Little's Law in manufacturing is the exact relationship WIP = throughput × lead time, which lets you predict a line's lead time from inventory and exit rate alone. Rearranged as lead time = WIP ÷ throughput, it turns a pile of work-in-process into a number you can promise a customer.
The general law is a queueing identity that holds for any stable system. On a shop floor it becomes a planning tool: you can walk a line, count the parts sitting between operations, divide by how many parts leave per hour, and know the average time an order spends in the plant, without a scheduler, a simulation, or a stopwatch on every station. This post applies the law to real production settings: single cells, whole value streams, order backlogs, and the WIP caps that make it a lever instead of a curiosity. For the underlying proof and the queueing form L = λW, see Little's Law for manufacturers.
What does Little's Law mean on a shop floor?
On a shop floor Little's Law means the average number of parts inside a boundary equals the exit rate multiplied by the average time each part spends there. Draw a box around a cell or a plant, and the three numbers you can read off the floor, parts in the box, parts per hour leaving, and hours each part waits, are locked together by WIP = throughput × lead time.
The reason this matters in production is that lead time is the hardest of the three to observe directly. You cannot easily follow one part from release to shipment across weeks of queues. But you can count WIP with a walk and measure throughput from the shipping log. The law hands you the third number for free. A plant that carries 4,000 units of WIP and ships 800 units a day has a 5-day average lead time whether the planner believes it or not.
The terms have concrete plant-floor meanings:
- WIP the average count of parts, jobs, or orders inside the boundary you chose, from the first operation to the last. Raw material not yet released and finished goods already shipped are outside the box.
- Throughput the actual rate work leaves the box, measured at the exit. It already bakes in downtime, scrap, breaks, and speed loss, so it is almost always below a machine's nameplate rate. This is the same exit rate covered in throughput in manufacturing.
- Lead time the average clock time a unit spends inside the box, most of which is queue time, not run time. This is the plant-level view of manufacturing lead time.
How do you predict lead time from inventory alone?
You predict lead time by dividing average WIP by throughput, both measured against the same boundary over the same window. That is the whole trick: inventory you can count on a walk becomes a delivery estimate without touching the schedule.
Say a machining department has 900 parts in process on an average day, on machines, in totes between operations, on the inspection bench, and it ships 150 finished parts a day. Little's Law gives a lead time of 900 ÷ 150 = 6 days. If sales is quoting 3-day lead times, the math says orders are actually sitting in queue for twice that, and the quote is fiction. The fix is not a faster machine; it is either less WIP or more honesty in the quote. When a plant finally measures WIP and exit rate with live floor data instead of guessing, the gap between the quoted lead time and the real one is often the most expensive surprise on the board.
By the numbers. John D. C. Little published the first general proof of L = λW in Operations Research in 1961, and the identity holds for any stable system regardless of arrival pattern or processing order (Little's law). The Lean Enterprise Institute lists lead time, WIP, and throughput among the core measures of a value stream, and Little's Law is the identity that binds them together (Lean Enterprise Institute, value stream mapping). In most job shops, queue time is the dominant share of that lead time, which is exactly why cutting WIP pays off (Lead time).
Why does cutting WIP shorten lead time without new machines?
Because throughput is set by your constraint, not by how much work you release. Rearrange the law to lead time = WIP ÷ throughput. If the exit rate is fixed by the slowest necessary step, lead time rises and falls in lockstep with WIP. Pile twice as much work in front of the constraint and every part waits twice as long; it does nothing to make the constraint faster.
Take a packaging line that finishes 1,200 cases a day and averages 3,600 cases of WIP across its stages. Lead time is 3,600 ÷ 1,200 = 3 days. Cut the released WIP to 1,800 cases with a release rule, start a new case only when one ships, and the line still finishes 1,200 a day, so lead time falls to 1,800 ÷ 1,200 = 1.5 days. Same equipment, same output, half the wait. This is the arithmetic under pull systems, kanban limits, CONWIP loops, and one-piece flow. They all work by capping WIP so the same throughput clears the floor faster.
