Find your bottleneck from WIP data, not gut feel.

Ask three supervisors where the bottleneck is and you will get three answers. Each one is sure. Each one points at a different machine. The shop cannot fix three bottlenecks at once, so it fixes none of them, and the late list grows. The work-in-process on your floor already knows the answer. You just have to count it the same way every day and do a little arithmetic.

This post covers what WIP tells you, the one formula that turns a count of jobs into lead time, a 15-minute morning routine for reading WIP by cell, and what to do once the numbers point at a constraint.

WIP is evidence, not clutter

The Lean Enterprise Institute defines work-in-process as "items of work between processing steps," and notes that in a lean system the standardized WIP is the minimum number of parts, including units inside machines, needed to keep a cell flowing.^[1] Everything above that minimum is waiting. In a batch-and-queue shop, large lots get processed and pushed to the next operation whether or not that operation needs them, and there they sit in a line.^[2]

Where that line is longest is not random. Goldratt's definition of a constraint is the thing that "limits a system from achieving higher performance versus its goal."^[3] By definition, every non-constraint has more capacity than the constraint does. The Theory of Constraints Institute puts the consequence bluntly: left unchecked, that imbalance "results in bloated WIP inventory, elongated lead times, and frequent expediting/firefighting."^[4] Work arrives at the constraint faster than it leaves. The pile grows there and nowhere else.

In my shops, the pile was always visible. Carts lined up along one wall, racks full in front of one cell. What nobody had was a count. Everyone had an opinion about which cell was the problem. Good people, scattered data, nobody looking at the same numbers.

Little's Law: the only formula you need

In 1961, John D. C. Little published a short proof in the journal Operations Research. He defined L as the mean number of units in a system, W as the mean time a unit spends in it, and λ (lambda) as the arrival rate, and showed that under general conditions L = λW.^[5] Fifty years later he wrote that the result had "become widely known as 'Little's Law'" and that it is, at bottom, a theorem in queuing theory.^[6]

Operations management rewrote it in shop terms. Little's own 50th anniversary paper gives the form used by Hopp and Spearman in Factory Physics: TH = WIP / CT, where TH is throughput in items per unit time, WIP is work in process in items, and CT is cycle time in time per item.^[6] Hopp and Spearman define throughput as the average output of a process per unit time, WIP as the inventory between the start and end points of a routing, and cycle time as the average time from release to completion.^[6] The third edition of their book, published by Waveland Press, devotes a chapter to these basic factory dynamics.^[8]

Rearranged, the law gives you three tools from one count:

Two conditions apply. The numbers are long-run averages, and the system has to be stable, which still allows WIP to rise and fall, batching, and the occasional breakdown.^[7] A single morning count will not be exact. A week of counts gets close, and that is good enough to settle the argument about which cell is the constraint.

What the law tells you about late orders

Little drew out the point every plant manager needs. If output is the goal, then "any increase in output requires either an increase in work-in-process inventory or a reduction in cycle time or both, unless some change is made in the manufacturing process itself."^[6] The same throughput can be had with a lot of WIP and a long cycle time or with a little WIP and a short cycle time.^[7] Most shops that are behind have chosen the first option without knowing it. They released more work to "get more things going," and the extra work just lengthened the line.

Hopp and Spearman also define a critical WIP level: the amount that achieves the bottleneck rate at the minimum cycle time in a line with no variability.^[7] WIP above that level adds waiting, not output.^[7] Real shops have variability, so the exact number is a benchmark rather than a target, but the direction is clear. Past a certain point, more jobs on the floor means later jobs, not more jobs.

Utilization makes this worse. Wait time in a queue grows faster and faster as utilization rises, and the Project Production Institute's summary of the Factory Physics work notes that pushing a bottleneck beyond 80 to 85 percent utilization makes a "cycle time explosion" very likely.^[7] The TOC Institute uses very high utilization of at least one internal resource, 95 percent or more, as the real test of an internal capacity constraint.^[3] If you have read why 85% machine utilization can still mean late orders, this is the same physics from the other side. Utilization tells you how busy the machine is. WIP tells you how long the customer waits.

The 15-minute morning WIP read

You do not need a system to start. You need a clipboard, a fixed route, and a fixed time. In my experience the discipline matters more than the tool: same person, same route, same time, every day, for at least two weeks before anyone draws a conclusion.

Walk the floor before first shift settles in. At each cell, count the jobs waiting to be worked, not the job on the machine. Write the count next to the cell's name. If jobs vary a lot in size, count standard hours instead of jobs. Then, back at the desk, divide each cell's count by what that cell typically finishes in a day. That number is days of queue, which is the cell's share of your lead time.

Reading the three patterns

One pile, downstream starved. This is the textbook constraint. Little's Law says the cell with the most days of queue is contributing the most lead time, and the starved cells behind it confirm that it is holding everyone up. Fix this cell first and nothing else until it is fixed.

