The Theory of Constraints (TOC) provides a simple and practical approach to the problem of managing complex systems. In this chapter, we discuss the application of TOC to production or manufacturing environments. Production/manufacturing environments are among the most complex of systems, characterized by high levels of dependency and variability. Planning the work of the many resources (often 100 or more), procuring the supply of materials from vendors, and coordinating all of these tasks in such a way as to meet committed delivery dates are truly challenging tasks. The development of computers and computer-based planning systems has been a major facilitator for these challenging tasks. Unfortunately, computers have not been a panacea and, in many ways, the use of computers has aggravated the problem-for example, it has been the author's experience that the nervousness1 in manufacturing supply chains is higher when the supply chain is managed by a sophisticated Enterprise Resources Planning (ERP) (or material requirements planning [MRP]) system.

In this chapter, we first show the application of the TOC approach to managing production environments-known as Drum-Buffer-Rope (DBR) and Buffer Management (BM). DBR and BM are the systems that emerge from the application of the Five Focusing Steps (5FS). DBR is the TOC methodology for planning and BM is the TOC methodology for execution control. The term planning is used for those activities that start with known market demand and generate the plans for managing the flow of material through the factory including identifying what purchased materials will be needed and when. Execution control refers to the actions that are taken during the execution phase of the plan developed previously. These actions are necessary to ensure that the plans are followed and include the corrective actions that must be taken when deviations from the plan threaten to compromise delivery dates and Throughput of the system.

FIGURE 8-1 Resource centric representation of a plant producing three products (A, B, C) with four resources (R1, R2, R3, R4).

After explaining these systems and their logic with simple examples, we next move to a discussion of complex, real-life flows. Real-life production environments are characterized by high levels of detail complexity and high levels of dynamic complexity.2 Many of these elements, especially the ones in detail complexity, are specific to the individual environment and make each one appear to be different and unique. However, the behavior of these systems as a whole are characterized more by the way their dynamic complexity relates one to the other. These relationships and their many apparently different operations exhibit similar behaviors with respect to operational performance as measured by on-time deliveries, system inventories, production lead times, and so on. Third, we present a classification of production operations based on the structure of the product as contained in the bill-of-material and routing or process information. We classify the product flows into four major types-V, A, T, and I-or a combination of these four types. The real power of this classification is that operations that belong to a particular V, A, T, or I type will share similar performance characteristics and business problems to others in the same group. The application of DBR in each type of plant is also discussed.

Managing Flow-Planning and DBR

The Need for a Focus on Flow

A production operation is characterized by a number of resources that typically occupy fixed spots on the factory floor. Materials move from one resource to another in accordance with the rules specified in the routing sheet for the specific material/product. Typically, we think of the factory or production operation from this spatial or static perspective. We will call this a resource centric view of the operation. For the simple case of a factory that has four resources R1, R2, R3, and R4 and makes three products-identified as Product A, Product B, and Product C-the resource centric view of the operation is depicted in Fig. 8-1. The solid black line (-) represents the path in which Product A moves from RM 1 (raw material 1) through the various resources as it is converted from raw material to a finished product. The dotted line (. . .) represents the path in which Product B moves from RM 2 through the various resources, and the dash-dot line (-.-.-.) shows the path followed by Product C from RM 3 through the various resources. This resource centric viewpoint is also the viewpoint of traditional management methods. Cost control is the primary goal of operations management and the traditional view is that resources drain or consume costs. The way to manage cost is to manage the efficiency of each resource and to make sure that no time is wasted at any resource. Goldratt (2003) has aptly captured this viewpoint of traditional operations management in the phrase: "A resource standing idle is a waste." Consistent with this view, most measurements in operations are resource centric (local departmental measures such as efficiencies, utilization, downtime, etc.) and are designed to capture information on what resources were doing every second of the day.

An alternate viewpoint of the same factory floor is to look at how materials flow. Materials move through the factory, flowing from raw material to finished product. Along the way, they are transformed or worked on by resources. The material thus flows from raw material to one resource to another until it is fully transformed and the finished product leaves the factory or operation. We call this viewpoint a flow centric viewpoint. From a flow centric viewpoint, the same factory in Fig. 8-1 would be represented as shown in Fig. 8-2. Since there are three separate materials, there are three separate flows. The transformation of any product, such as Product A (represented by the solid line), from raw material (RM 1) to finished product can be represented by a unique sequence of operations-resource R1 performing operation 010, then resource R3 performing operation 030, etc. We have chosen the vertical direction to depict the time sequence of these steps. In this particular case, the three separate products are produced in very similar fashion; they follow identical paths. Figure 8-3 shows the resource centric and flow centric views when the three products have very dissimilar routings. These diagrams are referred to as Product Flow Diagrams or PFDs.

