Cycle time, WIP, and why limiting work speeds it up
Four numbers describe how work moves. The relationships between them are more useful than any one of them alone.
Lead time and cycle time are different promises
Lead time runs from the moment you commit to something until it is done. It includes all the waiting before anyone started, and it is what the person who asked experiences.
Cycle time runs from starting work until done. It is what the team controls.
The gap between them is queue time, and on most boards it is larger than the work itself. A team with a three-day cycle time and a three-week lead time does not have a speed problem; it has a queueing problem, and working faster will not touch it.
Quote the 85th percentile
Cycle time distributions have long right tails, which means the average sits well below what people actually experience. An average of five days with a tail out to thirty is not a useful planning number for anyone.
"We finish 85% of items within nine days" is a commitment you can keep and a stakeholder can plan around. It is also falsifiable, which is what makes it worth saying.
Little's Law and the case for WIP limits
WIP = throughput × cycle time. Rearranged: at a fixed throughput, cycle time is
proportional to work in progress.
Twenty items in flight at five per week gives a four-week cycle time. Ten items in flight gives two weeks. Nobody worked faster — the queue got shorter. This is arithmetic, not philosophy, and it is the whole argument for limiting work in progress.
The instinct when someone asks for something is to start it. The law says starting it makes everything already underway finish later, including whatever they asked for last month.
The steady-state assumption
Little's Law is exact for a system in steady state: arrivals matching departures, nothing accumulating, a consistent definition of "in progress". It still returns a number when those fail.
If your backlog is growing, if items are being abandoned rather than finished, or if "in progress" quietly changed meaning halfway through the quarter, the arithmetic is correct and the answer describes nothing.
Aging is the only actionable metric
Cycle time, throughput and lead time all describe the past. Aging work in progress describes items on the board right now, and it is the only one you can still act on.
An item aged eleven days against an 85th percentile of nine is already an outlier and is telling you so while there is time to intervene. Most teams discover this in a retrospective instead, which is the same information arriving too late to use.