The Goal: A Process of Ongoing Improvement, Third Revised Edition

by Eliyahu M. Goldratt

  weeks or more," says Jake.

  E.M. Goldratt

  The Goal: A Process of Ongoing Improvement

  Captured by Plamen T.

  210

  "But we need them now," I say. "How come they're not

  being worked on?"

  Jake shrugs his shoulders. "You know which ones you want,

  we'll do 'em right now. But that goes against them rules you set

  up in that there priority system."

  He points to some other skids of materials nearby.

  "You see over there?" says Jake. "They all got red tags. We got to do all of 'em before we touch the stuff with green tags.

  That's what you told us, right?"

  Uh-huh. It's becoming clear what's been happening.

  "You mean," says Stacey, "that while the materials with

  green tags have been building up, you've been spending all your

  time on the parts bound for the bottlenecks."

  "Yeah, well, most of it," says Jake. "Hey, like we only got so many hours in a day, you know what I mean?"

  "How much of your work is on bottleneck parts?" asks Jo-

  nah.

  "Maybe seventy-five or eighty percent," says Jake. "See, everything that goes to heat-treat or the NCX-10 has to pass

  through here first. As long as the red parts keep coming—and

  they haven't let up one bit since that new system started—we just

  don't have the time to work on very many of the green-tag parts."

  There is a moment of silence. I look from the parts to the

  machines and back to Jake again.

  "What the hell do we do now?" asks Donovan in echo to my

  own thoughts. "Do we switch tags? Make the missing parts red

  instead of green?"

  I throw up my hands in frustration and say, "I guess the only

  solution is to expedite."

  "No, actually, that is not the solution at all," Jonah says, "because if you resort to expediting now, you'll have to expedite all

  the time, and the situation will only get worse."

  "But what else can we do?" asks Stacey.

  Jonah says, "First, I want us to go look at the bottlenecks,

  because there is another aspect to the problem."

  Before we can see the NCX-10, we see the inventory. It's

  stacked as high as the biggest forklift can reach. It's not just a

  mountain, but a mountain with many peaks. The piles here are

  even bigger than before we identified the machine as a bottle-

  neck. And tied to every bin, hanging from every pallet of parts is

  E.M. Goldratt

  The Goal: A Process of Ongoing Improvement

  Captured by Plamen T.

  211

  a red" tag. Somewhere behind it all, its own hugeness obscur ed

  from our view, is the NCX-10.

  "How do we get there from here?" asks Ralph, looking for a

  path through the inventory.

  "Here, let me show you," says Bob.

  And he leads us through the maze of materials until we reach

  the machine.

  Gazing at all the work-in-process around us, Jonah says to

  us, "You know, I would guess, just from looking at it, that you

  have at least a month or more of work lined-up here for this

  machine. And I bet if we went to heat-treat we would find the

  same situation. Tell me, do you know why you have such a huge

  pile of inventory here?"

  "Because everyone ahead of this machine is giving first pri-

  ority to red parts," I suggest.

  "Yes, that's part of the reason," says Jonah. "But why is so much inventory coming through the plant to get stuck here?"

  Nobody answers.

  "Okay, I see I'm going to have to explain some of the basic

  relationships between bottlenecks and non-bottlenecks," says Jo-

  nah. Then he looks at me and says, "By the way, do you remem-

  ber when I told you that a plant in which everyone is working all

  the time is very in efficient? Now you'll see exactly what I was talking about."

  Jonah walks over to the nearby Q.C. station and takes a piece

  of chalk the inspectors use to mark defects on the parts they

  reject. He kneels down to the concrete floor and points to the

  NCX-10.

  "Here is your bottleneck," he says, "the X-what-ever-it-is machine. We'll simply call it 'X.' "

  He writes an X on the floor. Then he gestures to the other

  machines back down the aisle.

  "And feeding parts to X are various non-bottleneck ma-

  chines and workers," he says. "Because we designated the bottle-

  neck as X, we'll refer to these non-bottlenecks as 'Y' resources.

  Now, for the sake of simplicity, let's just consider one non-bottle-

  neck in combination with one bottleneck . . ."

  With the chalk, he writes on the floor:

  Y —> X

  E.M. Goldratt

  The Goal: A Process of Ongoing Improvement

  Captured by Plamen T.

  212

  Product parts are what join the two in a relationship with

  each other, Jonah explains, and the arrow obviously indicates the

  flow of parts from one to the other. He adds that we can consider

  any non-bottleneck feeding parts to X, because no matter which

  one we choose, its inventory must be processed at some subse-

  quent point in time by X.

  "By the definition of a non-bottleneck, we know that Y has

  extra capacity. Because of its extra capacity, we also know that Y

  will be faster in filling the demand than X," says Jonah. "Let's say both X and Y have 600 hours a month available for production.

  Because it is a bottleneck, you will need all 600 hours of the X

  machine to meet demand. But let's say you need only 450 hours a

  month, or 75 percent, of Y to keep the flow equal to demand.

