A Closer Look at Variability

So let us take a more in-depth look at how variability affects an enterprise value-added chain. Figure 35-5 depicts workflow in an organization. The time to complete each task consists of set-up time plus work time plus set-down time plus resource queue time. In the case of a typical shop floor scheduling routing, this represents three different resources processing three individual tasks on a single part. In the case of a project network, this represents three different resources processing three tasks that are supporting a single part, or it could be tasks that simply receive information or results from the predecessor task in order to work on the successor task. The material required to support these tasks will be procured, placed in inventory until transported using an entirely different schedule. This is a very complex effort indeed.

The queue time or white noise is the time when no value-added or productive time is being realized. In other words, Fig. 35-5 is showing how the elapsed time is still being accumulated between the productive tasks while providing no value-added to the output of the organization. The accumulative effect of productive time plus queue is equal to the total cycle time. Therefore, it follows that for any organization to improve, any scheduling algorithm must be able to synchronize and leverage the availability of the resources in order to eliminate this excessive idle time to maximize the Throughput of the organization. This means that to increase Throughput, which was previously defined as sales minus TVC, the organization must accelerate the flow of work or, more precisely, the rate of work flow. That is, with a given amount of resources, the organization must be able to deliver the final product to the market sooner. This rate of workflow is the key to being more responsive to the customers and increasing the company's profits.

FIGURE 35-5 Elements of task time for a dependent series of tasks.

The company must find ways of reducing the cycle time of producing and delivering its products. The easiest and most effective way is by reducing the queue time (Fig. 35-5) of resources waiting to be utilized, the principle reason WIP starts to increase, which leads to resources being significantly more productive (Little's Law; Hopp and Spearman, 2000). There are different ways this can be accomplished but one fact is indisputable-every company is susceptible to variability; therefore, any successful solution must be able to better manage the uncertainties, changing of priorities, and schedule changes. This variability causes major impact to the work schedules, which become the single greatest contributor to the resource queue. There is a direct relationship between resource queue time and productivity. So if cycle time must be significantly reduced, any breakthrough scheduling algorithms must reduce variability when possible, thereby reducing queue times. This must be done while providing real-time information to managers to mitigate the increased risk of managing variability.

The uncertainty of resources, material availability, and required technical information, subjected to the effects of unforeseen common and special causes, can be expressed as variability. Lacking a deeper understanding of how the variability can be mitigated leads to many instances of companies actually using the wrong scheduling tool while attempting to better manage the variability. For example, they may be using only project management tools when this may not be the best scheduling algorithm because they view themselves as a "project management" company, when in reality different parts of the company may be subjected to different kinds or types of variability, which means more than one scheduling algorithms is required. Alternatively, they may only be using production planning and control scheduling tools because they view their company as a floor scheduling/manufacturing company. In addition, regardless of which tool they decide to use, in many cases they may decide to schedule and manage their material requirements by imbedding them in whatever work-scheduling tool they are using instead of using the appropriate material management algorithm.

Many of these mistakes can be traced to a lack of understanding of the origin and the cause and effect of the variability. In order to clarify this confusion, a further classification of variability is required. This will help companies decide which algorithm is appropriate and lead to more effective planning, scheduling, and management of their environment.

There are three different types of variability significantly affecting an organization.

Type 1-This occurs when most of the variability is within the task itself and not in the resource queue (see Fig. 35-5). The most significant known or anticipated variability will be in the work being performed in task A, B, and C. Remember in planning we are identifying what work must be done and the resources required within the tasks. This is not implying that in execution there will not be variability due to lack of resources, or set-up or set-down time. In fact, it is very highly likely that many of the tasks will be impacted by the required resources not being available. Conversely, some of the resources will spend time in the queue waiting for predecessor tasks to finish, thus allowing the successor task to start.

Type 2-This exists when the variability within the task itself is relatively low and most of the variability is in the queue. This assumes well-defined manufacturing processes and well-defined tasks. In Fig. 35-5, Tasks A, B, and C the variability is low because this particular work or something very similar has been done many times before. In companies using MRP or MRPII, the manufacturing routings are readily available and will be incorporated into the master schedule. The same can be said for the set up and set down; the required time is well known and variability is minimum.

Type 3-Occurs when the variability is in the demand pattern of material requirements. This can be within the company if the part is currently in inventory or is a component being manufactured internally. Sometimes the material is outsourced and must be delivered in time to support the company's master schedule. This is further complicated by having to anticipate future market demand for all products, which of course determines what material is needed, the quantity, and precisely when it must be available.

