Lean Six Sigma · Interactive lesson
Little’s Law
Learn the simple relationship that connects work in process, flow rate, and time—then watch it come alive in a production flow, a Lean value stream, and a queue.
Why this matters at work: Little’s Law turns three measurements you can observe into one practical constraint. It helps teams estimate lead time, set sensible WIP limits, understand congestion, and ask better questions about why work is taking so long.
Foundation
What is Little’s Law?
Little’s Law describes average flow through any system with a clear entrance and exit. It says that the average number of items inside equals the average rate at which items flow through multiplied by the average time each item spends inside.
The “items” can be parts, customers, patients, orders, calls, support tickets, or software tasks. The law does not predict the journey of one particular item. It connects three averages measured across many items and over time.
Average items in the system = average arrival rate × average time each item spends in the system.
More items arriving, or items staying longer, means more items accumulate inside the same system boundary.
This is an accounting relationship—a constraint on the averages—not a claim that one variable alone caused another to change.
Customers, parts, orders, tickets, jobs, patients, calls, etc.
Items entering per unit time.
Waiting time + processing/service time inside the chosen boundary.
Start by drawing the system boundary. An item begins contributing to L when it crosses the entrance and stops when it crosses the exit. Its W is every moment between those two events—including waiting, transport, inspection, rework, or service that happens inside the boundary. L, λ, and W must describe the same items, boundary, and observation period.
Keep the units compatible. If λ is measured in parts per hour, W must be expressed in hours. For example, 20 customers/hour × 0.25 hour/customer = 5 customers in the system on average.
Why the equation makes sense
Imagine adding up every minute contributed by every item while it is inside the boundary. You can view that total two ways: as the average number present over time, or as the number of items flowing through multiplied by their average residence time. Both views count the same accumulated item-time, which is why L = λW.
Interactive 1
See Little’s Law in motion
Each moving P is one part. It begins counting toward “Active now” when it enters Arrival and stops when it leaves Inspect. The complete trip through arrival, queue, process, and inspection is W; the rate at which new parts enter is λ; and “Predicted average WIP” is L.
Try this guided experiment
- Start at 12 parts/hour and 15 minutes. The equation predicts 3 parts in the system on average.
- Double λ while holding W fixed. Predicted average WIP doubles because twice as much work enters per hour.
- Restore λ, then double W. Predicted average WIP doubles again because each part occupies the system twice as long.
- Compare “Predicted average WIP” with “Active now.” One is a long-run average; the other is a changing snapshot.
How quickly new work enters the system.
From crossing the entry boundary until leaving the exit boundary.
12 parts/hour × 15 minutes ÷ 60 = 3.00 parts in the system on average.
The moving picture will fluctuate at any instant. Little’s Law relates long-run averages, not a single snapshot.
This time-compressed, discrete animation is an illustration rather than a full stochastic simulation. After changing a control, allow the parts already in motion to finish their trips before interpreting the new pattern.
More parts enter during the same time span. If W stays fixed, more parts must be present on average.
Each part occupies the system longer. If λ stays fixed, more parts overlap inside the boundary.
The law narrows the explanation: λ increased, W increased, or both changed.
Interactive 2
Little’s Law in Lean: connect WIP to flow time
Lean uses different operational words for the same three quantities: L becomes average work in process (WIP), λ becomes throughput, and W becomes average flow time. In a sufficiently stable value stream, average arrivals and average completions balance, so throughput can stand in for the arrival rate.
When the process is observed in a sufficiently stable operating regime, arrival rate and departure throughput balance, so the equation is commonly written using throughput.
How to read the visual
The orange numbered tiles picture the average work occupying the chosen value stream. Set throughput and WIP, then read the calculated flow time. Press Halve WIP or Double WIP to see the mathematical effect while throughput is held fixed.
The average number of finished parts leaving per hour—not the theoretical maximum capacity.
All started-but-not-finished work inside the value-stream boundary.
At the same throughput, less average WIP mathematically means less average time in the system. This is why Kanban WIP limits can be powerful.
Reducing WIP does not guarantee throughput stays fixed. An excessively low WIP limit can starve downstream work. Little’s Law constrains the averages; it does not by itself tell you which operational change caused them.
Capacity, batching, rework, downtime, priorities, and variation can all change the operating point. Use process knowledge to explain the cause.
A value stream averages 24 units of WIP and completes 6 units/hour.