How do you use Little's Law on a value stream?
You use it as a five-minute check you can run on any line, then as a lever to shorten the quote. The steps below turn the identity into a floor routine.
- Draw the boundary. Pick exactly where the system starts and ends, one cell, a department, or raw release to shipping dock. All three terms must refer to the same box or the numbers will not reconcile.
- Count WIP on a walk. Tally the average parts inside the box: on machines, in queues, on benches, in transit. Use a typical day, not a peak or an empty Monday.
- Read throughput at the exit. Take the actual completion rate over a representative window from the shipping or completion log, not the machine nameplate.
- Solve for lead time. Compute lead time = WIP ÷ throughput and compare it to the lead time you quote. A wide gap means orders queue far longer than anyone admits.
- Cap WIP and re-measure. Set a release rule or kanban ceiling that holds WIP down, then recount. If throughput held while WIP dropped, lead time fell exactly as predicted. Lock the cap into standard work so it does not creep back.
What does Little's Law look like across different production systems?
The law is unit-agnostic, so it works on a cell, a whole plant, or an order backlog as long as the terms share a boundary and a time base. The table shows the same arithmetic in three settings. Notice that lead time always lands in the same time unit as throughput.
| Production system | Average WIP | Throughput | Lead time = WIP ÷ throughput |
|---|---|---|---|
| CNC machining cell | 900 parts | 150 parts/day | 6 days |
| Packaging line | 3,600 cases | 1,200 cases/day | 3 days |
| Weld shop, per shift | 240 parts | 60 parts/hour | 4 hours |
| Open order backlog | 75 orders | 25 orders/day | 3 days |
The backlog row is the one planners forget. An order sitting on the schedule is WIP the moment the customer expects a date, so the same law predicts your quoted lead time from the size of the open-order book divided by how fast the plant closes orders. If the book is growing, the boundary is not stable and the snapshot will understate lead time, a warning covered next.
Where does Little's Law break down in a real plant?
Little's Law is exact only for a stable system averaged over a long enough window, meaning WIP, throughput, and lead time are not trending. During a ramp-up, when WIP is growing faster than parts exit, a snapshot understates lead time because the parts entering have not yet cleared. The same happens when the boundary leaks: parts scrapped mid-stream, jobs split or merged, or a rush order jumped to the front all drift the counts.
The law also speaks only to averages, not to spread. Two lines with identical 3-day average lead times can have very different worst cases if one has erratic queues. To promise a delivery date you still need to understand variability, which is where statistical process control and buffer thinking come in. And the throughput term must be the honest exit rate; if you plug in a nameplate speed instead of what machine monitoring actually records, the predicted lead time will be optimistic by exactly the amount your line loses to downtime and micro-stops. Used with those caveats, the law is a reliable compass, not a fragile estimate.
How does Little's Law connect to OEE and lean flow?
Little's Law and OEE attack lead time from two ends. OEE raises the honest throughput of the constraint by recovering the six big losses, and Little's Law converts any WIP cut into a proportional lead-time cut at that throughput. Raise OEE on the bottleneck and you lift the throughput term; cap WIP and you shrink the numerator. Do both and lead time falls faster than either alone. The Theory of Constraints frames which lever to pull first: if you cannot safely raise constraint throughput yet, the only honest way to shorten lead time is to drain WIP, exactly what the law predicts.
Map the flow with value stream mapping to see where WIP piles up, then price the lead-time payoff of draining each queue with the law. Watch how much of the floor's real work is queue versus run in a capacity utilization and cycle time view. Seeing true WIP and true exit rates in real time, rather than reconstructing them from a spreadsheet a week later, is exactly what a live operations layer gives a planner, the connected-floor data behind the CLS case study and the analytics on Harmony's platform.