What should a team do when a board column is at its work-in-progress limit?

## The one move that is never right Starting a new card. It is worth being blunt about this, because it is the honest instinct and it is always wrong at a blocked column. Starting adds nothing to the cards that are stuck, it adds an item that will eventually queue at the same stage that is already saturated, and it makes every item currently in flight finish later, because there is now more work competing for the same constrained stage. The team ends the week having started more and finished the same amount. The second move that is never right in the moment is raising the number. Whether the cap is correct is a real question, and a good one - but it is a question to answer calmly, on evidence, at a time when nobody is blocked. Answering it while blocked always produces the same answer, and after the third time the number no longer means anything. ## A ladder of good moves 1. **Swarm the blocked column.** Help finish a card that is already there. Review the change, pair on the hard part, take a subtask, write the tests. This is first because it directly attacks the reason the column is full. 2. **Help the stage that is actually constraining the board.** If you cannot work inside the blocked column - no access, no qualification - look at what that column is waiting for and help there instead. 3. **Remove the impediment rather than the card.** A card stuck for a reason outside the team is not waiting for effort, it is waiting for an event. Chase the approval, get the missing input, fix the flaky check, escalate the dependency. Time spent making an event happen sooner is the highest-value work available. 4. **Take work that creates no new card in flight.** Improving tooling, building test data, cleaning up the environment that keeps breaking, writing the thing a colleague is blocked on. This is genuinely useful and it does not add to work in progress. 5. **Stand down deliberately, visibly and briefly.** If none of the above exists, the honest answer is to wait, and to say that you are waiting. ## Why deliberate idling is the senior answer Interviewers ask this question because most candidates cannot say step 5 out loud. Admitting that the right thing to do is nothing feels like admitting to slacking, so people invent work instead - and invented work is the most expensive kind, because it consumes attention that the constraint needed and it usually creates a card. The underlying idea is that a team optimises for **finished work, not for occupied people**. Keeping everybody busy at all times is only optimal if every stage is equally fast, which is never true. Once one stage is slower, someone has to wait; the only choice is whether they wait visibly, or wait invisibly behind a pile of half-done work. Visible waiting is cheaper, and it creates the pressure that gets the constraint fixed. The practical form matters. Idling should be short, obvious and accompanied by a raised blocker - not a quiet afternoon. If it recurs, it is data, and it belongs in the conversation about the constraint. ## Spike or chronic? The senior distinction is between a block that happened today and a block that happens every week. | | One-off spike | Chronic constraint | | --- | --- | --- | | Looks like | one column full for a few hours | the same column full most weeks | | Usual cause | a hard card, an absence, a bad day | one qualified person, an external approval, a skills gap | | Right response | swarm it and move on | change the structure, not the number | | Wrong response | a process change | swarming again and hoping | Swarming a chronic constraint is a treadmill: the team pays for the same rescue every week and never fixes the cause. The fix there is structural - cross-train a second person into the constrained skill, get the approval delegated, split the stage, reduce the demand entering the board - and it is usually a conversation with someone outside the team. ## When raising the number is legitimate It is legitimate when the team decides it deliberately, outside the pressure of the moment, with a reason that will still be true next month: the stage genuinely gained capable people, or the column turned out to cover two different kinds of work that should be separated. Even then, change one number at a time and give it several weeks before judging. ## The pressure case A 7-person team building a legal-document review product has a regulator's audit date 23 working days out, and the legal-check column has been full for 3 days. The instinct is to raise every cap so people can keep moving. That is exactly backwards: raising the caps will not make the single qualified reviewer faster, and it will leave the team on the audit date with 14 items part-finished rather than 8 items delivered. The pressure is an argument for tighter limits and ruthless finishing, plus one structural move - a second qualified reviewer, or a delegated sign-off - started today rather than at the deadline.