Key takeaways
- A WIP limit is not an effort quota: it is a visual signal that tells you when to stop starting and start finishing.
- The simplest starting formula: limit = number of people touching the column + 1 buffer, then adjust by measurement.
- The test runs over two weeks, one column at a time, comparing lead time and queue sizes before and after.
- The cumulative flow diagram makes the real effect visible: if the "In Progress" band thickens, the limit is too high or the column is poorly split.
Why a WIP limit changes more than a board
On a Kanban board, everyone looks at the columns. Few people look at the number of cards piling up in them. Yet that is where lead time is decided.
A WIP (work in progress) limit is a cap: "no more than N cards in this column at any given time." When the cap is reached, you do not pull the next card. You finish what has been started.
Pawel Brodzinski reminds us in Milk Kanban that the original Kanban is first and foremost a visual signal: a card placed on the last brick of milk says "we need to reorder," with no board or column. The WIP limit is exactly that signal, applied to a column. It does not measure effort, it triggers a decision.
Without a limit, an "In Progress" column becomes a storage area. Cards enter quickly, leave slowly, and no one sees the problem until the customer is waiting.
How to calculate a first limit, without getting it wrong
There is no magic formula, but a reasonable starting point keeps you from going in blind.
The basic rule
- Count the number of people actually working in the column (not on the team, in the column).
- Add 1 to absorb the unexpected (a blocked card, a review, a test).
- The result is your starting limit. Example: 3 developers on "In Progress" → limit 4.
This rule is deliberately rough. It serves to set a number, not to defend it. Calibration comes afterward.
Split before capping
A catch-all column ("In Progress" containing dev, review, and test) makes any limit useless: the cap applies to stages that do not have the same duration. First split into columns that represent real stages of the flow. Then cap each column separately.
On a tool like Ever Earlier, this translates into custom columns with drag-and-drop and an activity log, which makes it possible to see how many cards moved and when. The log serves as raw material for measurement.
The per-column calibration exercise, over two weeks
Goal: find the limit that reduces lead time without starving the team. One column at a time, two weeks, three measurements.
Week 0: measure the current state
- Record, every day at the same time, the number of cards in the target column.
- Record the average time between entry and exit of that column for completed cards.
- Record the number of blocked cards (with a reason: waiting for info, dependency, bug).
Week 1: apply the limit
Set the calculated limit. When it is reached, the rule is simple: you do not pull a new card. You help finish, you unblock, you review.
Expect resistance. In the first hours, someone will say "but I have nothing to do." That is precisely the signal that the upstream column is poorly fed, or that the limit is too low.
Week 2: adjust by one notch
- If the column stays full and people are waiting: raise the limit by 1.
- If the column often empties and lead time increases: lower the limit by 1.
- If lead time drops and the upstream queue does not grow: keep the limit.
Repeat the exercise on the next column. Two weeks per column, no more. Beyond that, the team loses the thread.
Measuring the effect: three metrics are enough
No need for a dashboard with twenty indicators. Three measurements tell the essential story.
Lead time
Time between a card entering the flow and leaving it. This is the metric the customer feels. If the WIP limit works, this number drops or stabilizes.
Throughput
Number of cards completed per week. Careful: a WIP limit can lower throughput in the short term, then raise it again. Do not judge over three days.
The cumulative flow diagram
A chart where each band represents a column, stacked over time. If the "In Progress" band thickens, you are accumulating unfinished work. If it stays thin and the "Done" band rises steadily, the flow is healthy.
Ever Earlier offers burndown, velocity, and cumulative flow reports, as well as a Gantt chart. These views do not replace the decision, they make it visible in meetings.
Three mistakes that cancel out a WIP limit's effect
A poorly used WIP limit becomes an arbitrary constraint. Here are the most common pitfalls.
Applying the limit to the whole team at once
If you cap five columns on the same Monday, no one knows which limit causes which effect. One column at a time.
Confusing limit with productivity target
A WIP limit is not a quota. It says "stop starting," not "produce more." If you use it to push, the team will work around it by artificially splitting cards.
Forgetting to split large cards
A card that takes three weeks blocks an entire column. User stories with acceptance criteria and points help spot the ones that are too large before they enter the flow. Ever Earlier's AI agent, which generates user stories from epics, serves exactly this purpose: split before capping.
Connecting the limit to existing rituals
A WIP limit does not survive if it is not discussed. It must appear in the rituals the team already holds.
- Daily: start with the capped column. Who is blocked? What needs to be finished today?
- Flow review: once a week, look at the cumulative flow diagram. Is the "In Progress" band thickening?
- Retrospective: adjust one limit, not five. Note the change in the activity log.
If your team uses sprints and a backlog, the WIP limit also applies during the sprint: it prevents pulling a new user story as long as the previous one is not finished. This is often where it produces the most effect, because the temptation to parallelize is strong.
Read next
Ineffective Daily Standup: The 3-Question Test
Your daily standup drags on, goes in circles, and no one dares say it. In one week, three questions are enough to decide: coordination or theater. Here is the diagnostic protocol, the thresholds to measure, and the asynchronous format for distributed teams.
Kanban or Scrum: Choosing the Right Method in 2026
Continuous flow or iterations? The choice depends mainly on your delivery frequency and the nature of your product. Here are concrete criteria to decide, and how to configure each approach in a tool like Ever Earlier.
Gantt Chart in Agile: Useful or Counterproductive?
The Gantt chart is not the enemy of agility: misused, it freezes; used well, it aligns. Here are the cases where it helps, those where it hurts, and how to combine it with sprints and Kanban.