CPG Scheduler
All posts

Little's Law for CPG Schedulers: Predict Line Lead Time from WIP and Throughput

schedulingcapacityinventory planningoeestockouts

Little's Law for CPG Schedulers: Predict Line Lead Time from WIP and Throughput

Here is a scene every scheduler knows. The filler is humming. The run-rate report looks healthy. The line has not gone down all shift. And yet the promised ship dates keep slipping, the expeditor is on the floor again, and someone in sales wants to know why an order released Monday still is not on a truck Thursday.

The instinct is to blame throughput — the line must be running too slow. But when you actually check the rate, it is fine. So what gives?

The answer is almost never the run-rate. It is the pile of work sitting in front of each job — the staged totes, the un-palletized cases, the orders queued behind the one on the machine right now. That pile has a name (work-in-process, or WIP) and it obeys a law so simple and so robust that you can schedule against it. It is called Little's Law, and it turns your "flow feel" into a number.

The law in one line

Little's Law is a queueing-theory theorem from John Little. In its formal form it says the long-run average number of items in a stable system (L) equals the long-run average arrival rate (λ) multiplied by the average time an item spends in the system (W):

L = λW

For the shop floor, rearrange it into the form you will actually use:

Lead Time = WIP ÷ Throughput

That is the whole tool. The average time a case (or an order) spends flowing through your line equals how much work is in the line divided by how fast work is leaving it.

The reason this is worth trusting — and worth arguing with your operations team about — is its robustness. Little's Law does not care about the details. As the standard statement puts it, the relationship "is not influenced by the arrival process distribution, the service distribution, the service order, or practically anything else." Your product mix, your changeover pattern, whether you run FIFO or hot-list a rush order — none of it changes the arithmetic. Over a stable window, lead time is WIP divided by throughput. Full stop.

The three terms in CPG language

Before you apply it, define the terms the way they show up on your floor.

WIP (work-in-process). Everything that has been released into the process but has not yet come out the far end as finished, shippable product. On a beverage or food line that includes blend tanks waiting to fill, product staged between fill and pack, cases waiting to be palletized, and — importantly for schedulers — the queue of released orders waiting their turn on the line. If it has been committed to the line but is not done, it is WIP.

Throughput (exit rate). The rate at which finished units actually leave the process — cases per hour that are done and shippable, not the nameplate speed of the filler. This is the number most people get wrong. Your real exit rate already has downtime, minor stops, speed loss, quality rejects, and changeovers baked into it. If you want an honest throughput figure, use your OEE-adjusted, schedulable run-rate rather than the equipment's theoretical maximum. (If that distinction is fuzzy, see OEE and the schedulable run-rate.)

Lead time (flow time). The clock time from when an order is released to the line until it is finished and ready to ship. Not the touch time. Not the run time. The wall-clock elapsed time, which includes all the waiting.

That last point is the whole game. Most of an order's lead time is not spent being processed — it is spent waiting in WIP. Little's Law is how you quantify that waiting.

A worked CPG example

Suppose a pack line is holding 4,000 cases of WIP — product staged, cases queued, orders released and waiting — and it is clearing finished pallets at a throughput of 500 cases per hour.

Lead Time = 4,000 ÷ 500 = 8 hours

So any case entering that line right now will, on average, take eight hours to come out shippable — even though the actual pack-and-palletize touch time might be minutes.

Now here is the part that changes how you schedule. Suppose to "keep the line busy" you release more work, and WIP climbs to 8,000 cases. Throughput has not changed — the line still clears 500 cases/hour.

Lead Time = 8,000 ÷ 500 = 16 hours

You doubled the lead time without touching the run-rate. The line is running exactly as fast as before. Every order now takes twice as long to clear, because every order has twice as much stuff sitting in front of it. This is the mathematical reason "the line is running but orders are late": someone added work to a busy line, and lead time grew in lock-step.

The iSixSigma framing of this is what lean and Six Sigma practitioners call Process Lead Time (PLT) = WIP ÷ Exit Rate. Their worked example is identical in shape: a process holding 50 units of WIP with a completion rate of 2 units/hour has a lead time of 25 hours (50 ÷ 2). Same law, smaller numbers.

Apply it to nested systems

One of the most useful properties of Little's Law is that it "applies to any system, and particularly, it applies to systems within systems." You can point it at the whole plant, at one line, at a single work cell, or at a single machine — and the arithmetic holds for each.

