle time scatterplot: reading the percentile lines
The scatterplot is the kanban report for promises. It puts one dot per finished item on the chart: the horizontal position is the finish day and the vertical position is the days it took. The dots show the spread that an average hides.
One dot per finished item. Blue is within the 85th percentile, amber is above it, red is above the 95th.
Illustrative example, not real team data. Percentiles use linear interpolation (Excel PERCENTILE.INC).
The 50th percentile is 5 days: half the items finished in 5 or fewer. The 85th is 10 days, and 34 of 40 items (85%) met it. The 95th is 13. The slowest item took 14 days, 2.3 times the average.
The Guide’s service level expectation has the same shape, a period of elapsed time plus a probability. For this team: “85% of items finish in 10 days or less.”
- Lines first: a narrow gap between the 50th and 85th means predictable flow. A wide gap means a few slow items drive your promises.
- Clusters next: a second, higher cluster is usually a different kind of work.
- Outliers: 1 dot sits above the 95th line, at 14 days. Open that item and ask what happened: a block, a hand-off, or a card that was really three.
Wait for 15 to 20 finished items before trusting a percentile. Our post on throughput and cycle time has the same arithmetic in a spreadsheet.
Control chart: what it adds over the scatterplot
This kanban report uses the same dots as the scatterplot, in finish order, and adds a centre line and an upper limit from a stable stretch of history. The scatterplot’s lines answer “what can we promise?”. The control chart’s lines answer “is the process still behaving as before?”.
Mean and upper limit (mean plus three standard deviations) come from the first 20 finished items. Red dots are above the limit.
Illustrative example, not real team data.
The first 20 items average 4.3 days with a standard deviation of 1.7, so the limit is 9.4 days. After the baseline, 6 of the next 20 items cross it, and those 20 average 8.0 days. A run of points beyond the limit signals a change. One lone point, like the 10-day item in week 5, is worth a look, not a conclusion.
Aging work in progress: the daily report
The three kanban reports above look backward. The Guide defines work item age as the elapsed time between when an item started and the current date, so it looks at what is open today.
| Open item | Age on Friday, week 12 | Compared with finished items | Action |
|---|---|---|---|
| Open item 1 | 10 days | At the 85th percentile or older | Find the block and finish it before starting anything new. |
| Open item 2 | 9 days | Past the median | Check it at the next daily meeting. |
| Open item 3 | 7 days | Past the median | Check it at the next daily meeting. |
| Open item 4 | 7 days | Past the median | Check it at the next daily meeting. |
| Open item 5 | 7 days | Past the median | Check it at the next daily meeting. |
| Open item 6 | 4 days | Within the median | Leave it alone. |
Illustrative example, not real team data. The 50th, 85th and 95th percentiles are 5, 10 and 13 days.
A finished item that took 14 days is history. An open item on day 10 can still be rescued.
Symptom to action: what to do when kanban reports look wrong
| What you see | Report | Likely meaning | First action |
|---|---|---|---|
| One band keeps widening | CFD | Work enters that state faster than it leaves | Stop starting work, help the downstream state, add a WIP limit to that column |
| Finished line flat while the top line climbs | CFD | Nothing is delivered while WIP grows | List blocked items and unblock the oldest first |
| Dots above the 95th line | Scatterplot | Blocks, hand-offs or oversize items | Review each as a case; add a rule to split or escalate |
| Wide gap between 50th and 85th lines | Scatterplot | Unpredictable flow, often mixed work types | Plot each work item type separately |
| Point above the limit, or seven in a row above the mean | Control chart | The process changed | Ask what changed that week: staffing, tooling, a dependency |
| Open item older than the 85th percentile | Aging | A promise is at risk now | Swarm before pulling new work |
Which kanban report for which question
| Your question | Open this | Good for | Weak at |
|---|---|---|---|
| Where is work piling up? | Cumulative flow diagram | WIP, bottlenecks, arrival versus departure | Exact times for single items |
| What should we promise? | Scatterplot with percentiles | Service level expectations | Showing where the delay happens |
| Is this month different? | Control chart | Telling change from noise | Forecasting |
| What needs help today? | Aging work in progress | Daily action | History; it forgets finished items |
Use the kanban reports together: the CFD shows review is the queue, the scatterplot shows its cost, the control chart confirms the change is real, and the aging view lists what to rescue. Our older guide to Kanban reporting covers why teams report, and the cycle time guide covers the metric these charts share.
