The Pattern: Flow, Pull, and Heijunka
A pattern is a dependable set of attributes that are prevalent in a particular group or company.
Many good companies have respect for individuals, and practice kaizen and other TPS tools. … But what is important is having all the elements together as a system. It must be practiced every day in a very consistent manner, not in spurts. 1
Ohno and others spent a tremendous amount of time on these three subjects. Ohno talks extensively about these elements in all three of his books, referencing just-in-time (JIT), kanban, supermarkets, continuous flow, pull, leveling, and using various other terms. These concepts have received considerable coverage by those who were involved in the evolution of the Toyota Production System (TPS). However, in my experience, these critical concepts are not viewed with the same importance as Ohno and others viewed them. Ohno said, “The first step in implementing Toyota-style production is to create flow; next is to establish JIT production.” 2 In Ohno’s development of TPS, flow came early, the first thing he did in the machine shop. Next came pull, after he’d observed American supermarkets. Heijunka was instituted after pulling.
In view of the Toyota Template, the pattern consists of the “critical must conditions.” Flow, pull, and heijunka are the concepts that are required for a JIT production system. Remember, Ohno was building a system based on the elimination of waste. Everything he did was focused on this goal. Flow, pull, and heijunka were done in his continuous efforts to banish waste. Some of the tools that were implemented along the way, such as standard work or 5S, can be used outside of a JIT system. This is the case in many tool approaches, because this “pattern” is ignored. Standard work, problem-solving, kanban, and so on can be implemented without flow, pull, and heijunka, but the effect on production is greatly diminished and difficult to sustain. These three are closely related, dependent on each other, and are critical “musts” for the success of a lean initiative.
CONTINUOUS FLOW: ALIGNMENT OF PROCESSES IN THE ORDER OF VALUE ADDED
[I]n the plant we must first rearrange the equipment in the order that people work and in the order that value is added to create a flow. In creating flow, it is not enough to look at just part of the process. 3
[A]rranging processes in the natural workflow was a precondition for linking the processes and lines as a pull system. 4
To put this in context, Ohno started to arrange equipment in this manner in the machine shop in the late 1940s to eliminate waste. By 1953, “stores” had been set up at the various stages of the manufacturing processes. This was the beginning of an effort to realize Kiichiro Toyoda’s idea that “it is best to have the various parts arrive alongside the assembly line just-in-time.” 5
There are many advantages realized when continuous flow is created. The most important one, cited by Ohno, is that creating flow is a “precondition for linking the processes and lines as a pull system.” 6 This is the origin of one-piece flow. Linking processes to one another in order makes it possible to implement a pull system. It’s a “precondition.”
One advantage of flow is the ability to combine work or move work around between processes as needed. When they’re connected in order, it’s much easier to see and understand where imbalances exist. When processes are not connected in continuous flow, these imbalances are more difficult to understand. There tends to be much unnecessary batch building and transporting of parts between processes. Additionally, it’s more difficult to differentiate between the normal and the abnormal.
The improvement of what Ohno called the “baton passing zone” is also possible. This is the area between processes where work hand-off occurs. Ohno equated production to a relay race where “a faster runner can make up for a slower runner in the baton passing zone.” 7 When processes are arranged in the order in which value is added to the product, connected processes can aid one another in times of need. For example, if the upstream process experiences a delay due to a scrap part, the downstream process can assist by moving over to help in catching up. There’s no telling how many times this happens in a day at Toyota. Likely thousands of times throughout the plant. Additionally, team leaders will step in on many occasions to assist team members when they experience a problem. “To improve the efficiency of the line, the supervisor must establish a baton passing zone where workers have a chance to catch up.” 8
A third big advantage is the potential to minimize, or even eliminate, the need to transport parts between processes. When they’re not aligned in the order in which value is added, parts from upstream must be transported to the downstream process. This requires additional people and equipment. When they’re aligned, transporting can be avoided by simply passing along the product with rollers or a conveyor, particularly in a one-piece flow situation. In fact, in one-piece flow areas, this is the norm.
Finally, inventory can be greatly reduced. One of the primary reasons for buffers between processes is the distance between them. When processes are not well connected, more buffer is required. If excessive buffers, or work-in-process (WIP), are required due to poor process connections, then money is tied up on the floor in inventory unnecessarily. Cash flow is negatively affected. It’s possible to have millions of dollars sitting on the production floor between processes in work-in-process (WIP) when it could be avoided.
Continuous flow is most important in processes where items are, or can be, made one at a time. Identifying these processes and closely linking them together in order should be done first. The opportunity to eliminate waste in the form of excess inventory and transportation by simply establishing continuous flow is the result. And more importantly, as Ohno said, flow is a precondition for implementation of a pull system.