The essence of the manufacturing operation-the transformation from raw material to a finished product-is reflected in the flow centric view. It is not surprising that the management of production/manufacturing operations should be based on a flow centric view and not a resource centric view. In his article, "Standing on the Shoulders of Giants," Goldratt (2009) presents the core argument that Henry Ford's assembly line process and Dr. Taichi Ohno's Toyota Production System (TPS) originate from a focus on flow.

By a flow centric view, we mean much more than looking at production operations in the format of Fig. 8-2. The primary role of production management is recognized to be to manage flow. Effective management of flow implies that the movement of all material through the factory will be smooth and fast with no stoppages. In any flow, obstacles to the flow result in buildup of material-a traffic jam-and are considered highly undesirable. Resource centric methods are interested in keeping resources busy and consider buildup of material unavoidable.

FIGURE 8-2 A flow centric representation of the plant in Fig. 8-1.

FIGURE 8-3 A comparison between the resource centric representation and the flow centric representation of a plant where the routings for the products are dissimilar. [ E. M. Goldratt used by permission, all rights reserved. Source: Modified from E. M. Goldratt (2003, 29)]

To understand the key difference between the two views, think of it this way. When walking through the factory you are bound to see idle resources and idle batches of material. Which bothers you more in the pit of your stomach? If the resource standing idle bothers you more, then you are exhibiting a resource centric view. If the batch of material sitting idle bothers you more, then you are exhibiting a flow centric view. What we have learned from Henry Ford and Taichi Ohno is that the flow centric view is the proper point of view for effective management of the system.

Traditional cost accounting-based management methods, unfortunately, are resource centric in nature. Operators typically describe themselves in terms of the resource or resources they operate-press operator, furnace operator, etc. Managers also describe themselves in terms of the resources they control-Press Department, Heat Treat Department, etc. The entire management control system is geared to track the activities of the resources and, in particular, to track, understand, and hence eliminate non-production or idle time on the resource.

Ford and Toyota Production Systems-A New Perspective

In a groundbreaking article in 2009, Goldratt provided a new perspective on the two production methods that have defined this field in the last 100 years-the Henry Ford assembly line system and the Dr. Ohno TPS. Everyone knows that Henry Ford is the father of the modern assembly line mode of production, but most have focused on how this system enables better utilization of resources (material is brought to the worker) and it achieves the dream of balancing capacity. Goldratt took a different point of view. Henry Ford's real objective was to improve the flow of products through his factory. He was so successful at improving flow that in 1926, the elapsed time between unloading iron ore from the boat to the same iron being loaded on a freight train as a finished automobile was an astonishingly fast 81 hours (Ford, 1928). The magnitude of this achievement is underscored by the fact that eight decades later, no automobile manufacturer can come close to Ford's achievement. Contrary to the traditional belief that one cannot achieve maximum or full output without ensuring that all resources are productive and producing all of the time, Ford's method produced far more output from the factory as a whole. In fact, a focus on flow can result in some resources running out of work occasionally. However, system Throughput is not compromised, but actually enhanced. The success of Henry Ford clearly demonstrated that the resource centric view had led to false assumptions, but this lesson was mostly lost in history from 1926 to the 1970s and the emergence of TPS.

Goldratt concludes that both Henry Ford's assembly line and Taichi Ohno's TPS were systems in which achieving smooth flow through production was a prime objective and that the generalized method they followed can be summarized by the following four principles: 1. Improving flow (or equivalently lead time) is a primary objective of operations.

2. This primary objective should be translated into a practical mechanism that guides the operation when not to produce (prevents overproduction).

3. Local efficiencies must be abolished.

4. A focused process to balance flow must be in place (Goldratt, 2009).

Of particular significance is the second statement. Goldratt points out that the assembly line and the Kanban system of Toyota are essentially systems that tell workstations when not to produce. For instance, in an assembly line if one workstation stops, all others have to stop because the line stops and there is no place to put material if any of the other stations continue to produce. Similarly, in a Kanban system when there are no Kanban cards, work centers stop working. In contrast, in most traditional production operations one of the key arguments for maintaining significant work-in-progress (WIP) queues is to decouple each work center from other work centers and their possible disruptions.

Henry Ford relied on space to limit production while Taichi Ohno developed the Kanban system3 to do the same. Of course, if we are introducing a system that intentionally stops resources from producing, then clearly Principle 3 (abolish local efficiencies) is unavoidable. What is interesting is that both Henry Ford and Taichi Ohno did not simply stop at limiting production, but leveraged these situations into opportunities for improving processes that streamlined and increased the volume of flow. When the built-in mechanisms-space or inventory-create a line stoppage, one has clear visibility about what caused the stoppage and, hence, points to the problem that needs to be solved to better balance the flow. The magnitude of improvements that both Ford and Toyota were able to achieve over their competitors in increased speed and reduced total cost stands as testimony to the effectiveness of their approaches.