  What happens when Y has worked its 450 hours? Do you let it sit

  idle?"

  Bob says, "No, we'll find something else for it to do."

  "But Y has already satisfied market demand," says Jonah.

  Bob says, "Well, then we let it get a head start on next

  month's work."

  "And if there is nothing for it to work on?" asks Jonah.

  Bob says, "Then we'll have to release more materials."

  "And that is the problem," says Jonah. "Because what happens to those extra hours of production from Y? Well, that inven-

  tory has to go somewhere. Y is faster than X. And by keeping Y

  active, the flow of parts to X must be greater than the flow of

  parts leaving X. Which means . . ."

  He walks over to the work-in-process mountain and makes a

  sweeping gesture.

  "You end up with all this in front of the X machine," he says.

  "And when you're pushing in more material than the system can

  convert into throughput, what are you getting?"

  "Excess inventory," says Stacey.

  "Exactly," says Jonah. "But what about another combina-

  tion? What happens when X is feeding parts to Y?"

  Jonah writes that on the floor with the chalk like this . . .

  X —> Y

  "How much of Y's 600 hours can be used productively

  here?" asks Jonah.

  "Only 450 hours again," says Stacey.

  E.M. Goldratt

  The Goal: A Process of Ongoing Improvement

  Captured by Plamen T.

  213

  "That's right," says Jonah. "If Y is depending exclusively

  upon X to feed it inventory, the maximum number of hours it

  can work is determined by the output of X.
And 600 hours from

  X equates to 450 hours for Y. After working those hours, Y will

  be starved for inventory to process. Which, by the way, is quite

  acceptable."

  "Wait a minute," I say. "We have bottlenecks feeding non-

  bottlenecks here in the plant. For instance, whatever leaves the

  NCX-10 will be processed by a non-bottleneck."

  "From other non-bottlenecks you mean. And do you know

  what happens when you keep Y active that way?" asks Jonah.

  "Look at this."

  He draws a third diagram on the floor with the chalk.

  In this case, Jonah explains, some parts do not flow through

  a bottleneck; their processing is done only by a non-bottleneck

  and the flow is directly from Y to assembly. The other parts do

  flow through a bottleneck, and they are on the X route to assem-

  bly where they are mated to the Y parts into a finished product.

  In a real situation, the Y route probably would consist of one

  non-bottleneck feeding another non-bottleneck, feeding yet an-

  other non-bottleneck, and so on, to final assembly. The X route

  might have a series of non-botjtlenecks feeding a bottleneck,

  which in turn feeds a chain of more non-bottlenecks. In our case,

  Jonah says, we've got a group of non-bottleneck machines down-

  stream from X which can process parts from either the X or the Y

  route.

  "But to keep it simple, I've diagrammed the combination

  with the fewest number of elements—one X and one Y. No mat-

  ter how many non-bottlenecks are in the system, the result of

  E.M. Goldratt

  The Goal: A Process of Ongoing Improvement

  Captured by Plamen T.

  214

  activating Y just to keep it busy is the same. So let's say you keep

  both X and Y working continuously for every available hour.

  How efficient would the system be?"

  "Super efficient," says Bob.

  "No, you're wrong," says Jonah. "Because what happens

  when all this inventory from Y reaches final assembly?"

  Bob shrugs and says, "We build the orders and ship them."

  "How can you?" asks Jonah. "Eighty percent of your prod-

  ucts require at least one part from a bottleneck. What are you

  going to substitute for the bottleneck part that hasn't shown up

  yet?"

  Bob scratches his head and says, "Oh, yeah ... I forgot."

  "So if we can't assemble," says Stacey, "we get piles of inventory again. Only this time the excess inventory doesn't accumu-

  late in front of a bottleneck; it stacks up in front of final assem-

  bly."

  "Yeah," says Lou, "and another million bucks sits still just to keep the wheels turning."

  And Jonah says, "You see? Once more, the non-bottleneck

  does not determine throughput, even if it works twenty-hour

  hours a day."

  Bob asks, "Okay, but what about that twenty percent of

  products without any bottleneck parts? We can still get high efficiencies with them."

  "You think so?" asks Jonah.

  On the floor he diagrams it like this . . .

  This time, he says, the X and Y operate independently of

  one another. They are each filling separate marketing demands.

  "How much of Y's 600 hours can the system use here?" asks

  Jonah.

  "All of 'em," says Bob.

  "Absolutely not," says Jonah. "Sure, at first glance it looks as if we can use one hundred percent of Y, but think again."

  "We can only use as much as the market demand can ab-

  sorb," I say.

  "Correct. By definition, Y has excess capacity," says Jonah.

  "So if you work Y to the maximum, you once again get excess

  E.M. Goldratt

  The Goal: A Process of Ongoing Improvement

  Captured by Plamen T.