Mistakes that make kanban reports mislead
- Averages instead of percentiles: the average here is 6.1 days, yet 6 items took more than 10. Promise with the 85th percentile.
- Mixing work item types: bugs, features and requests take different time. Filter by type before you read any line.
- Ignoring blocked items: a blocked item still ages and still counts as WIP. Flag blocks and review them daily.
- Backdating or bulk-closing: dates entered after the fact make steps on the CFD and false cycle times.
- Making a chart a target: people judged on cycle time close cards early. Use the charts to ask questions.
Build these kanban reports weekly in 30 minutes
One row per item: start date, finish date, work item type and the date it entered each column.
Add the week’s finished items and recompute the 50th, 85th and 95th percentiles. Compare the 85th with your service level expectation (10 days in the example).
Look for a widening band, a flat finished line and steps.
Look for a point above the limit or seven in a row on one side of the mean.
Sort open items by age and mark every one past the 85th percentile.
Pick one change, such as a WIP limit on the widening column.
To build the board that feeds these reports, start from the Kanban board template and see Kanban board examples. How measurement fits the wider method is in Kanban principles vs practices.
Produce kanban reports from your own data
AgileFever offers a free Flow Metrics Analyzer. Upload a CSV export from Jira, Azure DevOps or any board, with a start and finish date per item. It shows cycle-time percentiles, throughput, work in progress over time, a scatter plot with 50th, 85th and 95th percentile lines, a Little’s Law check and a forecast. The page says the file is read in your browser and not uploaded, and it asks for your name, work email and role first. It does not draw a CFD or a control chart, so use your board tool or a spreadsheet for those.
To learn to improve a service with these reports, AgileFever delivers the Kanban Systems Improvement (KSI) course from Kanban University. Its product page calls it a sequel to the KMP1 course covering pull systems, queuing theory and flow; follow Kanban University for the exact curriculum. As listed on November 12, 2026, KSI is $1,200 or Rs 27,999, and buyers in India pay 18% GST at checkout on top of the rupee price. Listed batches are November 16 to 18, November 21 to 22, December 14 to 16 and December 19 to 20, 2026, some across three weekdays and some across a weekend. See the Kanban Systems Improvement page for current details.
Additional discounts are available. Contact us to know more.
Questions people ask
What are the main kanban reports?
The four most used are the cumulative flow diagram, the cycle time scatterplot, the control chart and the aging work in progress view. The Kanban Guide defines the metrics behind them (WIP, cycle time, work item age and throughput) but does not name the charts, so tools label them differently.
How do you read a cumulative flow diagram?
Each band is a state and its height is the number of items in it. Parallel bands mean stable flow. A band that keeps widening is a queue. A flat finished line means nothing is being delivered. Horizontal distance is roughly how long items took.
What do the 50th, 85th and 95th percentile lines mean?
They are cycle times that a share of finished items met. If the 85th percentile is 10 days, 85% of finished items took 10 days or less. Use the 50th as the typical case, and the 85th or 95th for promises, depending on the cost of being late.
What is the difference between a control chart and a scatterplot?
Both plot finished items by date and cycle time. A scatterplot adds percentile lines to forecast how long future items will take. A control chart adds a mean and limits from a stable baseline to show whether the process has changed. The Kanban Guide names neither.
How much data do kanban reports need?
Aim for 15 to 20 finished items before you trust percentiles, and about 20 stable items for control limits. With fewer, one item moves the lines. Collect dates for a few more weeks before you set a service level expectation.
