The general rule that emerges from this example is the following. In trying to determine whether intervention is required, we are comparing two time periods. The first time period, time available, is the amount of time that is actually available to finish the batch on time. This is the time from Today/Now to the Date/Time when the batch is due. The second time period, planning or standard production lead time, is the amount of time that is required to complete the batch. As the ratio of available time to planning production lead time becomes smaller (this will happen naturally with the passage of time), the degree of certainty that the batch will finish on time will diminish. We refer to the ratio of available time to standard production lead time, expressed as a percentage, as the buffer status of the batch.

Buffer Status (%)= (Available Time)/(Standard Production Lead Time) 100% Figure 8-10 shows the buffer status in the form most frequently used, which is by assigning each work order a color based on the buffer status. If the time remaining for an open work order is less than one-third of its standard production lead time, then the buffer status is less than 33 percent. (If the batch was released on time, then we have less than one-third of the standard lead time available to complete the batch on time.) Such a batch should be flagged. Production personnel will have to investigate where the batch is currently located and determine whether corrective action is necessary. The rule that every batch whose buffer status is smaller than 33 percent should be flagged for investigation is an empirical rule based on experience. If the point at which a warning signal is issued is too lax (buffer status of, say, 50 percent), then we are likely to receive too many warning signals, creating unnecessary work. Conversely, if the point at which the warning signal is issued is too tight, then very few warnings are issued and, more seriously, there may not be sufficient time to react. Due to the fact that the actual touch times (process + setup) are typically 10 percent or less of the actual production lead time, if we know of serious problems 10 days before the batch is due (for a batch with a standard or planned lead time of 30 days), we should be able to expedite the product to finish on time. In the next section, we discuss how this works in practice, in a production operation that has hundreds of batches in process at any given time.

FIGURE 8-10 Designation of buffer status by a color. Comparison of time remaining (Due Date of Order to Today) to the planned buffer time to assign color to a work order. Status and Action: Red-Time remaining less than one-third of the buffer. Expedite. Yellow-Time remaining between one-third and two-thirds of the buffer. Monitor and plan. Green-Time remaining greater than two-thirds of the buffer. Do nothing.)

Buffer Management-The Process

In accordance with the previous discussion, each open work order or production batch will have a buffer status that can be calculated. Note that the buffer status does not depend on where in the process a specific work order/batch is located. Based on the buffer status, the work orders are color-coded into three different categories.

Green Work Orders: A work order is assigned the color green when the buffer status is greater than 67 percent. For a green work order, there is plenty of time still available to complete the order. No matter where in the production process this order happens to be, there is no cause for concern and it is reasonable to expect that the order will in fact be finished on time.

Yellow Work Orders: A work order is assigned the color yellow when the buffer status is between 33 and 67 percent. For a yellow work order, disruptions have eaten into the normal flow and there is a risk that additional disruptions might make these orders late. However, for now there is no need for intervention.

Red Work Orders: A work order is assigned the color red when the buffer status is less than 33 percent. The time left for finishing this order on time is small (relative to what we would like to have as expressed by the standard lead time). It makes sense to see where in the process this order is located. If the order is near completion, no intervention may be necessary. If the order is still in the early stages of processing (or even waiting for material release), intervention is required to mitigate the risk of a late order.

Each work order thus has a color code assigned to it based on the buffer status at that point in time. As time evolves, the buffer status may change. At the beginning of the shift, production managers should construct the list of work orders that are red at that control point. Each of these orders should be investigated to determine if corrective action is called for. Then responsibility for the corrective action should be assigned. The next day, actions should be reviewed to make sure they were done and the new list of red orders should be investigated. In fact, the primary activity of the daily production meeting is the BM process.

The assignment of colors gives us an opportunity to refine the priority system inherent in the DBR process. The simple FIFO rule can be modified as follows: Red orders first, then yellow orders, and then green orders. If a work center is working on a yellow order and a red order arrives at the work center, it is sufficient that the red order moves to the head of the queue and be processed immediately after the order currently being processed.

