CHAPTER 11

Collocate

Job Shop Versus Work Cell Layouts

One tactic to reduce transportation time is to collocate sequential steps of a process. We will discuss this idea in the context of the traditional manufacturing setting, but later in the chapter, we will also describe how this idea can be applied to administrative processes. Although transportation time is the most apparent type of waste that collocation addresses, our discussion will identify other types of waste that are also reduced by collocation.

One way to arrange a factory is called a job shop layout, where similar pieces of equipment are located in their own department (see Figure 11.1). For example, all the drill presses that drill, tap, and bore holes are located in the same area, all the milling machines are in the same area, and all the heat-treat ovens are next to one another. Equipment in this type of factory is usually very flexible, which means that each piece of equipment can perform a variety of operations depending on how it is configured. For example, a drill press, with which some readers may be familiar, can drill a range of hole diameters depending on which drill bit is used. A fixture to hold a work unit in place might be required to ensure that the hole is drilled in the proper location at the proper angle. Fixtures are often suited to a particular part because each part or product is shaped differently. Setting up a drill press for a particular operation on a particular part therefore would entail installing the correct drill bit and attaching the correct fixture. The act of preparing the drill press for this operation would be called a setup or a changeover. Similarly, other general purpose, flexible equipment can be set up differently depending on the specific operations required for different parts: These machines include lathes, horizontal milling machines, vertical milling machines, and heat-treat operations.

Figure 11.1 Job shop layout

The advantage of job shops is that the general-purpose equipment can be set up for many different operations and is therefore flexible to permit a wide variety of operations that might be required for a wide variety of parts produced. Further, any particular sequence of operations required to make a particular part can be accommodated by routing the part to the appropriate sequence of production departments. For example, Figure 11.1 shows the paths of two different parts, which vary due to their different processing needs. Job shops might make hundreds or thousands of parts or products and overlaying the part paths generated by all those processing paths, similar to Figure 11.1 but with many more lines, would look like an incoherent bowl full of spaghetti. This is the motivation for a term often used to describe the flow of work through a job shop: jumbled flow.

One downside of job shop layouts is that transportation is required to move parts from one department to another. The necessity of transportation motivates parts to be moved in batches: As long as a material handler is moving goods from one department to another, it makes sense to move more than one part at a time. If parts were not moved in these batches, then additional resources (people, forklift trucks, etc.) would be required and, at some point, would cause the profitability of the operation to suffer. Working on items in batches, in turn, causes excess inventory and additional lead time. Rather than proceeding directly to the next step after being worked on, most items in the batch must wait for the remaining items in the batch to be completed, thus causing waiting time as items await their turn for processing at the next processing step. Our knowledge of Little’s law (Chapter 2) tells us that this additional waiting time must always be accompanied by additional in-process inventory. Conversely, observing the work-in-process waiting at various machines would allow us to infer from Little’s law that batches must increase lead time. One further disadvantage of batch production is that if one step in the process starts to produce defective items, then many more defective items will be made before the defect is found compared to when the downstream processing steps were executed immediately and the defect was observed at the subsequent step. This increases waste due to defects, rework, and repair.

An approach to remedy these shortcomings is to construct manufacturing cells where the processing steps are located next to one another on the factory floor or, in other words, the process steps are collocated. In some cases, this means that heavy equipment must be moved from the department where similar machinery resides (we will call this the home department of the equipment in question) to another part of the factory. In other cases, only small hand tools might be required along with simple workbenches, and relocating equipment or purchasing additional equipment in this case is less of an issue. People, too, must be assigned to these cells from other parts of the factory. This layout, where the sequential processing steps are collocated, is also called a product layout because the equipment is laid out according to the products’ processing needs rather than by the functionality of the equipment.

With collocated process operations, we no longer need material handlers to move goods from one step to the next. Operators can simply pass the item to the next step themselves, perhaps by pushing it along a roller table. The motivation for batch production is thus eliminated: Each station can process one item at a time and immediately pass it along to the next station. Waste due to transportation is thus significantly decreased and, with batch sizes of one, inventory and lead time are reduced.