Different Tools for Different Types of Variability

If there are three types of variability, this leads to a requirement for separate and distinct algorithms for planning, scheduling, and execution. The three commonly used algorithms are as follows: Project Management-Type 1 variability, which relies heavily on the concept of critical path methodology and establishing well-defined relationships of the tasks. Once the tasks have been identified, the correct sequencing will yield the project network. This network becomes the schedule for managing the resources and executing the project. Again, in Fig. 35-5, the greatest uncertainty or variability is captured within the individual tasks. The project network schedule will not have any protection against variability in the resource queue. Typically, the amount of protection time for variability placed within the task is two or three times the actual productive time required.

Production Floor Scheduling-Type 2 variability, which relies on developing well-defined relationships of the tasks and identifying resources. This algorithm does not use the concept of critical path methodology. As shown in Fig. 35-6, Task A, B, and C and set up and set down have very little variability. This drives most of the variability into the resource queue. In fact, if one looks at the ratio between the times scheduled to accomplish all of the tasks in manufacturing of the individual product to the productive time (Fig. 35-5), which is the actual touch time needed, this confirms most of the time in the schedule is placed in the resource queue. It is not uncommon to schedule the manufacturing cycle time with 10, 20, or more times than the actual touch time (Schragenheim and Walsh, 2004).

FIGURE 35-6a Traditional project network with buffering within each task.

FIGURE 35-6b Critical chain project network with strategic time buffers.

Material Management and Inventory Control-Type 3 variability is managed by providing safety stock of specific physical parts and finished goods (stock buffers) to protect against the changes in forecasted demand patterns. This also requires scheduling and managing material that may be outsourced or is provided directly by multiple suppliers. Material requirements have to be carefully coordinated using tools like MRPII to support the company's schedules. In addition to receiving material from suppliers in order to manufacture the company's products, this also includes scheduling materials to and through the stocking points to their final destination through the distribution channels.

Regrettably, most companies feel that, in spite of their best efforts and willingness to implement a multitude of process improvement initiatives, they fall short of achieving the anticipated returns. The reason companies fail to achieve their objectives significantly is a lack of focusing on improving the entire system as depicted in Fig. 35-4. Rather, the tendency is to focus on improving individual functional areas of the company without truly understanding the net effect it will have on the profit or return on investment. Figure 35-4 shows the local variability experienced by the individual functional areas of the company (departments, work centers, etc.). The source of the variability may be caused by disruptions within the functional area or other functional areas, late deliveries from suppliers, or changing market demand patterns.

Defining the System

The first step in developing a systems approach to improving the company is building a combined work flow diagram of the design, production, and distribution and supporting networks. Starting at a high level (Fig. 35-4) will lead to additionally granular diagrams until you have defined the level of required detail. A word of caution-keep the work flow diagram at a fairly high level or you will get bogged down in needless detail. The diagram can be developed further with as much detail as needed when action plans are being developed. This macro-to-micro approach has proven helpful in analyzing and creating effective company systems architecture of the types of planning, scheduling, and control systems. At times, the different types of variability may appear not to be that clear cut; if so, I encourage you to make the effort to identify which type of variability is involved. This effort will give you a better understanding of what lies ahead. Perhaps it may be a hybrid environment where more than one algorithm must be implemented as part of a system.

The TOC Approach

Regardless of the source or cause of variability, it is far more important to know how it is impacting the company rather than just how it is impacting an individual functional area. Variability is the key indicator of how valid your assumptions are and how well the planning is being executed. In other words, if you are measuring this variability it will be an indicator of how effectively the planning and scheduling is deviating from what you thought was going to happen. However, in order to do this monitoring, there must be a common metric tying all of the individual functional areas to the company's Throughput. This metric is time. By using time as the overarching metric, it is now possible to evaluate if the individual functional areas throughout the company are staying within a predetermined acceptable time burn rate. It is now possible to see if the variability is consuming an unacceptable amount of time. This provides a process for evaluating the potential impact the variability in any part of the company will have on performance. TOC focuses on time management and ties this to the disruption this variability is causing within the schedule.

Visualize a time bank, referred to as a time buffer, providing additional time to individual functional areas if needed to protect the schedule from variability. Then the time buffers are placed strategically in the schedule, providing significant protection while protecting the delivery dates of your products or services being provided to clients. Once this connectivity is established, then it is possible to monitor the time buffers. This is called Buffer Management (BM). Buffer time and BM are new and important concepts first developed by Eliyahu Goldratt while developing TOC, and are the basic building blocks for the business strategies and solutions discussed next.