Another feature of assigning color codes to work orders is that they provide information about the adequacy of the production lead times that have been established. If the number of red orders is too low, then it is a clear indication that the production lead time is larger than it needs to be. The allowed time is large enough that few, if any, orders are experiencing disruptions that have any consequence. This suggests an opportunity to reduce the planning production lead time used in the DBR system and eventually to reduce the lead time quoted to customers. Conversely, if the number of red orders is large, then it suggests that a large number of orders are experiencing significant disruptions relative to the time allowed for their processing. In this case, the production lead times are too aggressive and need to be increased. It has been our experience that the total number of orders in the red should be around 10 percent. If the percentage of red orders exceeds 15 percent, then we should consider increasing the size of the buffer. If the percent of red orders is less than 5 percent, then we should consider reducing the size of the time buffers.

Complex Production Environments and a Classification Scheme

Real-life production environments are much more complex than the simple flows used in explaining the DBR system. Even a medium-sized factory has hundreds, and often thousands, of parts and products and has tens, and often hundreds, of different resources. In other words, the detail complexity of production operations is immense. When the focus is on the detail complexity, one is overwhelmed and tends to believe that each operation is unique and there is little that can be transferred in learning from one operation to the next. Typically, production is lumped with the business as a whole and is described in terms of the industry segment to which they belong, such as an auto plant or food and beverage plant. In this section, we present a scheme to organize the production operations based on their product flow characteristics. This classification scheme will bring together elements of detail complexity and dynamic complexity that have an impact on managing the production operations effectively. By managing effectively, we mean deliver the products as promised to customers, while keeping investments in resources and inventory to a minimum. The development of the classification begins with a change of perspective, from a view centered on resources to a view that is centered on product flow.

The Fundamental Elements of the Classification Scheme

Since we are used to the resource centric view of production, we have to learn how to view the same operation in a way that focuses on flow. Such a view is provided by the view of production represented in Fig. 8-2 and Fig. 8-3-the view of operations from the point of view of the materials. It is a time-oriented description of the manufacturing process. As indicated earlier, the resulting diagram of the production operation is referred to as a PFD.12 We now explore PFDs in more detail.

Consider the simple case where we have three different raw materials (RM-A, RM-B, RM-C) that are fabricated into three component parts (A, B, C) and that these are assembled together into a finished Product D. This simple production operation has only one finished product.

To construct the PFD, we begin with raw material A (RM-A) at the bottom left-hand side of the diagram (Fig. 8-11). Each step in the fabrication of the component part A is represented by a box vertically above the box for RM-A. If the fabrication process consists of four steps (this information is typically contained in the routing file or process sheet for component A in the company's ERP system), then we have a series of four boxes in a vertical line as shown in Fig. 8-11. For clarity, inside the box we have designated the process step and the resource used in that step-again a piece of information found in the routing file. The first step is designated A-010 and is performed by resource R1. The second step is A-020 performed by R2, and so on. Similarly, the fabrication process for component B made from RM-B consists of three steps and is represented by the series of three boxes B-010, B-020, and B-030. Finally, component C made from RM-C requires four process steps and is designated by the boxes C-010, C-020, C-030, and C-040.

FIGURE 8-11 A detailed product flow diagram for an assembled product.

In just the part of the PFD that we have constructed thus far, two characteristics of production operations (characteristics that make it difficult to manage these operations) stand out. One factor inherent in the PFD is the dependency of operation B-020, for example, on operation B-010. This type of dependency is referred to as material dependency. Simply stated, B-020 cannot be performed unless B-010 has been completed. Every stage in a PFD depends on the preceding stage. If a box in a PFD has an incoming arrow, this indicates material dependency. The material from the box at the base of the arrow is an absolute requirement for the box at the tip of the arrow. The boxes RM-A, RM-B, and RM-C have no incoming arrows as they are the beginning of this production operation. If we were looking at the entire supply chain, then clearly these boxes would be linked to the suppliers of these materials.