  215

  inventory. And this time you end up, not with excess work-in-

  process, but with excess finished goods. The constraint here is

  not in production. The constraint is marketing's ability to sell."

  As he says this, I'm thinking to myself about the finished

  goods we've got crammed into warehouses. At least two-thirds of

  those inventories are products made entirely with non-bottleneck

  parts. By running non-bottlenecks for "efficiency," we've built

  inventories far in excess of demand. And what about the remain-

  ing third of our finished goods? They have bottleneck parts, but

  most of those products have been sitting on the shelf now for a

  couple of years. They're obsolete. Out of 1,500 or so units in

  stock, we're lucky if we can sell ten a month. Just about all of the

  competitive products with bottleneck parts are sold virtually as soon as they come out of final assembly. A few of them sit in the

  warehouse a day or two before they go to the customer, but due

  to the backlog, not many.

  I look at Jonah. To the four diagrams on the floor, he has

  now added numbers so that together they look like this . . .

  Jonah says, "We've examined four linear combinations in-

  volving X and Y. Now, of course, we can create endless combina-

  tions of X and Y. But the four in front of us are fundamental

  enough that we don't have to go any further. Because if we use

  these like building blocks, we can represent any manufacturing

  situation. We don't have to look at trillions of combinations of X

  and Y to find what is universally true in all of them; we can

  generalize the truth simply by identifying what happens in each

  of these four cases. Can you tell me what you have noticed to be

  similar in all of them?"

  Stacey points out immediately that in no case does Y ever

  determine throughput for the system. Whenever it's possible to

  E.M. Goldratt

  The Goal: A Process of Ongoing Improvement

  Captured by Plamen T.

  216

  activate Y above the level of X, doing so results only in excess

  inventory, not in greater throughput.

  "Yes, and if we follow that thought to a logical conclusion,"

  says Jonah, "we can form a simple rule which will be true in every

  case: the level of utilization of a non-bottleneck is not determined

  by its own potential, but by some other constraint in the system."

  He points to the NCX-10.

  "A major constraint here in your system is this machine,"

  says Jonah. "When you make a non-bottleneck do more work

  than this machine, you are not increasing productivity. On the

  contrary, you are doing exactly the opposite. You are creating

  excess inventory, which is against the goal."

  "But what are we supposed to do?" asks Bob. "If we don't

  keep our people working, we'll have idle time, and idle time will

  lower our efficiencies."

  "So what?" asks Jonah.

  Donovan is taken aback. "Beg pardon, but how the hell can

  you say that?"

  "Just take a look behind you," says Jonah. "Take a look at the monster you've made. It did not create itself. You have created

  this mountain of inventory with your own decisions. And why?

  Because of the wrong assumption that you must make the work-

  ers produce one hundred percent of the time, or else get rid of

  them to 'save' money."

  Lou says, "Well, granted that maybe one hundred percent is

  u
nrealistic. We just ask for some acceptable percentage, say,

  ninety percent."

  "Why is ninety percent acceptable?" asks Jonah. "Why not

  sixty percent, or twenty-five? The numbers are meaningless un-

  less they are based upon the constraints of the system. With

  enough raw materials, you can keep one worker busy from now

  until retirement. But should you do it? Not if you want to make

  money."

  Then Ralph suggests, "What you're saying is that making an

  employee work and profiting from that work are two different

  things."

  "Yes, and that's a very close approximation of the second

  rule we can logically derive from the four combinations of X and

  Y we talked about," says Jonah. "Putting it precisely, activating a resource and utilizing a resource are not synonymous."

  He explains that in both rules, "utilizing" a resource means

  E.M. Goldratt

  The Goal: A Process of Ongoing Improvement

  Captured by Plamen T.

  217

  making use of the resource in a way that moves the system toward

  the goal. "Activating" a resource is like pressing the ON switch of a machine; it runs whether or not there is any benefit to be derived from the work it's doing. So, really, activating a non-bottle-

  neck to its maximum is an act of maximum stupidity.

  "And the implication of these rules is that we must n o t seek to optimize every resource in the system," says Jonah. "A system of local optimums is not an optimum system at all; it is a very inefficient system."

  "Okay," I say, "but how does knowing this help us get the

  missing parts unstuck at the milling machines and moved to final

  assembly?"

  Jonah says, "Think about the build-up of inventory both

  here and at your milling machines in terms of these two rules we

  just talked about."

  "I think I see the cause of the problem," Stacey says, "We're releasing material faster than the bottlenecks can process it."

  "Yes," says Jonah. "You are sending work onto the floor

  whenever n o n-bottlenecks are running out of work to do."

  I say, "Granted, but the milling machines are a bottleneck."

  Jonah shakes his head and says, "No, they are not—as evi-

  denced by all this excess inventory behind you. You see, the mill-

  ing machines are not intrinsically a bottleneck. You have turned them into one."

  He tells us that with an increase in throughput, it is possible

  to create new bottlenecks. But most plants have so much extra