Culture: Creating continuous flow is another contributor to the building of the lean culture. It requires multi-skilled workers, reducing boredom and improving teamwork in the baton passing zone. When employees are dependent on one another and help each other, information, communication, and the general working conditions are improved. It’s also critical to the property of waste elimination.
PULL: PACES AND PRIORITIZES PRODUCTION, AND PROMOTES PROBLEM-SOLVING
Looking at things individually they say they are doing a good job of producing gears, or that they are using robots very well, or that they can do the work with just 3 people. But these items can only be sold when they are together as a set. 9
[O]nly considering the production plan for each process, we would produce parts without regard to the later processes. Waste would result – defective parts on the one hand, huge inventories of parts not needed immediately on the other. This reduces both productivity and profitability. Even worse, there would be no distinction between normal and abnormal states on each assembly line. 10
The Toyota Production System is a pull method. To understand its tremendous success, one has to grasp the philosophy behind it without being sidetracked by particular aspects of the system, such as Kanban. 11
It takes great effort to practice the 6 rules of kanban … thus a half-hearted introduction of a pull system brings a hundred harms and not a single gain. 12
I’ve often heard people say that the processes must be stabilized before a pull system can be introduced into production. What this means exactly is not clear. Generally, having standard work, inventory accuracy, and 5S are mentioned. All desirable conditions, for sure. No doubt, having stability on the manufacturing floor is a condition required to achieve efficiency.
Stabilizing production consists of two very important concepts. They are pull production and heijunka, or level loading. Along with flow, these were the concepts that changed the game for Toyota. If there is a silver bullet to TPS, continuous flow, pull, and heijunka are the three conditions that must be met.
An old-timer in the assembly section told me how President Kiichiro Toyota said to him that the most efficient way to assemble parts in an assembly plant was when each part arrived Just in Time. 13
At Toyota, the decision to pull production through the plant was the beginning of Kiichiro Toyoda’s idea that parts arrive at the workplace in the quantity and at the time needed. This requires a well-thought-out plan. Pulling is a major change if pushing with schedules has been the practice. In the beginning, the actual orders must be pulled through the plant. The pulling of individual parts is a result of the pulling of the orders. On more than one occasion, I’ve heard this idea of pulling interpreted as simply setting up parts on a kanban. Some call it a “supermarket strategy.” Setting up parts on a kanban is a good idea only if the orders are being pulled through the plant with heijunka. The fact is that the kanban does not work well unless the orders or jobs are pulled. When a kanban system is set up for a group of parts in a push system, other problems will ensue. Pulling begins at the end of the plant, in shipping, and moves upstream. Pulling must be the method of making things in lean manufacturing. Ohno said, “Kanban must work effectively to maintain just-in-time in the plant. And for Kanban to be effective, stabilization and production leveling are indispensable conditions.” 14 This is a very important point. Heijunka and pull are indispensable conditions.
In a push system, such as Material Resource Planning (MRP), orders are pushed to the various areas of the floor based on a schedule over a given period. From an efficiency standpoint, there are two problems with MRP systems. One is timing. Products, or sub-assemblies, related to the same order are manufactured at various rates in the departments where they’re made. Each area works in the silo of its own schedule. Their focus is to meet their schedule. Typically, the problem is not that each area can’t make the items on its schedule. The problem is timing, or synchronization.
One timing issue is that the process receiving the orders doesn’t work toward the needs of its customers. In an effort to maximize their work in an efficient manner, they work at their own pace and in their own order, irrespective of customer needs. No attachment to the customer exists. This order may be determined simply by the fact that the easiest items are made first. Or maybe, the material needed for a particular item happens to be available. No matter what the reason, each department is attempting to maximize its own productivity without regard to the effect on its customer.
In batch build processes, parts should be batched to maximize process efficiency. In a push system, the batch quantities change with every order. Since the batch quantity is linked to the projected demand over time for a particular item, this projected demand, and thus the batch quantity, can be highly variable. Therefore, the batch quantity will always be a suboptimal number. Furthermore, because the exact quantities are made based on the order, when a quality problem arises anywhere in the process, a shortage will occur.
For example, the processes must batch by part number to minimize changeover of equipment or material. In this case, the entire time period’s worth, say 1 week’s worth, of Part A is built, then Part B, and so on. By the end of Monday, the process has built all the week’s requirements for A and B. The problem is that on Monday, their internal customer has 10 orders requiring parts A, B, C, D, and E. The internal customer will receive parts C, D, and E later than required, which prevents them from building the 10 orders. Timing!