In spite of the tremendous success of their methods and the volumes of articles and books written about their methods, the focus on flow did not spread to all parts of the manufacturing industry. In a small country like Japan, given the clear success of Toyota as a business and their attribution of this success to TPS, one would expect wide adoption of TPS. In fact, less than 20 percent of the manufacturers have implemented TPS and few of these manufacturers have achieved Toyota's level of success. What is the cause for this low level of adoption and success? Certainly, it is not lack of desire or knowledge. Almost every company has attempted to adopt TPS or Lean Production as it is also known. There is an ocean of material on TPS and Lean and Toyota has been very open about its techniques. Goldratt (2009) concludes that the core issues are twofold: 1. The resource centric mindset is still the prevailing viewpoint. This explains why even when TPS is applicable and is adopted, the results are less than what is possible.

2. The specific mechanisms for preventing overproduction-space in the case of Ford's assembly line and inventory in the case of Ohno's TPS-are not applicable to all manufacturing environments.

In his article, Goldratt proposes a different and more universal mechanism for preventing overproduction. He proposes the use of time. To prevent overproduction or producing early, one should not make the material available early. Exactly how we determine the time when material should be released and the additional rules for managing flow are described in the following.

Production Operations and the Five Focusing Steps of TOC

In this section, we discuss the application of the core principles of TOC to production operations. As discussed in other chapters, the 5FS provide the rules for determining how any given operation should be managed. These steps (Goldratt, 1990b, Chapter 1) are listed below: Step 1: Identify (or choose) the system constraint.

Step 2: Decide how to exploit the system constraint.

Step 3: Subordinate all other decisions to the above.

If we desire to improve the performance of the system to a level higher than possible with the current constraint, then we must Step 4: Elevate the system constraint.

This step can change the constraint or the decisions on how to exploit the constraint. Hence, the need for Step 5.

Step 5: If, in Step 4, the constraint is broken, then go back to Step 1. Don't let inertia become the constraint.

Production is only a part of most manufacturing business organizations, that is, it is a subsystem. The true constraint of the business may or may not be in the production subsystem of the organization. If the constraint is chosen to be another subsystem or the market, then the role of production in the five-step process is under Step 3-Subordination. In this case, production should be managed by the rules of Simplified Drum-Buffer-Rope (S-DBR) discussed in Chapter 9.

The other possibility is that we choose the constraint to be in the production operation. More specifically, the capacity of a specific work center is chosen to be the constraint. By this choice, the company is making the statement that its business strategy is to make money by finding the best ways to exploit the available capacity at this work center. Clearly identifying the specific resource that will be the constraint (Step 1) and then finding the rules to exploit the capacity of this constrained resource (Step 2) are the key elements of the DBR system for managing production operations.

In this section, we present the DBR system for managing the flow of products in production operations. The scope of decisions that are involved in exploiting the constraint goes far beyond managing the flow of products. For example, the choice of which products to market has a significant impact on the total Throughput potential of the factory. An excellent discussion of this case (referred to in the TOC literature as the PQ example) can be found in The Haystack Syndrome (Goldratt, 1990a, Chapters 1113) and is also discussed in Chapter 13 of this book. For our purpose, it is assumed that we know what products are being sold and who the customers are. The challenge we are addressing is how best to manage the flow of products so we are able to satisfy this customer demand while keeping inventories and expenses to a minimum.

Characteristics of Production Operations

Every production operation is characterized by the following elements.

There Is a High Degree of Dependency

Dependency in this context means that certain operations or activities in the plant cannot take place until certain other operations or activities are completed. Some examples of dependency in a manufacturing operation are as follows: The routing sequence of required operations to manufacture a product is a simple example of manufacturing dependencies. In the typical case, the production process cannot begin until the required materials have been procured; individual operations cannot be performed until the prior operation specified in the routing has been performed; and the assembly operation cannot begin until all required components have been fabricated or purchased.

Another obvious example of dependency is the same resource being required to process more than one operation. These operations can be different steps in the routing of the same product (rough milling and finish milling, for example) or steps on different products (rough milling of Product A and rough milling of Product B). The possibility for creating blockage for one product when the resource is occupied with another product is obvious.

Other examples of dependency include: Resources cannot be set up until the setup person is finished with another job.

Work cannot begin until the setup or changeover is complete.

The first piece of a lot cannot be inspected and approved until the inspection gauges are calibrated.

The number of dependencies in even a small production operation is staggering.

