What does an expedite swimlane on a Kanban board cost a team that uses it twice a week?

## Classes of service and the lanes that show them A **class of service** is a group of work items that the team has agreed to treat differently - not by importance in the abstract, but by a written difference in how items are selected and handled. The classes most teams recognise are: - **Standard** - the default. Pulled in order, no special handling. - **Expedite** - genuinely urgent; jumps the order and interrupts work already under way. - **Fixed date** - close to worthless after a date, so it must be started late enough not to waste capacity and early enough to land. - **Intangible** - no immediate customer visible in it, but real future cost avoided; the class that gets starved unless somebody names it. A **swimlane** is a horizontal band across a Kanban board that makes one of these classes visible. The lane is not the point; the policy attached to it is. A lane with no rule of its own is a colour, and it costs board space to say nothing. ## What an expedite policy usually says - Only one expedite item is in play at a time. - It is pulled ahead of everything else at every column. - Whoever is interrupted stops, leaves their card where it is, and returns to it afterwards. - A named person can invoke it, and the invocation is looked at afterwards. That rule is deliberately expensive to use. Expedite works precisely **because it is rare**: the whole system yields to one item, which is affordable when it happens now and then. ## What twice a week costs For a four-person team, twice a week is roughly nine invocations a month, against a standard flow of maybe twenty-odd items. The exception is now approaching a third of the work, and the costs compound. | Symptom | What it actually means | | --- | --- | | Half-finished standard cards parked in activity columns | Every interruption leaves work stopped, and stopped work is neither delivered nor forgotten | | Standard items landing unpredictably | Their elapsed time now depends on how many interruptions arrived, which nobody controls | | Promises quietly padded | The team is insuring against interruptions it cannot predict | | Expedite invoked for merely annoying work | The label has stopped steering behaviour, because everyone has watched it used loosely | | Nobody can say where the urgency comes from | Each invocation looks like a one-off, so the pattern is never examined | The deepest cost is the last one. A recurring source of urgency handled as a stream of exceptions is a source that is never named, never staffed and never fixed, because the board keeps presenting it as bad luck. ## What to do about it 1. **Classify the last two dozen invocations by source.** Production defects, a commitment made outside the team, an external dependency slipping, a date somebody forgot. 2. **If one source dominates, it is not an exception - it is a class.** Give it a standing lane with declared capacity and its own written rule, so it is planned for rather than allowed to interrupt. 3. **Use the fixed-date class for real deadlines.** A release timed to an annual industry conference is a fixed-date item from the day the date is agreed; expediting it in the final fortnight is exactly the failure the fixed-date class exists to prevent. 4. **Tighten who may invoke expedite** and make each invocation visible afterwards - not to punish anybody, but to make the pattern countable. 5. **Then look again.** If the rate falls, the lane has gone back to doing its job. ## When a swimlane earns its place - It earns it when the class carries a **rule that genuinely differs**, not just a different colour or label. - It earns it when that difference is one a person has to act on at the moment they pick work up. - It costs board space: lanes cut the columns into narrow rows, and a small team's board with four lanes is mostly empty rows. - **Never a lane per person.** That turns a shared system into private queues and hides imbalance. - **Never a lane per component or per customer** unless each really does carry its own handling rule; otherwise put a marker on the card. A small team can usually carry two or three lanes honestly. Beyond that the lanes stop being read, and an unread distinction is worse than none: the board looks like it is saying something when it is not. ## The signal to listen for The healthiest question about an expedite lane is not 'how do we get through these faster' but 'why did this arrive as an emergency at all'. A lane invoked twice a week is not a lane doing its job well. It is the board reporting, honestly, that the team's real intake has a shape nobody has agreed to yet.