TOC recognizes the existence of interdependency and variability in all organizations; in fact, all of the TOC business solutions are firmly grounded in these tenets, providing tools to better leverage the organization's Throughput. The interdependencies of the different functional area resources and their corresponding statistical fluctuations, which are manifested as variability, are shown in Fig. 35-4. The three TOC business solution algorithms are as follows.

Project Management

Critical chain is the longest path recognizing task and resource dependency.6 Time buffers of aggregated safety are placed strategically throughout the project, providing much greater protection against variability for the critical chain than conventional critical path methodology. During project execution, monitoring the individual rate of buffer penetration against predetermined acceptable levels will provide real time risk management information. In most cases, this information will be provided early enough to allow for the required action to be taken before the promised delivery date is impacted.

Figure 35-6a is a project network where task A (using Red Resource) is scheduled to take 8 days to finish. The successor tasks when completed will feed into task D, the last task in the project. Traditional project management tools are typically used to schedule work in a Type 1 variability environment. Project management does not normally have a resource queue to provide protection against variability in the scheduling algorithm. Everyone knows that in execution, variability will cause many of the tasks to take longer than anticipated, so the common practice is to embed additional safety time within the task itself.

The TOC project management algorithm, Critical Chain Project Management (CCPM), removes the protection or safety time placed in the individual tasks and schedules only the known time. Then part of the total time removed is placed in the high-risk integration points as a feeding buffer throughout the project network. An additional portion of the removed safety time is placed after the last task as a project buffer. The critical chain is task A + task B + task D and when combined with the project buffer placed at the end of the last task we establish the duration time of the project. In essence, removing the safety time previously embedded in the individual tasks and strategically placing 50 percent in time buffers throughout the project provides much better protection from variability by aggregating the safety time at strategic points (see Fig. 35-6b). This buffer protection allows for establishing control limits and monitoring the rate of time penetration into the feeding and project buffers, providing valuable real-time information of precisely when and where variability is affecting the project. This is crucial for effectively prioritizing where the resources are used when you are resource limited. A more in-depth explanation of the critical chain solution can be obtained in the book Critical Chain (Goldratt, 1997) and in Section III of this Handbook.

FIGURE 35-7a Serial line showing product/service flow.

FIGURE 35-7b Serial line with a buffer inserted prior to the capacity-constrained resource.

Production Floor Scheduling

Drum-Buffer-Rope (DBR)7 provides buffer protection against variability at the most critical parts of the operation. Monitoring the buffer penetration will indicate when and where action must be taken, ensuring very high on-time deliveries. This scheduling algorithm is typically used in a Type 2 environment, where the task itself has low variability and there is a considerable resource queue. Therefore, the most fertile area for reducing cycle time is not in improving the time to perform the task but rather reducing the queue.

Figure 35-7a depicts a simple routing of tasks required to build a product in a manufacturing process in a Type 1 environment. The routing is built in isolation and is subsequently added to the master schedule, which is used for scheduling many other products. In conventional scheduling algorithms, the paradigm is one of loading the master schedule until every resource is fully utilized. However, the DBR approach schedules the constrained resource to no more than 85 to 90 percent (in simplified DBR), which provides a time buffer for protecting the constrained resource against variability. (See Chapter 9.) The capacity constrained resource (CCR)-a resource that if not managed effectively will become the constraint-in this case X, (see Fig. 35-7b) has less capacity than the other resources in this process. This means that the resource determines how much can be produced. Therefore, this scheduling less than constraint capacity also means that all of the other resources by definition have additional sprint (protective) capacity to respond whenever variability is causing disruptions. A CCR time buffer is placed in front of the constrained resource, which means the resources in front can start to work and deliver their output to the CCR before they are needed. This tying the rope from the CCR to the gating operation allows delaying release of the work order to the floor until a buffer time ahead of when needed by the CCR. It is common for WIP to accumulate in front of the CCR so that when variability impacts a resource, the CCR will be protected from the disruption. Whenever the disruption is resolved, the resources use their sprint capacity to catch up until the flow is back to normal. By monitoring the control limits of the buffers, management knows when and where to take action before the effects of the disruption impact delivery dates.

FIGURE 35-8 Typical material flow in a manufacturing operation.

This approach significantly reduces the WIP inventory, which reduces the resource queue, a prerequisite for reducing cycle time. It follows if we reduce the cycle time without hiring additional personnel, then we increase the company's Throughput.

Material Management and Inventory Control

TOC Replenishment8 is when stock levels are based on dynamic buffer stock levels that are much more agile and responsive to changing demand patterns than conventional min-max methodology.

