Critical Chain: A Business Novel

by Eliyahu M. Goldratt

time," I say. "And we all suffer from it. Call it meetings, call it emergencies, call it other jobs. The impact is the same. Lead time inflates. If you think about it, whenever you give a time estimate you know that the actual time is just a fraction of your estimate, but you intuitively factor in the impact of multi-tasking."

"Wait a minute," says Mark. "There is something wrong here. In our company, we have to come out with a new product every six months or so. Therefore, if we increase all the safety times, we increase the lead time of all projects. That means that more projects will be going on at the same time." I don't see his problem. "Correct, but what's your problem?" I ask. Slowly, he continues. "More projects means multi-tasking. And according to what we said, it means that lead times will len...

How come we allow multi-tasking?" "Isn't it obvious," I answer. "Without multi-tasking we might run into times where people don't have enough

work. Efficiencies will go down."

"Isn't the only important thing to guarantee that the projects will be successful." Mark joins in. "And on the bottleneck, with multi-tasking or without it, we'll have enough work."

"We found three mechanisms to put safety in. Now it seems that we also found three mechanisms to waste that safety. One we called the student syndrome, there is no rush so start at the last minute. The second is multi-tasking. The third involves the dependencies between steps; these dependencies cause delays to accumulate and advances to be wasted.

"Don't forget that in practice, machines are not working smoothly,"

"And don't assume any neat pattern. A worker may slow down, tools can break, materials jam. You never know; you only know that it happens."

"Under the realistic conditions that Rick so vividly described, how can we guarantee that our X machine will always have enough material so that it will be able to work constantly?"

"Put enough stock in front of it," Jim collaborates.

"Good idea." And Johnny draws, in front the circle with the X, a hump that is supposed to represent a pile of stock. "Now, as Rick told us, Murphy hits; one of the work centers upstream has a problem. The flow of ...

Jim's suggestion there's a pile of material ready. We can continue to utilize our X machi...

"But what, unavoidably, happens during this time? The flow of material to X stops and X continues to work, pulling from its pile. The amount of inventory in the pile must go down."

"Should we agree that Murphy does not strike only once? That sooner or later one of the feeding machines will, once again, stop?"

"And if so, from time to time, the pile will continue to be drained. Can we afford to let the init...

"If we do, the next time Murphy hits one of the feeding machines, our X machine will be starved, it won't work for one hundred percent of its time. So, if we don't want this initial pile to be drained to zero, when the problem in the feeding machines was corrected and the flow resumed, what did we have to do?" "Rebuild the pile in front of X," Jim plays Johnny's trivial question and answer game.

"But to do that," Johnny says triumphantly, "the feeding machines have not only to supply the ongoing rate of X, at the same time they also have to rebuild the stock. Quickly, before Murphy hits again. Which means . . ." he looks directly at me, "which means that each one of them must have more capacity than X."

"Conclusion. If we want to utilize even one resource to one hundred percent, all its feeding work centers have to have more capacity. Since Murphy is not negligible, and the feeding machines have limited time to rebuild the stock,...

"How much excess capacity do the feeding work centers need to have?" And Johnny answers, "It depends on the magnitude of the breakdowns, not less on the frequency of the breakdowns, and, of course, the amount of the stock you allow or want to build in front of X."

"It's the same," Johnny answers. "Production lead time is the time elapsed from the minute the first soldier stepped on a particular point until the last soldier stepped on the same point. So, the bigger the distance between the first row and the last row, the longer the lead time. Work-in-process inventory and lead time are twin brothers."

"The officer will stop the troop, regroup the soldiers and then they'll continue. Stopping the troop means that throughput has been lost. We see that even in this simple ‘plant' there are problems. On average, lead time is too long, and from time to time throughput is lost."

Still not knowing the answer, I try to joke, "We can tie the soldiers to each other, with chains." "That's the assembly line," Jim jumps up and starts to pace. "That's the conveyor belts of the assembly line."

"Suppose that in an assembly line one work center is faster than the work center downstream from it. The conveyer belt between them will be full of products, the chain is tight. If that conveyor belt is full, our fast work center cannot continue to produce at it's own rate.

"The same is true for Just-In-Time," Jim is saying slowly. "JIT doesn't use conveyor belts, it uses containers, of which a limited number are allowed to accumulate between work centers. It's exactly the same concept."

"So, what is the essence of these methods?" He continues to ask, "Why do they work so well?" And then answers, "All they've done is to put a cap on the amount of inventory they allow to accumulate between each two centers. Once the local inventory reaches its cap, the work center generating it is not allowed to continue producing at one hundred percent of its capability."

"Now we see methods, the assembly line and JIT, that do work. Which means that either they follow the five steps or the five steps are wrong." Thank you, Jim. Now I know how to continue. "It's apparent that assembly lines and JIT do not follow the five steps. Not only don't they start with identifying the bottleneck, they are not even considering the existence of a bottleneck at all. So where is the mistake in the five steps?" Johnny looks at us, then at the board. Then he sits down. "I don't follow you," Charlene says to Jim. "You talk as if it's all or nothing. What happens if JIT follows just one step? Won't it yield better results than a method that doesn't follow any?" "It will," I agree. "But which of the five steps does JIT follow? It's apparent that it doesn't obey the first or second." "What about the third?" she asks. "It forces work centers to work at less than their maximum local efficiency. It does force subordination." "Yes, it does," I once again agree with her. "But then . . ." Something is still wrong. "But then," Johnny continues my trend of thoughts,

"then, if we follow all five steps, not just one, we are bound to get a better method. And that's exactly what we get."

"I haven't gone that far. I just want to check if the problems that I've identified in projects also exist in assembly lines."

"As we said, the chains actually symbolize a restriction on the amount of inventory allowed to be accumulated locally," I say. Johnny already drew a hump of inventory in front of the X soldier. Mindlessly I add

humps between each two soldiers. "Johnny," Jim says, "can't we regard these humps as queues before machines?" "That's exactly what they are." "Let's not talk inventory," I impatiently say. "Let's talk time." "Go ahead." Johnny is very patient. "If this work center has a problem," I point arbitrarily to one of the soldiers, "then the hump represents the time that the next soldier can still work before he will have to stop. In a way," I slowly say, "these humps represent the protection a work center has. Protection against problems occurring upstream." "You could say so," Johnny agrees. "They represent safety." "Safety. Safety. You are right. Here is the connection to projects, don't you see?" Y-E-S! "I told you that we have to relate to inventory as representing time. In production we protect a work center with inventory, in projects we protect a step with safety time." "I see what you mean," Jim comments, "and I agree. Still I think that there is a difference. In projects the situation is worse."

"Why?" "Because, if there is a stoppage, inventory does not disappear. In projects...

Charlene comes from left field with "I have a problem with all of that. Why do we do that? Why do we try to protect the performance of each work center? I thought that we already agreed that local efficiencies don't count." Somehow these sentences sound familiar to me. Then I recall that Ruth's complaint was "We try to protect the performance of each step." I also recall what Ruth said next. "We put in so much safety, and the project...