Production Operations Are Subject to a High Degree of Variability

Variability exists in manufacturing operations in the form of both random events and statistical fluctuations. Random events are those activities that take place at irregular intervals, have no discernable pattern, and by nature are unpredictable. Examples of random events include: A significant customer order is suddenly cancelled.

A key vendor's plant is crippled by a strike and the critical materials are not readily available.

Tools, fixtures, gauges, etc., are suddenly unavailable due to unexpected breakage.

Statistical fluctuation or common cause variations in manufacturing environments refer to the fact that all processes have some degree of inherent variability. Examples of statistical fluctuations include: Receipt of materials from vendors can vary in quantity, quality, or timing from purchase order to purchase order.

Time to set up a resource varies each time the resource is set up.

Actual customer orders are different from the forecast.

Process yields may change from one lot to another.

We will use the term variability to describe both random events and statistical fluctuations.

The existence of these two phenomena-dependency and variability-combine to make the task of controlling the performance of manufacturing operations very difficult. In fact, the day-to-day role of a shop floor manager is nothing more than attempting to cope with the almost endless stream of disruptions and their impact on a wide range of activities.

At a single step in any process, it is not safe to assume that the effect of statistical fluctuations will average out and the performance of the process will be the average rated performance for that step. One of the dramatic effects of having both dependencies and fluctuations is that this averaging out does not occur. As discussed in detail in several other works (Goldratt and Cox, 1984; Srikanth and Umble, 1997; Schragenheim and Dettmer, 2001). "Disruptions/fluctuations will not average out for the total system and most individual resources will be forced to perform below their capability" (Srikanth and Umble, 1997, Vol. 1, Chapter 4).

Resource Capacities Are Unbalanced to Each Other and to the Market Demand

The ideal goal that every operation strives to achieve is that of a balanced capacity plant-every resource has just enough capacity to meet market demand. A major effort of most manufacturing operations is to manage the capacity that is available so that there is no wasted or excess capacity. In spite of this enormous effort, the perfectly balanced plant does not exist in reality. This is due to two factors. The first factor is that capacity comes in finite increments-resources must be purchased in whole units, labor must be hired for one full shift, etc. Thus, if we need 2.67 units of a particular resource we have to end up with 3 units.

The second factor that makes it impossible to have the ideal perfectly balanced plant is the combined effect of dependency and fluctuation. As discussed in the previous section, resources downstream will feel the impact of disruptions in upstream processes in a very biased fashion-they feel the impact of negative variations, but not those of the positive variations (see Srikanth and Umble, 1997, Vol. 1, Chapter 4). As a result, resources downstream will fall further and further behind, unless they have available capacity to catch up. If the plant were perfectly balanced, there would be no catch up capacity available and the plant would fall further and further behind. Without an appropriate amount of reserve capacity, the plant will be unable to operate effectively. As the plant falls behind schedule, managers will be forced to increase capacity (through overtime, hiring additional labor, etc.) at the resources that have the most delays. Thus, in the end, managers are forced to run unbalanced plants.

The total available capacity of a resource can be broken down, based on the previous discussion, into three categories: productive capacity, protective capacity, and excess capacity.

Productive capacity is defined as resource capacity that is required to produce a quantity of product sufficient to satisfy the agreed upon output of the system (Sullivan et al., 2007)4. Protective capacity is the resource capacity needed to protect the Throughput of the system by ensuring that some capacity (above the capacity required to support system Throughput) "is available to catch up when disruptions inevitably occur. Non-constraint resources need protective capacity to rebuild the bank in front of the constraint or capacity constrained resource (CCR) and / or on the shipping dock before Throughput is lost" (40). Excess capacity (22) is defined as resource capacity that is in excess of what is required to protect Throughput of the system. Protective and excess are also called idle as most of the time they are not used; protective engages when Murphy strikes to rebuild buffers.

It is a far better strategy to acknowledge that perfectly balanced plants are not attainable and are not even desirable. This means that most real life production operations are unbalanced and many resources will have idle (composed of protective and excess) capacity available. The availability of this idle capacity allows us to design a system under which the operation as a whole will perform at a higher level of reliability (less fluctuation) than individual operations.

Applying the Five Focusing Steps to Production Operations

We are now in a position to design a system that can operate at a very high degree of reliability while producing the highest levels of output possible. Since we do not have a balanced plant, it is clear that at least some resources will have more capacity than needed to meet market demand. In fact, in any dependent chain of resources there will be one resource that has the least capacity relative to demand. If the capacity of this resource is the same or less than the capacity required to meet market demand, then the resource is referred to as a bottleneck. The weakest bottleneck is the constraint of the system.

The rules one must use to get optimal performance from any system are derived under TOC through the application of the 5FS. The resulting approach is referred to as the DBR method of managing production operations. The application of the 5FS would proceed thusly.