The predominant conventional inventory control algorithm is based on determining the maximum amount of inventory carried for an item in the various stocking points in the company as depicted in Fig. 35-8. These stocking points can occur anywhere needed in the production flow as required to protect Throughput. This algorithm is based on determining at what quantity level you reorder (min), triggering an order to get back to the maximum inventory level. The min level is the minimum quantity level that triggers the reordering procedure to get back to the maximum level. The min level is based on the average demand during replenishment lead time and the amount of safety stock.

The TOC replenishment approach is based on dynamic stock buffers, which are based like all TOC algorithms on managing time. This causes the inventory levels to increase or decrease in real time based on the fluctuations of market demand. Now to be clear, the stock buffers are physical material for supporting the manufacturing operations or finished product in a make-to-stock environment. The greatest source of variability is due to ever-changing market requirements for the company's products. So it may not appear that the replenishment solution is managing time; therefore, an explanation is in order. The objective is to maintain inventory levels that provide materials in a timely manner to support the manufacturing schedules. Therefore, the focus is on ensuring that as customer requirements change, the inventory levels of the needed material will be available. This change in focus, unlike the min-max methodology, allows for more frequent ordering to replenish current demands and the continuously changing trends. This allows the carried inventory in a company to be aligned closely to the market needs with significantly reduced levels of inventory. An agile and responsive material management and inventory control solution is needed for supporting the internal critical chain and DBR schedules that are producing higher and accelerated Throughput levels of performance.

Throughput Accounting for All Methods

The TOC approach provides a common overarching metric, Throughput (T) dollars which are defined as sales (S) dollars minus Truly Variable Costs (TVC) dollars as a rate; that is, T in dollars per period of time. The significance is managers now have an unburdened, absolute, and real measurement that can be used across the organization. Every work center, department, and functional area has a common metric on which they can focus. This enables companies to make decisions focusing on what is best for increasing Throughput with individuals and every individual support function measured on their contribution to Throughput. As previously stated, the limiting factor to increasing Throughput is the company's constraint; therefore, it follows that all support functions must always make this metric their top priority. Now for the first time, every part of the company has the same common metric for measuring the flow of value added being generated. This also provides managers with the individual contribution to Throughput that each part of the company is generating.

Buffers for Time Management

The other critical contribution the TOC approach provides is the concept of time management. There are many very effective ways to manage operations within a business. Henry Ford used the concept of placing material on conveyor belts to control the flow. Dr. Ohno revolutionized the world of manufacturing by controlling the release of material and work performed as late as possible, thus reducing the queue, the key to improving Throughput. Dr. Goldratt (2009) decided a more effective way was managing time; this also is a way of reducing the queue providing the advantages pioneered by Dr. Ohno. However, it also provides a means for protecting against variability by strategically placing time buffers that will send a signal when to release material to be worked on. In Fig. 35-7b, the material is released earlier in time, a buffer time, to reach the CCR when needed as determined by the schedule. The buffer is divided into three regions: green (all is well), yellow (caution), and red (schedule is being jeopardized). The time buffers are an integral part of the TOC solution. Once the proper buffer levels are established, they also become the control limits. By monitoring the penetration of the buffers, it will indicate when and where variability is affecting the schedule allowing management to take action in a timely manner. In almost all cases, there is enough time to intervene without affecting delivery commitments. It is important to understand that buffer penetration indicates the system is experiencing disruption and monitoring and taking action when required is key to keeping the system in control.

A key factor then is the focus BM provides on the highest priority problems.

Since the three TOC scheduling and business solutions are focused on maximizing Throughput, providing risk management using strategically placed time buffers provides the basis for a powerful means of improving productivity across the enterprise. It appears that, using common metrics, we can now assess the impact any specific task is having on any part of the organization even though they may be using different scheduling algorithms and may even be in different functional areas.

Think of time buffers as aggregating a portion of the total required time and placing it in strategic parts of the schedule in order to provide significantly more effective protection. This is in stark contrast to conventional approaches that simply release material much earlier than needed to allow additional time to combat variability. It also addresses the challenge, when using Dr. Ohno's approach, of having to store physical inventory throughout the manufacturing process or improving every single process. This will drive the work or availability of services being provided to commence earlier in time. The buffers are also control limits; therefore, in execution we can monitor how much the time buffers are being expended, which is simply a reflection of how much variability is impacting the schedule. If we have previously determined the acceptable level of buffer burn rate, then it is easy to observe if our system is in control or if any action must be taken. If action must be taken, we will precisely see exactly where the action must be taken.

Applications

An example of a systems approach using TOC tools for managing the development of bio medical devices is depicted in Fig. 35-9.