A second form of dependency that is highlighted in a PFD is between steps A-010, B-010, and C-010. All of these processes require the same resource, R1. This is an example of the type of dependency referred to as resource dependency. If R1 is engaged in step A-010 and there is only one resource R1, then B-010 and C-010 cannot be performed. Another resource dependency can be seen between stages C-020 and C-040. Both require the same resource R2. In addition, R2 will have to complete C-020 and R3 complete C-030 before C-040 can be started.

In Fig. 8-11, we complete the PFD for this simple operation by adding the assembly operation. An assembly operation, by its very nature, requires more than one input material. Just as the arrow from RM-A to A-010 represents the fact that RM-A is an input to the processing step A-010, the arrows from A-040, B-030, and C-040 all converging on box D-010 indicate that all of the components A, B, and C are required to perform this assembly step. If even one of them is missing, the assembly operation cannot proceed. In Fig. 8-11, the arrow from PP1 to D-010 represents the fact that a purchased Product PP1 is required (in addition to parts A, B, and C) to perform operation D-010. We refer to assembly stages as convergence points in the PFD-multiple products/materials are assembled together to make a single product. A convergence point (a control point) represents a high degree of dependency since all materials represented at the base of the multiple arrows are necessary for this operation to be performed.

FIGURE 8-12 Product flow diagram illustrating a divergence point.

In addition to the linear and converging flows, there are cases where the flow shows a divergence. Just as convergence is characterized by the coming together of multiple materials into a single product or component, divergence (a control point) is characterized by a single material being transformed into several different output materials. Consider, for example, a case in the textile industry. Figure 8-12 shows the case of a specific type of yarn being processed at the next stage-the dye house. At the dye house, color is applied to this yarn. We know that for the same yarn, different colors can be applied (red, blue, green, etc.) and we also know that red yarn is a distinct and different product than blue yarn. In the language of the PFD, the dye house is a divergence point-the same input material (untreated yarn) can leave the dye house as any one of a multitude of colored yarns. The divergence point at the dye house shows up in the PFD as a single yarn diverging at the dye house into a multitude of different boxes.

Material dependency, resource dependency, convergence points, and divergence points are the fundamental elements of a PFD. As discussed in the next section, production operations can be classified into families based on which element is the dominant element in the PFD of that particular operation. If divergence is the dominant element, then we have a V-plant. If convergence is the dominant element, then we have an A-plant. If both divergence and convergence exist (and exist at the same stage), then we have a T-plant. If we have neither divergence nor convergence, then we have a simple case of resource contention and the plants are classified as I-plants.

V, A, T, and I Flows-Descriptions and Examples

V-Plants

V-plants are dominated by the presence of divergence points throughout the product flow. The PFD for a plant that exhibits divergence at every step is shown in Fig. 8-13. Notice that this diagram resembles the letter V; hence, the name V-plant. In addition, in most real life V-plants, the different products share common resources at most stages in the process. A steel rolling mill provides a good example of a V-plant. The first step in the process is annealing where the sheets of steel are softened in preparation for rolling. At the rolling operation, a given piece of steel can be rolled into any of a large number of different thicknesses. Rolling represents a divergence point. At each divergence point, the number of distinct products increases. For example, after rolling each of the different thicknesses, steels can be heat-treated to many different products with different strength and hardness characteristics (based on the manner of heat-treating). Each of these steels, now with unique thickness and mechanical properties, can be cut into desired widths at the slitting operation. From just a few varieties of steel coils at the start of the operation, one can end up with thousands of finished products-characterized by thickness, mechanical properties, width, and length.

FIGURE 8-13 Product flow diagram illustrating a typical V-plant.

The existence of divergence points gives rise to three primary characteristics of a V-plant regardless of the specific industry or materials.

1. The number of end items is large compared to the number of raw materials. Because divergence points exist throughout the different stages of production, by the time several stages are completed, the number of different products can be very large as can be seen in the rolling mill example.

