A work-in-progress limit is a policy: a maximum number of items allowed in a state, enforced when it is inconvenient. It is not a slogan on a wiki, and it is not 'we try not to start too much.' If you break the limit whenever a sponsor is in the room, you have a poster.
Why a limit makes work finish faster
Uncapped WIP is how a team starts twelve things and finishes none. Each item waits on review, on a specialist, on a decision. People context-switch and call it collaboration. Calendar time per item grows even as utilisation looks heroic. The motorway is full; nothing is arriving.
A limit forces a choice: finish or help finish, rather than open a new card. That is uncomfortable in organisations that reward visible starting. It is also how review queues shrink and blockers become the main event instead of a footnote. You cannot hide a bottleneck when the column is full and the policy says stop.
Little's Law, in the form you can say aloud: if you keep more items in play, and you do not magically raise throughput, each item spends longer in the system. Cut WIP and, if throughput holds, cycle time falls. That is not a slogan. It is a relationship you will measure in the next lesson. The limit is how you choose the WIP term on purpose.
Set a limit you can defend, then defend it
Start from the real column, not a blog post. Count how many items already sit in 'In development' in a typical week. Set the limit at or slightly below that, or at roughly the number of people who can actually work that state — then watch. A limit of 1 on a team of eight is a stunt. A limit of 40 is a compliment to chaos.
Write it on the column. When someone wants to pull past it, the answer is not 'just this once' unless you have a class-of-service rule for that once. The adult move is: swarm the stuck item, split the card, or escalate the blocker. Starting a fifteenth item is how you vote against the policy while nodding in the retrospective.
Limits are experiments. If a column is always empty, the limit may be too tight or the upstream policy is starving the system. If a column is always slammed against the cap and items age, you found a constraint — capacity, skill, or wait on another team. Change one thing. Recalibrate with a week of data, not with a feeling in a crisis.
The politics: who is allowed to break the rule
Sponsors will ask you to 'just start' the urgent extra. Translate: they are asking to raise WIP and therefore cycle time on everything else, including their first urgent extra. Say that in English. Offer a swap — what stops so this can start — or an explicit temporary limit raise with an end date. Silent yes is how policies die.
Personal WIP matters as much as column WIP. A person with six 'in progress' items is a private push system. Pairing, swarming, and 'only one active card' are policies too. If only the board is limited and every human is a hoarder, you have decorated the constraint.
Do not use a limit to punish. Use it to make the constraint discussable. A team that hits the cap and holds a five-minute huddle is using the system. A PM who uses the cap to scold is using a ruler. You will get compliance and hidden work in drawers.
Summary & next steps
A WIP limit is an enforced maximum, set from the real system, adjusted with data. It exists so you finish and expose bottlenecks. Breaking it without a class-of-service rule is how you return to a full motorway.
Next: flow metrics — cycle time, lead time, throughput, the cumulative flow diagram, and Little's Law as something you can see, not recite.