Another advantage of cells is that they enable process steps to be coordinated and synchronized. With the process steps for any product being widely dispersed in the job shop layout, it is difficult to know which process step might be the momentary bottleneck or to know the progress status of the items in a customer’s orders. At best, one needs an effective information technology system to determine status and, at worst, the entire factory would need to be scoured to determine the status of an order. With a manufacturing cell, or work cell, the entire process and all the work-in-process can be viewed in one glance. This allows a worker or a manager to see which step may be holding up the process at a given time by just looking for the inventory backup, which visually signals the need for remedial action, such as repairing a machine or replacing defective material. Additionally, if the amount of inventory between the stations is limited, then the steps of the process are better synchronized—that is, they are working on the same product at the same time, at the same rate, with little inventory buildup between. Limiting in-process inventory in this way also reduces process lead time, as we could infer from Little’s law.

General Criteria for Work Cell Design

The primary difficulty in implementing a work cell is ensuring the proper utilization of equipment and people in the cell and in the remainder of the plant, with the remainder of the plant meaning the home departments where the equipment in the cell formerly resided. Specifically, the equipment and people in the work cell should not be overloaded with work. However, they should not be substantially underutilized because, as we shall see, this can cause the equipment and people in the home departments to be overloaded.

Utilization is most simply defined as the percentage of available time that equipment or people are used for productive activities.¹ Available time might be defined as all the hours that the production workforce is present at the factory. In that case, if a factory operated on one shift, the shift was scheduled for eight hours each day, and 240 working days were scheduled in each year, then 1,920 hours of production would be available each year:

If the parts produced over a year caused a particular machine to be used 1,700 hours, then its utilization would be

This calculation could be made over a month, a year, or any period as long as the quantities in the numerator and denominator corresponded to the same time period.

Assume that the machine referred to in this example is a drill press and it was moved from the drill press department to a work cell. Further, assume that there were six drill presses in the drill press department before one drill press was moved to the work cell. Also assume that 11,100 hours of drill press time are required per year by a company’s orders. With all the drill presses located in the home department, the utilization of the drill press department is

Once one drill press is removed from its home department and moved to a work cell, the utilization of the machines in the home department is

Thus because the utilization of the drill press in the work cell (88.5 percent) would be less than the utilization if the machine had remained in its home department (96.4 percent), the utilization in the home department after reassigning the drill press would increase to 97.9 percent in this case. In the worst case, it is possible that the utilization in the home department could go above 100 percent, which would indicate that the five drill presses do not provide sufficient machining time to complete the required work. In such a case, it probably means that the work cell is infeasible, another piece of equipment must be purchased, overtime work is required, work must be outsourced, or the capacity discrepancy must otherwise be resolved. Table 11.1 illustrates that the utilization in the home department can go either up or down when a machine is relocated from a drill press department with six machines to a work cell depending on how much the machine would be used in the work cell.

Table 11.1 Effect of cell on home department machine utilization

Scenario	Total plant drill press hours required per year	Drill press hours required for work cell	Home department utilization before work cell	Home department utilization after reassigning one drill press	Drill press utilization in work cell
1	11,100	2,000	96.4%	94.8%	104.2%
2	11,100	1,800	96.4%	96.9%	93.8%
3	11,100	1,639	96.4%	98.6%	85.4%
4	11,100	1,400	96.4%	99.5%	80.7%
5	11,100	1,200	96.4%	101.0%	72.9%

We have thus identified these considerations for designing a work cell:

1.The products manufactured in the cell must have a common set of sequential process steps.

2.The work cell capacity utilization cannot be unacceptably high, and it should be sufficiently high to justify creating the cell.

3.Relocating machines from the home department to a work cell must not increase the utilization of the home department machines to unacceptably high levels.

