Late Night Discussions
Number 8

by Dr. Eliyahu M. Goldratt

"Jonah, have you noticed that, lately, wherever one goes, the phrase 'paradigm shift' keeps popping up?"

"What is the big surprise?" he says. "What do you expect when organizations are actually, right now, struggling to go through a whole set of paradigm shifts? Alex, what is actually bothering you? Are you looking for a better understanding of this phrase?"

"I guess so," I agree. "Paradigm, I think I understand. It is supposed to represent one's way of looking at a subject. Shift, probably indicates that this way, this conceptualization, has been changed. What I'm missing is the importance people seem to give to the combination of these two words."

"There is a bit of confusion," Jonah admits. "We are using this phrase a little bit too freely. You see, paradigm shift implies that the shift causes a drastic change in our actions combined with the emergence of many, very important, new questions."

"What you say definitely fits the most popular example of a paradigm shift: Copernicus' suggestion, that the planets are not circling around Earth but rather they and Earth itself are actually circling around the sun. But Jonah, do you really claim that industry is currently going through such revolutions? Can you give me even one concrete example? Actually it shouldn't be too difficult," I add sarcastically, "since you said that industry is right now struggling to assimilate a few of those paradigm shifts."

"No problem," Jonah calmly says and puts his book aside. "Why don't we start with a paradigm shift that you yourself put your production plants through, the logistical-paradigm-shift."

He took me by surprise. I certainly caused my plants to go through a major change (see The Goal for details) but I never thought about it as a paradigm shift.

Jonah, realizing my astonishment, elaborates, "Your paradigm of the shop floor was changed the minute that you shifted your definition of resource efficiency. Alex, can you recall what your previous understanding of that term was?"

"Frankly, it's quite embarrassing," I laugh. "Resource efficiency was defined by two parameters: what percent of their time the resources (people and machines) devoted to production, and how fast they produced."

"Fine," he says. "Now let's understand a little bit better the paradigm that stemmed from this definition. Look on the bind that material managers found themselves in," he explains. "On the one hand they had to guarantee that production would be able to achieve high traditional efficiencies; that any resource (especially people) would never run out of work and that the work would be done efficiently, that is, a lengthy setup would not be 'wasted' on a relatively short run. This situation forced the scheduler to consider releasing material required for remote customer orders or even for forecasts."

"That's obvious," I say. "But where's the bind? Where is 'on the other hand'?"

"You are forgetting the customers with their specific demands for specific quantities on specific dates."

"Yes, I see. Whenever an urgent demand, or a change in an existing demand occurred, material managers wished that all resources were not already busy with less urgent work. In other words, to enable fast response to changes, it would have been advantageous to consider only the short term demands."

"And how did we cope with this apparent conflict?" Jonah puffs on his cigar.

"I don't know how others have coped with it, but I'll tell you how we did it, if you promise not to laugh. We simply split the job. We had people in charge of pushing material to the floor, and from the other end we had a whole bunch of expeditors trying to pull out the already late orders. This awkward situation, as you are very well aware, no longer exists, and I do admit that we abandoned our traditional measurements, we no longer chase workers' efficiency. But Jonah, we didn't replace it with any new measurement."

"Really? What is your definition for resource efficiency now?" he continues his questioning.

For a while I don't answer. We don't have any new formal definition. "Satisfying the customers?" I half ask, half say.

"This is what you call a definition? " he gestures with his cigar.

"The extent to which resources supply what the customers demand," I correct my definition. "I guess that's the best I can give you."

"Not good enough," he says, "you are using a more elaborate one." When he realizes that I'm not following him, he continues, "this definition of yours, 'the extent to which resources supply what the customers demand' might be sufficient if the customers' demands would have been stable both in quantity and in mix of products demanded. Is it the case in your company, or even for most organizations? If not, what is missing from your definition?"

"You're right," I readily agree. "It's not just the amount supplied to the customers, it is also the speed in which we are able to react to the customers' demands."

"Very well," he says. "Once you shifted your concern from chasing local efficiencies to defining resource efficiencies in terms of customer satisfaction, what different actions did you have to take?"

I smile. If Jonah wants me to play back what he caused me to realize sometime ago, it's fine with me. That doesn't mean that I'm going to let him off the hook on the subject of paradigm shift. In a good mood I start.

"Shifting our attention to what comes out at the end of the pipe forced us to acknowledge that we are dealing not with isolated links but with a chain of resources that are dependent in their performance on each other. From there it was a tiny step to ask 'what determines the strength of the chain?' Of course, the weakest link. We started to concentrate on squeezing the maximum from the weakest link – the constraints - while the concern on all other resources was no longer their local efficiencies but rather that they will never starve the constraints.

"Jonah, I admit that it caused a drastic change in actions. We have choked the quantities of released material to be in line with the rate of consumption of the constraints. We timed the release to be always some paranoia lead time before the constraints will need it. In those product lines where we didn't have any resource constraints, we choked and timed the material release according to the due dates of the orders.

"I gladly admit that the results were dramatic. Due to the fact that we were concentrating on the constraints, throughput of the plant was considerably higher, inventories took a nose dive and customer service - even though they kept on surprising us – reached almost 100 percent. This is all old news. But Jonah, according to your own definition we still cannot call it a paradigm shift; one major thing is missing. This shift didn't cause us to have any open questions. Isn't this condition, opening new frontiers, a very important condition?"

"Alex, Alex, you never stop to amaze me. You don't have any open questions. How lovely. Let me ask you, my brilliant friend, haven't you realized that you are actually just guessing the time interval that you have to release the material before it must be consumed by the constraints - the time buffer?"

"Yes, of course," I answer cheerfully. Jonah is grasping at straws. "But what am I going to gain by determining my paranoia more scientifically? Another relatively small reduction in inventory?"

"No, no, you are missing the point," Jonah laughs. "You are, what you call, paranoid because you are very well aware that Murphy exists. You simply don't want that when Murphy hits one of the non-constraints the temporary stoppage of the work-stream will cause the constraints to starve. You release a time-buffer ahead of the theoretical time so that the constraints, resources or customer orders, will always be protected by products ready for their immediate consumption. What determines the magnitude of the protection is your estimation of the magnitude of the stoppages that Murphy may cause in the smooth flow, and the spare capacity available at the non-constraints to recoup from any such delay."

"That's correct," I answer, much less confident of myself.

"Thus, you must have realized that there is a trade-off between the amount of spare capacity on the non-constraints and the length of the time-buffer that you use for timing the material release. To guarantee very good customer service you are currently overprotecting for most orders; that is unavoidable when you use a time-buffer based on averages. What would have been the results if you knew how to determine the time-buffer for each order individually according to the queues that it might face in its journey to the constraints? In other words, if you knew how to use a dynamic time-buffer rather than a fixed one based on averages? The additional gain in inventory reduction might be relatively small, but the additional shrink in average response time can easily be another factor of two.

"The open opportunity is actually much bigger," he keeps on hammering. "Today you cope with product mix changes by oscillating the available capacity through, mainly, overtime. This avenue is not only expensive and annoying but also quite limited. What will happen if a significant change in mix occurs or the level of sales goes up? Your excellent customer service will deteriorate, and the customers, who by now are spoiled, will definitely not like it. Have you thought about using a dynamic buffer rather than oscillating the capacity?

"How can it be done?" I ask in astonishment.

"Well my friend, here is your open question. One out of many."

This "Late Night Discussion" is Copyright © 1991 Dr. Eliyahu M. Goldratt