2. All end items are produced in essentially the same way. All products are processed through the same basic operations-rolling, heat-treating, slitting, etc.

3. The equipment is generally capital-intensive and highly specialized. The evolution into capital-intensive equipment is not difficult to understand. Since every product goes through the same sequence of operations, there are a relatively small number of basic operations performed repeatedly. Because the focus of improvement under the traditional cost-based system is to reduce the product's direct labor content, the equipment naturally became specialized, high-volume, capital equipment.

The one characteristic that all V-plants share is that despite having high levels of finished goods inventory, there is constant scrambling to meet customer requirements. The capital-intensive nature of the equipment, which typically comes with lengthy setup times and the presence of divergence points, is at the heart of this problem. The lengthy setup times encourage supervisors to increase batch sizes, to minimize setups by combining batches whenever possible, and to produce families of products together. All of these actions, which are consistent with cost-world thinking,13 result in a mismatch between customer required priorities and production priorities. In addition, the large production batches cause the production lead times to increase. The result of all of these actions is that lead times are long and unpredictable and this ultimately leads to missed due dates.

FIGURE 8-14 The WIP profile of each resource (in hours of work for that resource) for a V-plant.

V-plants typically face the following concerns: 1. Finished goods inventory is large.

2. Customer service is poor.

3. Manufacturing managers complain about constantly changing demand.

4. Sales and marketing managers complain about the lack of responsiveness from manufacturing.

5. Interdepartmental conflicts are common within the manufacturing area.

DBR in V-Plants

It is important to recognize first that in almost all cases, there is a considerable effort underway to address the problems faced by a typical V-type plant. Each of the issues is assigned a cause and a solution is either being designed or in implementation. However, the problems persist in most cases. A properly implemented DBR solution will address many of the root cause issues that underlie the V-plant problems and thereby help mitigate most of these problems at the same time. If a capacity constraint exists and these are the only conditions in which the full DBR system would be considered, then identifying which resource is the capacity constraint is the first task. In V-type plants, this is a simple task. Since the resources are involved in the flow of most products, material naturally accumulates in front of the resource with the highest load. The CCR is thus the resource with the largest in-process queue (measured in hours of work for that type of resource). In the case shown in Fig. 8-14, the constraint is resource R3. It is also true that the personnel in the plant have a common and usually correct knowledge of the constraint. As an aside, it should be noted that the presence of high levels of Finished Goods (usually the largest bank of inventory in the flow) suggests that setups are considered large at many key resources in the operation.

The next key step is to establish the drum. The challenge in most V-plants is the fact that the load placed on a specific resource is significantly influenced by the number of setups that result from this mix. In other words, changing the product mix can change the resource that is most loaded. For example, a textile mill running very large batches of a given color can significantly reduce the total load at the dyeing resources but can cause major problems at the cutting and sewing operations. This is because a single color material will have to be fabricated into apparel of many different sizes and styles and this causes overloads at these work centers. The key to establishing the drum is to find the proper balance between the market demand and the schedule at the constraint that satisfies the requirements for a drum: 1. It satisfies market demand.

2. It maximizes Throughput for the system.

3. It does not create new constraints.

The other factor in designing and implementing a DBR system in a V-type plant that needs special attention is the existence of a large number of divergence points. Each divergence point is a schedule control point and needs to be managed as such. Detailed lists that show the different products that need to be produced and the exact quantity that needs to be produced for each product are required at each divergent point resource.

The schedule control points are material release, constraint(s), divergence points and shipping.

A-Plants

A-plants are characterized by the existence of convergence points wherein a large number of component materials are assembled together into a few end items. The component parts are usually made up of parts that are fabricated in the plant (or other plants/departments in the division) and parts that are purchased from outside vendors. The typical PFD for an A-plant is shown in Fig. 8-15. One characteristic of A-plants, which is different from characteristics of T-plants, is the fact that the component parts tend to be unique to a single end item. Several levels of subassemblies may be involved prior to final assembly. Since the overall product flow is convergent rather than divergent, the product flow diagram resembles an inverted V, thus the designation A-plant.

