APICS is the leading professional association for supply chain and operations management.
 
APICS Magazine > APICS Magazine - Landing Page - Everyone

Send the Right Signals

By Rajan Suri | March/April 2012 | 22 | 2

Send the Right SignalsTaking an alternative approach to kanban

The ability to mass customize individualized products is incredibly important for many businesses today. These companies offer a rich combination of options, resulting in tens of thousands of possible end products. Managing the flow of such a huge variety of products can be extremely difficult, and factories that make large numbers of items have a reputation for long lead times and late deliveries. Although kanban systems have helped improve flows, kanban is ineffective for low-volume production. The good news is there are some emerging techniques available to support varying factory environments.

Ten years ago, Tom Schabel, president of Alexandria Extrusion Company (AEC) in Alexandria, Minnesota, was looking for ways to improve his operation. AEC specializes in small-batch and custom aluminum extrusions. While the company had an excellent reputation for quality, it suffered from long lead times and poor on-time delivery performance. Around the same time, P&H Mining Equipment in Milwaukee, Wisconsin—which makes large custom equipment such as mining shovels—was facing similar issues. Both businesses had looked into kanban and decided it did not fit their operations.

AEC, P&H, and other companies with similar goals partnered with the University of Wisconsin-Madison to prove out a competitive strategy called quick response manufacturing (QRM). As part of this effort, they pioneered the implementation of paired-cell overlapping loops of cards (POLCA), an alternative to kanban.

Developed at the Center for Quick Response Manufacturing along with industry partners over the last 18 years, QRM is a strategy for reducing lead times by 80 percent or more, heightening quality, and cutting costs at businesses that manufacture low-volume and custom products. While QRM is applied enterprise-wide—including in the office and the supply chain—a key part is reorganizing 
the shop floor into QRM cells, which are highly flexible and perform a variety of opera-tions. For instance, a company could have sev-eral types of fabrication cells feeding multiple subassembly cells, 
in turn feeding final assembly cells.

The next step is to route jobs through the right combination of cells to make the needed end items. Material requirements planning (MRP) is used to map out this routing and to create the initial schedule. Then, as jobs are launched, POLCA helps keep them flowing and on target with delivery dates.

Kanban systems have been successful at controlling jobs, so why not use this methodology? In short, kanban is a pull system: Parts are pulled through the factory as they are used in downstream operations. Starting from finished goods, when a container is shipped, a signal is sent to the previous operation to restock it. That previous operation has partially completed material waiting in its stock. When it draws a container to work on, it sends a signal to its previous operation to resupply that material. These signals are in the form of kanban cards, and each specifies a part number and quantity. Thus, kanban requires partially completed products stocked in containers at each stage of the manufacturing system and supply chain.

If you make parts with high demand, material keeps moving, and kanban works well. However, consider what happens when you make products with low demand. Suppose a company makes axles for non-automotive applications such as construction equipment. Let’s say this company stocks a container with six axles of a certain type, and it gets an order from a distribution center about once a year for one of these containers. When this order arrives, the container is shipped, and a kanban signal is sent to the previous operation. Within a couple of days, the previous operation completes the six axles and restocks the warehouse. These axles will sit in the warehouse for a year before another order is received.

So the business would have an inventory turn rate of only once a year for this product—in an era when management expects to achieve turns of 20 or more. Not only are axles sitting in the warehouse, but there also are partially completed products sitting throughout the whole flow path. In fact, kanban systems have a name for these intermediate stocks: supermarkets. When volume is high, items move through the supermarkets quickly. But in low-volume environments, instead of reducing inventory, this approach actually adds inventory.

Now suppose this company receives an order for two custom-engineered axles for a prototype vehicle. Kanban won’t work at all because it is a replenishment system. You ship finished goods, and then send a signal to replenish them. But you can’t have something in finished goods if it has never been made before; it hasn’t even been engineered yet.

Kanban was not intended to operate in environments of low volume or custom products. However, POLCA is designed specifically to meet these needs.

When material flows between any two cells, they are connected by a POLCA loop, which contains a number of circulating POLCA cards. (See Figure 1.) These cards are specific to this loop and called A/B cards. When Cell A is scheduled to start a job destined for Cell B, it needs to have an A/B card in order to launch the job into Cell A. If a card is available, the job is started and the card is kept with the job. When Cell A completes the job, it is sent along with the A/B card to Cell B. Cell B’s operation will be discussed later on in this article—but, for now, note that when Cell B finishes working on this job, it sends the job to the next cell and sends the A/B card back to Cell A. Thus, a card coming back from Cell B conveys the message “we finished an A/B job; you can send another.”

This highlights a distinction between kanban and POLCA. Kanban is an inventory signal (when a quantity of parts is used up, it tells the previous operation to restock that inventory). On the other hand, POLCA is a capacity signal (returning POLCA cards signify availability of capacity in downstream cells).

This means that, unlike kanban cards, POLCA cards do not list a part number and quantity, only the names of the two cells. To enhance visual management, cells are assigned colors, and each POLCA card has two colors. The left half has the name and color of the originating cell, and the right half has the name and color of the destination cell. (See Figure 2.)