Some phrases in the previous points are uncomfortably vague—for example, sufficiently high and unacceptably high levels. This is because universal rules of thumb are difficult to specify here, and precise quantification depends on many factors, some of which we will mention later in this chapter. However, capacity utilization above 100 percent qualifies as unacceptably high: People or machines cannot operate for more hours than are available. (Note that while overtime production is possible, if one were to include those hours as available, then the previous statement holds.)

These criteria must be considered when constructing a cell, which requires these tasks:

1.Identifying a group of products that share a common sequence of process steps;

2.Specifying how many machines (alternatively, other resources such as people) to relocate from the home departments to the work cell; and

3.Completing previously listed items while maintaining acceptable utilization rates in the cell and in the home departments.

To illustrate the process of designing a work cell, we will use the data for the operation described in the example titled Job Shop Inc.

An Example: Designing a Work Cell at Job Shop Inc.

The first step in designing a work cell is to determine a set of parts that share a common sequence of operations. In the data provided for Job Shop Inc., this can be done visually by finding paths for different parts, each of which requires the same sequence of process steps. The visual depiction of part paths for Job Shop Inc. is jumbled, except for three parts, parts 1, 2, and 3 that share a common sequence of process operations: M4, L1, and D2. Note that the three process steps identified do not compose the entire processing for parts 1, 2, and 3. Although a work cell might perform all the processing for all the parts it manufactures, it is not required to do so.

Having established candidate products for a cell, we now evaluate the utilization of resources within that cell if it contained one of each of the required machines: one milling machine setup for operation M4, one lathe setup for operation L1, and one drill press setup for operation D2. (We will also refer to these machines as workstations.) First, we need to determine the rate at which the cell would produce each part. The pace of production for each part is determined by the machine with the slowest cycle time, and so when the cell is producing product 1, product 2, and product 3, it will complete a part every 15 minutes, 18 minutes, and 27 minutes, respectively.

Based on the annual volumes of these parts, we can then calculate that 145,200 minutes, or 2,420 hours, of production time are required in the work cell each year.

The utilization is thus

Clearly, producing all three parts in the work cell would overload the cell: There are not enough working hours in the year.

We can determine why the cell is overloaded by looking at the required utilization of each machine as if each machine in the cell was operating independently. Based on the cycle times at each machine, we would find that workstations M4, L1, and D2 have utilizations of approximately 97 percent, 122 percent, and 75 percent, respectively. For example, the calculation for workstation M4 is

The utilizations of workstations L1 and D2 are calculated similarly. The utilization statistics clearly indicate that workstation L1 is the most significant contributor to the cell’s utilization being greater than 100 percent. More specifically, the issue seems to be product 3, which takes a long time to produce on operation L1. The operation on L1 for product 3 takes significantly more time than on other operations at other workstations for product 3. Note also that the differences in the cycle times across the workstations for product 3 cause other operations to have idle time while L1 completes its operation. Attempts to redesign the cell to resolve this situation and, particularly, the overload on L1 and the imbalance among the operation cycle times might include the following:

1.Relocating more machinery to the cell—in this case another lathe to perform operation L1;

2.Removing product 3 from the cell;

3.Producing some of the requirements for products 1, 2, and 3 in the cell and some in the home departments using the traditional job shop routing; and

4.Increasing the number of hours available for the cell to operate, either through a second shift or overtime.

We will investigate the first two options next.

If we relieve the overload on operation L1 in the cell by relocating another lathe to the cell and setting it up to perform operation L1, then the cycle times for all products on workstation L1 are cut in half: Twice as many units can be produced in the same time. The workstation utilization results are shown in Table 11.6. We also need to compute the utilization of the machines in the home department by subtracting from their workload the number of hours of production performed by each workstation in the work cell as shown in Table 11.7. For example, the utilization of the machines that remain in the lathe department if two lathes were put in the work cell is computed as follows:

In this case, while relocating a second lathe from the home department to the cell resolves the overutilization problem in the cell, it causes an overload in the home department. Thus it is not a feasible work cell design unless another lathe is purchased.

