Adrian Colyer wrote a great piece on applying the Universal Scalability Law (USL) to management:
He starts by two insights stemming directly from USL:
C(N) = N / ( 1 - α(N-1) - β*N(N-1) )
Where C is the organizations’ capacity (max throughput), N is the number of people, α and β are the contention and coherence coefficients respectively.
- Contention – measures the impact of waiting/queuing for resources on capacity, more commonly known as a the “bottleneck factor”. The contention coefficient reflects your ability to effectively delegate work and not become a bottleneck. Smaller coefficient reflects better delegating ability.
- Coherence – measures the cost of getting agreement of what is the right thing to do. The coherence coefficient reflects the decision making autonomy within the organization. The more people need to be involved in making a decision, the higher the coefficient will be, and the capacity return from adding more people will decrease and may even become negative.
When trying not to become a bottleneck, we have to fight against our natural tendency to try and be helpful to anyone who asks us to help or contribute to a project. Little’s Law shows how by doing so we’re not being helpful at all:
average_wait_time = work_in_progress / throughput
Since our throughput is fixed (there are only that many hours in a day), the only thing we control is our “work in progress”, the number of concurrent projects we take on. The more projects we take on, the longer our average wait time gets and we become more and more of a bottleneck on these projects…
This gets compounded by the fact the request don’t come into our queue at a uniform rate. We know from Kingman’s Formula:
That as our utilization (ρ) increases, wait time increases exponentially. Working too close to 100% utilization will grind things to a halt. Adrian suggests choosing a WIP limit that will result in spending about 60% of your time on operational activities – ones that will slow down the rest of the org if you don’t process them promptly. The rest of those operational activities should either be delegated or discarded. The remaining 40% should be used for more strategic activities that will not have a negative cascading impact on the organization if they are put on hold during short-term periods of high operational load.