How does Cell A decide which job to start next? Based on ship dates of end items, the MRP system back-schedules requirements in the normal way and calculates start dates for jobs at each cell. With POLCA, these are called authorization dates because cells need to follow additional rules before starting jobs.

For each cell, the MRP system compiles the usual dispatch list (say, once a day or once a shift), which shows all the jobs that have yet to be started. This list is ordered by authorization date, with the earliest at the top. Only jobs with start dates of today or earlier are authorized. The cell team takes the first authorized job on the list and checks the next cell for this job (provided in the dispatch list). Suppose the job is going to Cell D next. The Cell A team checks if it has an A/D POLCA card. Each team has a bulletin board where it organizes its available POLCA cards. (See Figure 3.) If an A/D card is available, the job is launched into the cell along with this card. If no A/D card is available, then the team must skip this job and go to the next one on the list.

Why is this rule beneficial? Let’s say there are five A/D POLCA cards. If the dispatch list for Cell A has a job destined for Cell D, but no A/D cards are available, this means all five cards are in use with other jobs. Either Cell D is backed up or enough work is on its way to Cell D, so sending another job means it will just add to work in process (WIP). On the other hand, other cells may be waiting for work from Cell A. If the next job on the list is for Cell B and an A/B POLCA card is available, this implies Cell B could use the work. Skipping the job for Cell D and working on the job for Cell B is a good idea. Thus, POLCA makes effective use of capacity by ensuring that upstream cells work on jobs that go somewhere—instead of jobs that end up in bottlenecks.

It may sound radical, but if no cards are available for any authorized jobs, the cell cannot start any jobs. Typically, if people don’t have work, a supervisor will look ahead in the schedule and start jobs even if they get made earlier than needed. In POLCA, jobs with dates in the future cannot be started even if the appropriate cards are available.

At first, this makes management uneasy. Yet there are good reasons why these rules have resulted in better operations. Whenever you put capacity into an unnecessary job, you steal capacity from another job that might have needed it—plus, you create WIP. Say you start a job ahead of schedule, and then a POLCA card arrives for a job that’s behind. For instance, if you have already set up your machine, you want to finish the job you started. This job adds to WIP, while the late job has to wait even longer.

Let’s return to the job that went from Cell A to Cell B. Suppose the next cell for this job after Cell B is Cell G, so there is a B/G POLCA loop as well. (See Figure 1.) Each cell implements the same scheduling logic, so, when the job arrives at Cell B, the team must wait for the job to be authorized and all other jobs above it on the dispatch list begun—or be waiting for POLCA cards. Then, it must check if a B/G card is available. If both conditions are met, the B/G card is allotted to the job, and it is started. However, the job also arrived carrying an A/B card, and that card will not be sent back to Cell A until the job is completed at Cell B.

With kanban, you send a card back as soon as you collect the material. But because POLCA is a capacity signal designed to work with jobs having varying work content, you don’t want a cell to send a signal until it actually finishes the job. So the job that was launched into Cell B still carries the A/B card with it. In addition, it now has the B/G card. While it is being worked on in Cell B, the job will thus have two POLCA cards. (See Figure 4.) Hence, POLCA loops overlap throughout the routing—except at the first and last cell.

During the past 10 years, POLCA has been implemented in many types of factories, including machining and fabrication, electromechanical assembly, and equipment manufacturing. Here are some benefits that these factories realized:

  • Companies don’t need to replace their MRP system. POLCA builds on the existing system and improves its performance.
  • POLCA ensures effective use of capacity and reduces congestion. For example, P&H Mining connected 12 shop floor cells and other facilities. During the first year, WIP was reduced by $3 million.
  • Frequently occurring bottlenecks are flagged by POLCA. When upstream cells are held up by the same cell, teams point this out and work with management to find a solution.
  • POLCA continually re-sequences jobs during their flow. Jobs that are behind bubble up to the top of the dispatch list—but, at the same time, areas that are bottlenecked are avoided and work is sent to other areas that can use it. After implementing QRM and POLCA, Alexandria Extrusion achieved near-perfect service levels, while quoting shorter lead times and carrying less inventory.
  • Companies have seen the elimination of “hot jobs.” Bosch Hinges, which makes specialty hinges in small batches, was constantly struggling to meet delivery dates. The owner attended a QRM workshop and pioneered the first implementation of POLCA in the Netherlands. Today, the company makes even smaller batches and has more work orders in process—plus, it has no rush 
    jobs, has no late deliveries, and is more profitable.

Providing customized products with short lead times is a powerful strategy. QRM and POLCA provide companies with an effective strategy and tools to achieve this objective and to retain and even grow jobs.

Rajan Suri is emeritus professor and founding director of the Center for Quick Response Manufacturing at the University of Wisconsin-Madison. He is the author of two books on QRM, including It’s About Time, The Competitive Advantage of Quick Response Manufacturing. Suri may be contacted through his website at rajansuri.com.

Comment

  1.