Table 11.6 Work cell utilization with two lathes

Operation	Number of machines	Annual hours of work at average demand	Utilization
M4	1	1,870	97.4%
L1	2	2,347	61.1%
D2	1	1,448	75.4%

Table 11.7 Home department utilization with two lathes in work cell

Department	Number of machines	Annual hours of work at average demand	Utilization
Milling	5	8,130	84.7%
Lathes	5	10,153	105.8%
Drilling	3	5,752	99.9%

Table 11.8 Work cell utilization without product 3

Operation	Number of machines	Annual hours of work at average demand	Utilization
M4	1	935	48.7%
L1	1	862	44.9%
D2	1	898	46.8%

Negotiating the requirements of satisfactory utilization in the work cell and the home department is therefore one of the difficulties of work cell design. One potential downside of cells is that unless an appropriate mix of products can be found for the cell, additional investment in machinery might be required. At the heart of this issue is the lumpiness of machines and other resources: We can relocate only an integer number of machines and sometimes placing one fewer machine in a work cell is too few machines and placing one additional machine in a cell is too many. And where machines are heavy and immobile, perhaps bolted to the floor with rigid connections to electricity and compressed air, they must be dedicated to the work cell on a full-time basis. Where resources required for production are people and the only tools needed are relatively inexpensive hand tools, however, the option does exist to assign people to cells on a part-time or as-needed basis, thus avoiding the integer lumpiness issue. We will explore this tactic later in this chapter.

The next alternative solution to resolving the overloading problem that we can explore is removing product 3 from the cell. The reader may verify that by doing so, the utilization of the workstations in the cell is as shown in Table 11.8. Furthermore, the utilization of the home departments is as shown in Table 11.9.

Table 11.9 Home department machine utilization without product 3

Department	Number of machines	Annual hours of work at average demand	Utilization
Milling	5	9,065	94.4%
Lathes	6	11,638	101.0%
Drilling	3	6,265	108.8%

This tactic, then, causes the machines in the cell to be underutilized while overloading the machines in the home departments. While overtime production might be a possible resolution to this problem in the home department, it increases production costs and compromises the flexibility of the home department to react to demand spikes.

If in this example we were not contemplating a cell with heavy machinery but rather a cell with people operating relatively inexpensive hand tools, then removing product 3 from the cell might be a viable solution. In that case, we might be able to staff the work cell roughly half the time, and for the remaining work time, the workers could work back in their home departments, which would relieve the overload indicated in Table 11.9.

Note also that our utilization calculations did not include the effect of changeovers. One important observation to make about work cells is that they eliminate the need for changeovers on the machines in the cell: Once the machines are set up for their specific operation, they are dedicated to that task for all working hours and will not be switched to another task. This is convenient because some unproductive changeover time is eliminated, as are the required labor hours and materials for changeovers. Setups are still required in the home departments, however, and productive time must be sacrificed in order to make the changeovers there, whereas that is not the case in a work cell. Thus the percentage of time that a workstation is in production can be closer to 100 percent for machines in work cells than for machines in the home department.

Work Cells in Service and Administrative Processes

Figure 11.3 shows the paths for two administrative processes: creating a purchase order and processing an employee grievance. Because administrative personnel tend to be located in office buildings according to their function, the process paths in Figure 11.3 look much like production paths in a job shop: Work tends to be done in batches, travel time and distance are excessive, waiting time between processing steps is excessive, work-in-process tends to be high, and the lead time is long. Therefore, work cells can potentially be used in administrative processes to resolve these performance issues, especially where work is paper based rather than electronic. Relocating people on a permanent basis to a work cell dedicated to a limited scope of work may be difficult, however, especially in small organizations, because some departments might be small. In that case, staffing work cells on a part-time basis might effectively reduce lead time.

Figure 11.3 Administrative process flow

With more and more administrative work being handled electronically, the time to physically move paper forms from department to department is reduced, thus reducing some of the motivation for administrative work cells. However, significant wait time might still persist in electronic work-in-process, so that there still might be some benefit to cells or at least in developing a signaling device to indicate when work must be done (i.e., a visual tool).

