Little's Law: what L = λW tells you about work, queues and stock

• LThe time-weighted average number of items inside the boundaryOrders
• λ (lambda)The average rate at which items enterOrders per day
• WThe average time each item spends inside the boundaryDays

In a stable system, items leave as fast as they arrive over the long run, so throughput can stand in for the arrival rate. Rearrange the formula to find whichever value is missing:

W = L ÷ λ

λ = L ÷ W

Units are a quick error check. Items divided by items per hour gives hours. An inspection area with 80 items on average and a throughput of 20 items an hour holds each item for 4 hours. If throughput is recorded per minute, convert it first.

Operations and kanban teams usually write the law as work in progress = throughput × cycle time. The math is the same. The risk is the word cycle time, which different groups use for different start and end points. The cycle time guide covers the competing definitions.

Replace loose labels with exact events:

Exact labels make the calculation easier to audit and harder to misuse.

The idea underneath is accumulated presence. Every item adds time while it is inside. Add up all those stays and you get the area under a chart of items in the system over time. Divide that area by the length of the period and you get the average number present; divide it by the number of items and you get the average stay.

Picture a small workshop that starts and ends a day empty. Four jobs come in during an eight-hour window, and each stays two hours. Together they add eight job-hours of presence.

The jobs do not need even gaps. They can overlap, or the workshop can sit empty between them. The law does not depend on the arrival pattern, the service times, the number of servers or the order of service. Those details change how well the system performs, but not the relationship.

The boundary decides what counts as inside. The law can cover a whole operation or one step, but the count, the rate and the time must all use the same boundary. For an illustrative online retailer:

These are three separate analyses. Dividing the orders waiting for picking by the full order-to-delivery time mixes two groups and answers no clear question.

The same goes for queues. Count only waiting items and you get waiting time; count waiting and in-service items and you get the full visit. An illustrative service desk with 6 people waiting, 2 being served and 24 customers an hour gives 6 ÷ 24 = 15 minutes of waiting, and 8 ÷ 24 = 20 minutes from arrival to exit. Both are right. Name the boundary so a five-minute difference does not become an argument about whose dashboard is wrong.

Decide too how cancellations, abandoned queues and rework count. Keep fulfilled and canceled exits apart so lost business does not look like productive flow, and note whether rework stays inside the same item's time or starts a new visit.

Run through these checks before relying on the answer:

• One kind of item: Count the same type of work each time.
• One boundary: Use matching entry and exit events for all three numbers.
• One time base: Calendar time or working time, not a mix.
• A normal period: A window that reflects ordinary work, not a holiday or a launch week.
• Balanced flow: Compare arrivals, departures and the change in backlog. A backlog that keeps growing breaks the long-run assumption.
• Exceptions written down: Cancellations, abandonment, rework and missing timestamps.

For each check, note the source field, who owns it and any known gap. A short definition sheet is worth more than another decimal place.

The equation earns its keep when it challenges a plan. These illustrative cases show four common traps:

The law links the outcomes; it does not create the change. Process changes such as WIP limits, smaller releases or removing a bottleneck have to do that.

The same rule in four settings. All figures are illustrative:

In the fulfillment case, measure order acceptance to shipment as well. If the internal figure improves and the customer figure does not, the change moved the waiting rather than removing it.

It is a relationship between averages, not a queueing model. Pair it with other evidence for these questions:

Two teams can both average four days while one finishes nearly everything near four days and the other mixes same-day work with month-long cases. Their customers have very different experiences.

A record another person could repeat:

For a time-weighted average, weight each count by how long it lasted. A system holding 10 items for six hours and 30 items for two hours averages (10 × 6 + 30 × 2) ÷ 8 = 15 items. A snapshot taken only in the busy hours would say 30.

1. 1

   Calculate the missing value. Then compare it with a separately measured figure, such as average cycle time from item history.

2. 2

   Explain any gap. Look for definition mismatches, missing histories, window edges and changing conditions. Do not assume the numbers agree just because the law is exact.

3. 3

   Keep a discrepancy log. For example: the open-ticket count includes reopened cases, but the completion report excludes them. Fix these before presenting results.

4. 4

   Pick one change with a clear mechanism. For example, release less work into a crowded step while watching completed output. Write the expected direction of change for count, flow and time.

5. 5

   Name a guard measure. Something that would show the change backfired, such as a growing queue upstream.

6. 6

   Compare like with like. Note changes in demand or work mix, and keep poor results. They may show that cutting workload also cut output, or that waiting moved elsewhere.

1. 1

   Mixing boundaries. Counting one group of items and timing another.

2. 2

   Mixing units. Throughput per week with time in days.

3. 3

   Using a snapshot as the average. One busy-hour count is not a time-weighted average.

4. 4

   Applying it to a growing backlog. If arrivals keep outrunning exits, the long-run averages do not settle.

5. 5

   Quoting the average as a promise. The law gives the mean, not a date for one customer.

6. 6

   Reading cause into the algebra. The law says what must fit together, not what will change if you act.