Systems & Hidden Structure
Why Lines Grow When Timing Varies
Also called: Queueing theory
- Established idea
- Formal theory
- Observed in experiments
- Working interpretation
Uneven timing can cause long waits even when there seems to be enough capacity on average. Queueing theory, the math of waiting lines, shows how randomness, busyness and crowding feed each other, so waits can grow much faster than the workload does.
How much spare capacity is worth keeping to handle sudden rushes of demand?
The idea
Imagine a coffee stand where one barista can serve 60 customers an hour, and 50 arrive each hour. On paper the barista has spare time. In practice, customers come in clumps and some orders take longer, so a line still forms. Queueing theory explains how long it gets. The key result is that waiting rises slowly at first and then sharply as the server gets close to fully busy. Two other facts are useful. Little's Law says the average number of people in line equals the arrival rate times the average wait. And cutting variation, in arrivals or in service times, shortens waits without adding any capacity.
Why it attracts me
It is one of the clearest cases where intuition fails and simple math corrects it. It also puts a price on running everything at full speed.
An example
The Queueing & Staffing Lab on this site lets you set the demand, the length of each job and the number of inspectors on duty, then shows what share of requests are answered within a target wait. The Queueing Simulation Lab plays the same kind of line out customer by customer (Simulating a System Event by Event).
Where it connects
Waiting lines are one place where uncertainty turns into concrete cost and risk (Probability, Risk & Uncertainty). They also explain why a bottleneck (Finding the Real Bottleneck) needs a small buffer of work in front of it, so it never sits idle waiting.
What this does not establish
The waiting-line results hold under their stated assumptions about arrivals and service times. Applying them to a person's week is a loose comparison, not a measured result.
Questions I'm still exploring
- How much slack is wise, and how much is simply waste?
- Which kinds of variation can be removed, and which must be absorbed?
- Does the lesson about schedules with no slack hold for a person's week, or is that too loose a comparison?
Sources and further reading
- John D. C. Little, "A Proof for the Queuing Formula: L = λW," Operations Research 9(3), 1961
- Wallace J. Hopp and Mark L. Spearman, Factory Physics (McGraw-Hill)
Working interpretation: drafted from my notes and interests for review. It is not a direct quotation, and I may still change it.