A wandering pile. When the biggest queue moves around from week to week, the floor is usually generating its own variability: big batches that arrive all at once, hot jobs that jump the line, machines pulled off one job to start another. The TOC Institute names expediting and firefighting as symptoms of the same root cause, non-constraints producing more than the constraint can handle.^[4] Stop looking for a machine to blame and look at the rules.

Piles everywhere. When every cell is carrying days of queue, more work is being released than the shop can complete. The TOC answer is to choke the release of raw material to the pace of the constraint.^[4] The lean answer is the same from a different direction: determine the inventory a process needs and keep reducing it, but only after you have reduced variability and improved capability, because cutting WIP without doing that work "will only disappoint the customer."^[9] In practice that means you slow release and fix the constraint in the same month, not one or the other.

Then follow Goldratt's five focusing steps

Goldratt first published The Goal in 1984, a novel about a plant manager named Alex Rogo whose shipments are always late while inventories keep rising, and a boy scout named Herbie who comes to stand in for the bottleneck.^[10] The method Rogo learns is the five focusing steps, which the TOC Institute also calls the Process of On-Going Improvement.^[4] Your morning WIP read is step one. The rest follow.

1. Identify the constraint. The worksheet does this. You are no longer guessing.
2. Exploit the constraint. Squeeze everything out of what you already have before buying anything. The TOC Institute's list of exploit actions includes measuring OEE at the constraint, staggering breaks so it never stops, staffing it fully with trained operators, putting a quality check immediately before it so it never works on bad parts, and keeping tooling and consumables ready.^[4] It warns that the constraint in most organizations is not well used, often under 50 percent on a 24x7 basis.^[4]
3. Subordinate everything else. The rest of the shop runs at the pace of the constraint. Upstream cells stop building piles in front of it. Downstream cells stop pulling parts out of sequence. Release is gated by what the constraint can consume.^[4]
4. Elevate the constraint. Only now do you add a shift, a machine, or a person. The TOC Institute warns that teams tend to skip the first three steps and jump straight to buying capacity.^[4] That matches what I saw: the new machine usually arrived after the old one had already stopped being the problem.
5. Go back to step one. Once the constraint moves, the old rules stop working. The TOC Institute's fifth step is to prevent inertia from becoming the constraint.^[4] Your morning read will show you where the pile went.

For a fabricator this cycle takes weeks, not quarters. The read takes 15 minutes a day. The arithmetic takes five. The hard part is not doing anything else until the first constraint is handled.

What it looks like when it works

The MEP National Network publishes case write-ups, and two of them read like the worksheet above.

Island Components Group, a Long Island motor and actuator maker, was meeting its daily target of 75 units per sub-assembly when its equipment effectiveness data flagged an inflated work-in-process number.^[11] Working with the New York MEP center, the team trained on just-in-time, jidoka, and standard work, set up kanban for feeder lines, and moved to one-piece flow with what the write-up calls strategic bottleneck management.^[11] Output rose from 75 to 120 units a day and WIP fell from 717 pieces to 156.^[11] Same people, same machines, same building.

Stainless Works, an Ohio header and exhaust manufacturer, had long lead times and no clear picture of capacity.^[12] MAGNET, the Ohio MEP center, walked the process and found that the routings in the ERP were feeding it wrong information.^[12] Once the routings were fixed and work centers set up for scheduling, welding was identified as the main bottleneck, and forward scheduling through the capacity-constrained weld department cut lead time from 12 to 14 weeks down to 5.^[12] Notice the order: see the real data first, find the constraint second, schedule around it third.

NIST MEP's own description of cellular flow lists the benefits as increased productivity, reduced lead time, reduced inventory, and faster resolution of quality problems.^[13] Little's Law explains why those four move together. Reduced inventory with the same throughput is a shorter cycle time by definition, and a shorter line means a defect gets found sooner, before a hundred more parts behind it carry the same flaw. LEI's definition of continuous flow is "producing and moving one item at a time (or a small and consistent batch of items)" with each step making only what the next step asks for.^[14] That is what a floor looks like when the piles are gone.

Keep the count going

The morning read works on a clipboard. What breaks it is not the method, it is the day the person who does it is out, the week two supervisors count differently, or the month when nobody has time to type the numbers into a spreadsheet and the trend disappears. The data was never scattered on purpose. It just lived in heads and on paper.

That is the problem HixonOps built its MES to handle: cell-level tracking so every job's location is recorded as it moves, and a live WIP view so the count by cell is already on the screen when you walk in. The arithmetic in this post runs on its own, every morning, from the same numbers everyone else sees.

The result I come back to is a steel products maker that had over $700,000 tied up in late orders. Six months later it was under $50,000. There was no new system. Everyone just looked at the same numbers every day. Start the count on Monday. Little's Law and Goldratt's steps will tell you what to do with it.