The improvement made in the retail hiring process as discussed in Chapter 6 can be viewed as being conceptually identical to implementing a work cell. Figure 11.4 (which replicates Figure 6.1 here for convenience) shows the value stream map for the retail hiring process where an applicant needed to visit the store on three occasions in order to complete the application and interview process. Between steps in the process, applicants would return home and wait until the next step, when they would return to the store. The lead time of this process was therefore very long, which posed a problem, particularly in the winter holiday season when the store needed to ramp up its sales associate workforce by 30 percent. Collocating the managers for the interview is much the same idea as collocating different types of machines in an area and greatly reduces transportation and waiting time.

Figure 11.4 Retail hiring process

Work cells might sometimes be beneficial in service processes as well. For example, a Harvard Business School Publishing case on United Services Automobile Association (USAA)² describes an insurance claims process where both complex and simple claims were handled in the same department as part of the same process. Complex claims required all 21 steps of the process, whereas simple claims required only a subset of those process steps. An argument can be made in this case for building a work cell to handle simple claims. Doing so would reduce the jumble due to mixed processing flows, reduce the transportation and waiting times between steps, and reduce the lead time. Reduced lead time fits with USAA’s culture and strategy of offering exceptional service, and so a cell makes sense in this application, not only for increased efficiency but also for increased customer satisfaction through a more responsive claims process.

Final Comments and Further Study on Work Cells

Our explanation of work cells has emphasized reducing transportation time, making small batches feasible, reducing inventory, and reducing lead time. It is important to reiterate another benefit of cells, which is particularly important in a manufacturing environment. Specifically, collocating operations in a work cell also coordinates and synchronizes process steps because inventory between steps can be limited to a maximum level dictated in part by the space between the stations. This prevents a workstation from getting too far ahead of those before or after it. In other types of processes, work-in-process inventory between two workstations is often stored in a warehouse area, and managing the inventory level is accomplished with a computer program, which is usually less effective. In addition, more work-in-process inventory is needed in these cases to compensate for uncertainties in workstation scheduling and transportation between steps. Work cells also increase the probability that quality problems will be recognized immediately, which has a positive influence on quality. Limiting the inventory between process steps also limits the number of possible defective units between steps, which reduces rework cost. Lastly, one can argue that the quality of a repaired product is never up to par with a unit that was made properly on the first attempt, so the quality reaching the customer is improved as well.

With only 10 products, identifying parts with common processing paths was very simple and we could do it visually. This is not practical when a factory produces hundreds or thousands of parts. In that case, we might well need a computerized algorithm to help us determine which parts have similar processing sequences so that they can be processed together in a work cell. The logic of such an algorithm, however, is conceptually identical to the simple method described here.

In the utilization computations that we made in this chapter, we have ignored an important discussion of how high capacity utilization can be before problems ensue. This is a complex topic, and a full discussion is beyond the scope of this book. Note, however, that utilization must be kept at some margin below 100 percent to allow for changeover time and variability in production requirements. Day-to-day variations in demand and season-to-season demand variations require us to have some slack capacity so that we might idle the cell during slack times and respond to demand when it surges. Alternatively, the cell can be kept busy all year, in which case the cell would work ahead during slack times to accumulate inventory for the peak sales period. In some cases this is acceptable despite the cost of holding inventory and its other downsides; however, in some cases where demand is very erratic, where demand cannot be forecasted accurately, where the precise configuration of the products cannot be forecasted accurately, where products become obsolete, and where raw material costs are declining, building inventory by leveling the production schedule may not be a good idea.

We have also ignored in our discussion how the production sequence through the cell might be managed to reduce the effect of the work imbalance presented by product 3. This is a finer level of detail that the reader may investigate by looking at references that deal with work cells in a more comprehensive fashion.³

Exercises

1.In reference to work cells, verify the capacity utilization computations in this chapter where they are not explicitly shown.

2.Download the Excel spreadsheet from http://mason.wm.edu/faculty/bradley_j/LeanBook, and follow the instructions and see if you can find a good process to put in an administrative work cell.