That is how you find where your lead time is actually being manufactured. Measure WIP and throughput at each stage:

  • Whole line: total released-but-unfinished ÷ finished cases/hour.
  • Bottleneck station (say, the filler): units queued in front of it ÷ its exit rate.
  • Finishing / palletizing: cases waiting to palletize ÷ palletize rate.

Apply the law station by station and the queue will announce itself. Wherever WIP is piling up relative to that stage's exit rate is where the delay lives. Almost always it is in front of your constraint — the slowest resource that governs the whole line's output. That is exactly the resource Drum-Buffer-Rope scheduling is built to protect and pace. Little's Law tells you how big the buffer in front of it has grown, and therefore how much lead time it is adding.

Three levers to cut lead time

Once you accept Lead Time = WIP ÷ Throughput, there are exactly three ways to make lead time shorter. There is no fourth.

1. Reduce WIP. Cap how much work you release onto the line. If you hold WIP to a target, lead time stays bounded no matter how tempting it is to "get ahead." A CONWIP-style rule — release a new order only when a finished one exits — is the cleanest way to enforce this. Lower WIP is not just faster; it also exposes and reduces waste, since problems can't hide behind a mountain of queued inventory.

2. Raise real throughput. Increase the exit rate of the constraint through faster changeovers (SMED), debottlenecking, or reducing minor stops. Note the word real: raising nameplate speed does nothing if the constraint's actual exit rate doesn't move.

3. Stop over-releasing to a full line. This is the one schedulers get wrong under pressure. When orders are late, the instinct is to push more work onto the floor to "catch up." Little's Law says that does the opposite — more WIP with unchanged throughput means longer lead time. You cannot expedite your way out of a WIP problem by adding WIP. This is the mechanism behind the production line utilization trap: drive a line toward 100% loaded and queues (and lead times) blow up.

Notice the first and third levers are about restraint, which is why they are hard. The disciplined scheduler resists the urge to release everything just because the line is available.

Use it as a promise-date and diagnostic tool

Little's Law is not just an explanation after the fact — it is a forward-looking planning instrument.

Quote realistic promise dates. If your line runs at 500 cases/hour and currently carries 6,000 cases of WIP, a new order released now sits behind roughly 12 hours of queue before it even clears. Use that to give sales an honest available-to-promise, instead of a date built on touch time that ignores the queue.

Detect silently growing queues. Track WIP and throughput on the same clock. If WIP is climbing while throughput is flat, lead time is degrading right now — before a single order is officially late. That is your early-warning signal to stop releasing.

Set WIP caps to a lead-time target. Work the formula backward. If you promise a 6-hour line lead time and you clear 500 cases/hour, your WIP cap is 6 × 500 = 3,000 cases. Hold WIP at or below that number and you will hit the promise. This turns a service-level goal into a concrete floor rule.

Pitfalls and assumptions

Little's Law is bulletproof, but only when you feed it honestly.

  • It assumes a stable (roughly stationary) system. The averages hold over a consistent window where things aren't wildly ramping up or draining down. Measure WIP and throughput over the same representative period. A snapshot taken mid-changeover will lie to you.
  • Use the process (bottleneck) cycle time, not a fast step. Lead time and cycle time are linked through WIP, but the governing cycle time is the constraint's, not that of some quick station upstream. Measuring throughput at a fast step will make your predicted lead time far too optimistic.
  • Changeovers are part of throughput. If your line stops for 45 minutes every product switch, that time reduces your real exit rate. Fold it into throughput; don't pretend the line clears cases during a changeover.
  • Count all the WIP, including queued orders. The staged product is obvious. The orders released but waiting are easy to forget — and they are exactly the WIP that drives promise-date slippage.

Takeaways

Little's Law gives schedulers a number where they used to have a feeling. Lead time is not a mystery that depends on luck or mix — it is WIP divided by throughput, every time, in any stable system.

  • When the line is running but orders are late, look at WIP, not run-rate.
  • Doubling WIP doubles lead time with zero change in speed.
  • The only three levers are: cut WIP, raise real throughput, or stop over-releasing.
  • Set your WIP cap from your lead-time promise and hold it.

Combine this with a run plan built on takt time and a schedule that respects your real, OEE-adjusted capacity, and you can predict a lead time before you promise it — instead of explaining a late order after the fact.

Sources

More from the journal