FIGURE 8-15 Product flow diagram for a typical A-plant.

One example of an A-plant is provided by aircraft manufacturing. The PFD contains several thousand components that converge to a single product. The components that arrive at the final assembly plant are themselves major assemblies (jet engines, for example). In addition, the number of distinct aircraft types is quite limited-for example, Boeing has fewer than 10 active models.

The general characteristics shared by A-plants include: 1. Assembly of a large number of manufactured and purchased parts into a relatively small number of end items. Each assembly point represents a decrease in the number of distinct parts and after just a few assembly steps, the number of distinct items drops dramatically.

2. The component parts are unique to specific end items. This is a key feature that distinguishes A-plants from T-plants. Consider aircraft, for example. While every aircraft has engines, the engine for each type of aircraft is unique. The engine for a Boeing 747 is completely different from the engine for a Boeing 777.

3. The production routings for the component parts are highly dissimilar. In the example of the aircraft, the routing for the manufacture of a jet engine blade is nothing like the routing for the manufacture of the compression chamber.

4. The resources and tools used in the manufacturing process tend to be general purpose. In an A-plant, the same resources are used to produce many different parts. Resources are quite flexible, in contrast to the highly specialized equipment in V-plants.

Since the major focus in traditional manufacturing environments is resource utilization and not product flow, it is not surprising that the flow through fabrication and into the finished components is erratic. In fact, the flow through all areas of an A-plant is wave-like, resulting in what is characterized as "feast or famine." This wave-like flow means that it is highly unlikely that all of the component parts are available when needed at assembly. The missing parts must be tracked down and expedited to assembly. The feast or famine syndrome also creates the perception that bottlenecks "wander."

The major concerns in an A-plant include: 1. Assembly is constantly complaining of shortages and expediting is a way of life in manufacturing and purchasing.

2. Unplanned overtime is excessive. Resources that were idle during the week suddenly find a wave of material that is needed urgently at assembly in their queue and this results in overtime.

3. Resource utilization is unsatisfactory.

4. Production bottlenecks appear to wander about the plant.

5. The entire operation appears to be out of control.

DBR in A-Plants

Unlike V-plants where the identification of the constraint is straightforward, the identification of the real capacity constraint is not straightforward. This is a direct result of the possibility of product flows for different component parts being different. This can create a situation in which multiple constraints appear to be present. In addition, the use of large production batches (chosen to reduce unit costs and improve resource efficiency) results in wave-like flow and the constraint appears to wander from one resource to the next. At first sight, it might appear that the resource load information from the computer planning systems would provide a simple means of identifying the constraint, especially since most of these plants have a computerized planning and control system. However, the data are unreliable to the extent that in the author's experience resource load data from the computer system is highly suspect. As explained in Srikanth and Umble (1997), the best way to identify the constraint is to analyze resources that use the most overtime on a regular basis with the parts shortage information (the daily shortage list from assembly). The resource that uses overtime and processes parts regularly on the shortage list must be the constraint.

Two key factors should be considered in setting up the drum in an A-plant. The first is that the assembly (convergent point) operation provides an excellent place to establish the drum. Subordinating everything else to a well-constructed assembly schedule is the easiest way to achieve a good smooth flow through the entire operation. The assembly schedule should be established in such a way as to: 1. Meet market commitments.

2. Be within the constraint's capabilities.

3. Achieve a smooth flow through all of the operation.

The second factor is that the batch sizes that are being used are often too large and should be significantly reduced. Small batches are key to achieving a smooth flow and they should be aggressively reduced. Keep in mind that a batch is too small only when it creates a capacity constraint due to the increased number of setups that might be caused.

The schedule control points are material release, assembly, shipping, and the physical resource constraint (if one exists).

T-Plants