A mid-size company developed new pharmaceutical and biomedical devices. Typically, the company takes a partially developed new product through the R&D phase, then laboratory testing, and then clinical trials. They would build a manufacturing plant to provide the product for testing. Every step in this very complex and exacting process must comply with the Federal Drug Administration (FDA) requirements and is subject to close oversight at every step. It is very expensive and can take years to obtain FDA approval.

Once FDA approval is achieved, this company would deliver the new product to one of the large multinational companies who in turn markets, mass produces, and sells the product. The benefit of reducing the new product development cycle is very significant. The large multinational companies fund the entire development effort at great expense; it could range from tens of millions to hundreds of millions of dollars. The company that brings the new product to market first will end up owning the market and will always be the predominant supplier. This is a very high stakes game indeed.

FIGURE 35-9 Integrated scheduling algorithm for a new product development.

The company has all three of the different types of variation present in their operations. As discussed earlier, it is crucial they recognize this and develop a synchronized solution set. In Fig. 35-9 the overall, or master schedule, is notionally presented as a 14-task critical chain project. This is Type 1 variability, where most of the variation is within the tasks themselves. The white "tasks" are not actually tasks, but rather are the aggregated time buffers that provide protection to the project when disturbances to the schedule happen. These buffers push the tasks to start earlier in time and since the safety time previously embedded in the tasks is removed, the duration of the project is significantly less while providing much greater protection.

The construction of the manufacturing plant is depicted in Fig. 35-9 as a subordinate critical chain project, which is also Type 1. The construction of the plant is synchronized to finish and be operational when required by a task on the master critical chain schedule. The company delayed starting construction of the plant by six months and it was completed and operational with time to spare. This allowed the company, in essence, to have an additional six months before the production line had to be baselined. The maturity of the product, due to having the results of six additional months of data, was such that zero changes were made to the schedule.

When the manufacturing plant became operational, it was now subject to Type 2 variation. The manufacturing lines, the processes, and the individual tasks had very little variability as required in order to obtain FDA approval. The scheduling algorithm used was Simplified Drum-Buffer-Rope (S-DBR; Schragenheim and Walsh, 2004), a version of DBR developed by Eli Schragenheim, (See Chapter 9) which released the raw material for manufacturing the product, delivering to the task on the master critical chain project in a timely manner. The manufacturing time was 50 percent less than what the company historically had taken on similar products.

The TOC replenishment algorithm was used to manage the material requirements for the entire company. This requirement is a Type 3 variation, subject to rapidly changing requirements of the product development cycle. The quantity of material being held in stock was significantly reduced, which made it easier to manage. The different sizes of the product being manufactured changed often; this is equivalent to the product mix changing often, and the replenishment solution allowed the plant to be visibly more responsive.

The next application of this approach, depicted in Fig. 35-10, is at the United States Marine Corps (USMC) Maintenance Center in Albany, GA. They are one of two maintenance repair and overhaul (MRO) activities that service all of the USMC tracked vehicles.

The vehicles are returned to be serviced after many years of use in the field, much of which has been in very demanding environments. The mission is to return the vehicles to almost new condition, as quickly as possible and at the lowest cost. In addition, many upgrades are designed, manufactured, and concurrently installed as part of the total effort. In addition, the condition of the vehicles is unknown until inspected. Furthermore, the demand pattern is unpredictable, which only adds to the uncertainty in an already complex scheduling environment.

The vehicle itself is scheduled as depicted in Fig. 35-10 as a nominal 14-task critical chain project because it experiences Type 1 variation. The actual number of tasks is much greater because it covers the major events, such as inspection, disassembly (in many cases, only the nameplate will remain on the production line), assembly, corrosion, paint, testing, etc.

FIGURE 35-10 The integrated scheduling algorithm in an MRO environment.

Some of the major components removed for repair and overhaul such as the engine are scheduled using a subordinate critical chain schedule since again this is a Type 1 environment. The many other components that are removed from the vehicle are sent to the support shop and scheduled using DBR since this a Type 2 environment.

The TOC replenishment algorithm is used to schedule and manage the shop consumable items and replacement parts. This is extremely challenging in a complex, high-mix, constantly changing MRO environment. This is an extremely demanding Type 3 environment of high and low product volume demand that changes on a daily basis.

All of the work has to be synchronized and come together at final assembly. This is only possible by scheduling backward from the vehicle delivery date and subordinating all efforts to the needs of the master critical chain. The uncertainty encountered in the Maintenance Center at Albany, GA is much greater than a repetitive manufacturing environment such as when the vehicle is originally produced.