If the team can sustainably reduce WIP to 12 while maintaining 6 units/hour, the average flow time becomes 2 hours.
Interactive 3
Little’s Law in queueing theory: connect waiting to congestion
Queueing theory separates the time spent waiting from the time spent being served. For the whole boundary, Little’s Law is L = λW. For the waiting line alone, the matching form is Lq = λWq. The boundary determines which number-and-time pair belongs together.
For the system boundary shown below, L includes customers waiting plus customers in service, and W includes waiting time plus service time.
How to read the queue visual
- λ is the arrival rate; μ is the server’s average service capacity.
- The purple dots approximate average queue length Lq. The blue server represents service capacity.
- ρ = λ/μ is utilization—the average fraction of time the server is busy in this model.
- L = Lq + ρ and W = Wq + 1/μ for this one-server model.
Begin at λ = 10 and μ = 14. Then raise λ toward μ. Notice that the waiting measures grow nonlinearly even though the arrival rate moves in equal steps.
μ
This demonstration assumes one server, Poisson arrivals, exponential service times and λ < μ. Under those assumptions, W = 1/(μ − λ). Little’s Law then links the resulting averages through L = λW.
As λ approaches μ, the M/M/1 model predicts a rapidly increasing queue and waiting time. At λ ≥ μ there is no finite steady-state average queue: arrivals are entering at least as fast as the server can sustainably remove them.
Go broader
One law, many kinds of flow
The vocabulary changes by industry, but the structure does not: define a boundary, identify what flows, measure how many are inside on average, measure the rate, and measure complete time inside.
L: parts in WIP
λ: completions per hour
W: start-to-finish flow time
L: callers waiting or served
λ: calls admitted per hour
W: wait plus talk time
L: patients present
λ: patient arrivals per hour
W: total department time
L: started-not-done work items
λ: completions per week
W: work-item lead time
L: open orders
λ: shipments per day
W: order-to-ship time
L: open tickets
λ: resolved tickets per day
W: open-to-resolution time
The most useful practical rearrangement is W = L ÷ λ. If a team has 40 open items and finishes 10 per week, the implied average time in the defined system is 4 weeks. That result is a diagnostic starting point: the law tells you what the averages require, while investigation tells you whether queues, handoffs, rework, downtime, or priorities created that operating point.
Apply it correctly
- Use one consistent system boundary and one consistent item population.
- Use the effective or admitted arrival rate when requests can be rejected, abandoned, or cancelled.
- Match time units and observe long enough for meaningful averages.
- Do not substitute today’s snapshot for average WIP.
- Define “flow time,” “lead time,” and “cycle time” locally; teams do not always use those words the same way.
Guided practice
Use the relationship before the final quiz
These three questions give immediate feedback. Translate the story into L, λ, and W, make the time units consistent, and keep the system boundary in view.
Choose an answer, then reveal the reasoning.
- Convert time units before multiplying. Five minutes = 5/60 hour.
- Use averages measured over a consistent system boundary and observation period.
- Do not confuse instantaneous WIP with average WIP.
Go deeper
History, limits, and references
Little’s Law began in queueing and operations research, but later work showed why its core accounting logic reaches beyond any one probability model or industry.
John D. C. Little published the first general queueing-theory proof in 1961. Shaler Stidham later gave a deterministic sample-path treatment, showing that the underlying relationship is broader than a particular probability distribution or queue discipline.
For a finite observation window T, the linked article defines Λ = N/T, where N counts every distinct item present during any part of the window—including items already inside when the window starts. It defines w as the average portion of each item’s residence time that falls inside that window. Those observer-relative quantities differ from the long-run λ and W, yet their product equals the exact time-average L for the window. This is useful when a process is changing or the reporting window is short relative to the time work spends in the system.
- Krishna Kumar. A Deep Dive into Little’s Law, The Presence Calculus Project. Requested lesson reference.
- J. D. C. Little. “A Proof for the Queuing Formula: L = λW.” Operations Research, Vol. 9, No. 3, 1961, pp. 383–387.
- S. Stidham. “A Last Word on L = λW.” Operations Research, 1972.
- W. J. Hopp & M. L. Spearman. Factory Physics: Foundations of Manufacturing Management, 2nd ed., McGraw-Hill, 2000.
Stuff in the system = Stuff per unit time × Time in the system.