Queueing-system conservation identity
Little's Law
In a stable system over a consistent boundary and time horizon, average work in process equals average throughput multiplied by average flow time.
L = lambda W
L is the time-average number of items in the system, lambda is the long-run effective arrival or departure rate, and W is average time an item spends inside the same system boundary.
The live queue uses an adjustable service time and arrival rate. The identity applies to long-run averages; the animation shows transient congestion and variability around them.
(items)
Arrival spacing and journey duration drive the animation. The equation constrains long-run averages, while the visible queue fluctuates.
- CHANGE
- Arrival rate
- WATCH
- work in process
- MEANING
- The live queue uses an adjustable service time and arrival rate. The identity applies to long-run averages; the animation shows transient congestion and variability around them.
Count the items or time their journeys: both views measure the same area.
A cumulative-arrival and departure diagram turns inventory into vertical distance and flow time into horizontal distance.
What it actually says
Little's Law is an accounting identity rather than a particular queue model. It requires no Poisson arrivals, exponential service times, first-in-first-out discipline, or single server.
The hard part is measurement discipline: L, lambda, and W must refer to the same customers, boundary, and observation regime. Throughput, not offered demand, is used when arrivals are rejected or abandoned.
"A useful law compresses a pattern. It does not erase the conditions that make the pattern true."
How the idea developed
The modern form emerged through observation, argument, and later refinement. The timeline separates the first insight from the version now used in textbooks and practice.[1]
Philip Morse states related queueing relationships.
John Little publishes a proof of L = lambda W under stationary assumptions.
Jewell gives a sample-path style proof that broadens understanding of the identity.
Operations, software, manufacturing, healthcare, and networking use the relation for flow diagnostics.
How the pattern works
The relation becomes useful only when its mechanism, measurement process, and operating range are visible.
Every item contributes one unit of inventory for every unit of time it remains inside.
Total item-time equals the sum of individual flow times.
Arrivals, inventory, departures, and time must be measured over the same system.
Long-run inflow and outflow must balance without unbounded accumulation.
L is the time-average number of items in the system, lambda is the long-run effective arrival or departure rate, and W is average time an item spends inside the same system boundary.
Where it earns its keep
Applications are strongest when the law changes a decision, measurement, model, or experiment rather than merely providing an analogy.
Estimate cycle time from WIP
ApplicationTeams can divide average inventory by actual throughput.
Separate active work, queues, blocked work, and rework with explicit boundaries.
Control work in progress
ApplicationDelivery systems can use WIP limits to reduce average flow time at a given throughput.
Little's Law does not promise that cutting WIP preserves throughput.
Relate census, discharge, and stay
ApplicationAverage occupied beds equal discharge rate times average length of stay under a stable boundary.
Case mix, boarding, cancellations, and seasonal nonstationarity need stratification.
Where it stops working
The identity may fail as an estimate when the observation window is short, the system is rapidly changing, inventory grows without bound, or censored items are omitted.
It gives averages only. Two systems with the same L, lambda, and W can have very different variability, percentiles, fairness, and service levels.
"Arrival demand always equals lambda"
Better: Use effective throughput when work is rejected or abandons."It predicts individual waiting time"
Better: It constrains an average, not a distribution or sequence."Reducing WIP automatically raises throughput"
Better: Capacity, batching, blocking, and starvation can reduce output."Any three dashboard numbers can be combined"
Better: They must share the same units, population, boundary, and time basis.Sources and further reading
Original publications and serious secondary scholarship are prioritized over summaries.
- Little - A Proof for the Queuing Formula L = lambda WThe original 1961 proof in Operations Research.https://doi.org/10.1287/opre.9.3.383
- Little and Graves - Little's LawModern exposition, history, and applications.https://doi.org/10.1007/978-0-387-73699-0_5
- MIT OpenCourseWare - Queueing TheoryUniversity materials on flow systems and queueing models.https://ocw.mit.edu/courses/15-072j-queues-theory-and-applications-spring-2006/
- Factory Physics - Little's LawOperations-focused explanation of throughput, WIP, and cycle time.https://factoryphysics.com/principle/littles-law/