In a previous post I’ve introduced the four assumptions behind Little’s Law and discussed the first two assumptions in detail. If you haven’t read those previous posts I encourage you to go back to understand the background. As a reminder, the four assumptions are listed below.
Four Assumptions of Little’s Law
- The average arrival rate is equal to the average departure rate.
- All work entering the system will eventually depart the system.
- The average age of work remains constant, neither increasing nor decreasing.
- Consistent units are used to measure WIP, Cycle Time, and Throughput.
Today we will discuss the assumption that the average age of work remains constant, neither increasing nor decreasing.
The age of a work item describes how long it’s been since a work item that is still in process was started. Any variances in average age of work describe a change in the amount of Flow Debt in the system. As discussed in the first assumption when this Flow Debt is either accumulating or dissipating there is a significant impact on the predictability of the system.
I’ve often described Work Item Age as a leading indicator of Cycle Time. Any work that is currently in progress will eventually finish. Every time a currently in progress piece of work finishes the work item age for that piece of work suddenly becomes its Cycle Time. If Cycle Time is used as an input to our forecasts (which it almost certainly should be) then a change in our average work item age suggests a coming change to our cycle time samples.
When the average age of work in progress is increasing, we are accumulating Flow Debt and there is an upcoming increase in cycle time samples. This will result in any forecasts that we’re making with our existing cycle time samples to be artificially low causing our forecasts to be shorter than what we can expect in reality. This means that any forecasts we make while in this scenario are going to be less reliable and we are more likely to miss our forecasted dates!
Similarly, when the average age of work is decreasing we can expect an upcoming decrease in cycle time samples. While this is certainly better than the previous case in that our cycle time samples will be artificially high resulting in forecasts that may be longer than reality, it is still describing an unpredictable system. This is once again a situation where the lack of predictability may be a good thing in the pursuit of a more effective system.
Click Here for the last post in this series with a short discussion around using consistent units as well as some of the practical steps you can take to help your system more closely match these assumptions.