Additionally, the customer must store a week’s worth of parts A and B that they received on Monday in their area. Too early is the same as too much. And remember, too much, or overproduction, is the worst type of waste. They may also have begun to build their 10 orders, only to find out later that they won’t have all the parts needed to complete their orders on Monday. Now, they must store partially built orders until parts C, D, and E arrive the next day. This results in lots of excess WIP, excessive handling, and storage problems. In addition, and maybe less obviously, this type of production is disrespectful and frustrating to employees who come to work to do a good job.
A second timing problem is that the upstream process and its downstream customers are not synced up—the same issue that confronted Toyota years ago. Synchronization cannot be scheduled. In a push system, because processes don’t run at the same pace, there is no connection between them. Each is working off its own schedule. Either the upstream process is too slow, causing wait time, or the upstream process buries its customer with more parts than they can use. And rarely the right ones. When this happens, it can become difficult to find what’s needed, even though lots of parts are present. The opposite can also be true. The downstream process can be slower or faster than its upstream supplier. How can the actual state of the current condition be determined? Ohno said:
[O]nly considering the production plan for each process, we would produce parts without regard to the later processes. Waste would result – defective parts on the one hand, huge inventories of parts not needed immediately on the other. This reduces both productivity and profitability. Even worse, there would be no distinction between normal and abnormal states on each assembly line. 15
There is some good news. You don’t have to throw out your MRP system. Just because the system makes the schedule doesn’t mean all the schedules must go to production en masse. Use the MRP system to make the schedule, and then, separately, pull the orders in the needed sequence through the plant. Toyota has a schedule, but it’s pulled through the plant.
Additionally, timing issues can often lead to capacity questions. Sometimes, when items are not at the right place, at the right time, in the right quantity, the knee jerk assumption is that it’s because there is a lack of capacity. Acting on this assumption could lead to additional expense such as overtime, adding labor, or even the purchase of additional equipment. The poor timing root cause is difficult to countermeasure when pushing using schedules and can lead to poor decisions.
Pulling is the countermeasure. It’s the only way to sync up products throughout production. A JIT system requires that components mate up as they move down the production line. When beginning to pull, there will likely be a delay at the point at which parts should come together. When this happens, a bottleneck exists. This problem is addressed like any other. It’s time to do some problem-solving! What caused the delay, and what needs to be done to countermeasure the problem?
This is a good time to talk about TAKT time. TAKT time is the time that should be taken to produce something. It’s based on customer demand. TAKT may differ in different parts of the plant and/or with different products. Daily total operating time is determined based on machinery operating at 100% efficiency during regular working hours: in other words, what could be made if the line ran at 100%.
On a multi-model production line, TAKT time allows parts of different types to be produced. TAKT time also allows the right parts to arrive at the right process at the right time. TAKT time is the rate at which each product should be made. TAKT time is a calculation that can change over time.
|
Another important aspect of pulling through the plant is the use of predetermined buffers. No buffers would obviously be the ideal, but the reality is that there are several good reasons to have buffers within and between processes and departments. At Toyota, we had many buffers for smoothing product flow.
One reason for a buffer is that it mitigates downtime between departments. This is especially true between areas with a lot of automated equipment. Equipment will fail from time to time. That’s just a fact. Having a buffer between these areas and the downstream customer to cover these times when downtime occurs makes good sense. Otherwise, this machine downtime could shut down an entire area. In fact, if there were no buffer between areas with lots of automation, every time a machine stopped in one area, this could result in stopping production in its customer’s area. To be clear, I’m not advocating large buffers because machines don’t run well. This is about normal expected production loss due to delays caused by machine problems.
Another reason could be the distance between supplier and customer. When customers and their suppliers cannot be closely located, then some buffer is needed between them due to the distance. This could be between two processes on a line or between two processes that are clearly separated. An example might be the distance between areas, departments, or lines.
A third reason for a buffer could be for heijunka. Every department may have different requirements for smoothing production. There could be a need to rearrange products for the downstream department’s heijunka requirements. If the next department needs some buffer to rearrange the incoming products in a way that provides for smoother flow through their area, then a buffer is needed for this activity. More about this one later.
And, of course, building in batches is another obvious situation where buffers are required. Some products in production are pulled one at a time, or one-piece flow, through the plant. But many parts must be built, and subsequently pulled, in batches. This situation exists with a machine that makes large quantities of multiple part numbers. A stamping press is a good example. A press can make many parts quickly, but the dies/material must be changed out when a different part number is made. Making one piece at a time makes no sense here. A press may make 40 different part numbers for an auto. The questions for the press operator are which part number should be made, when should it be made, and how many should be made? (Right part, right quantity, right time.) In push systems, a schedule is used to make this determination, and we’ve covered the timing and other problems with schedules.
In 1956, I toured U.S. production plants at General Motors, Ford, and other machinery companies. But my strongest impression was the extent of the supermarkets prevalence in America …. we made a connection between supermarkets and the just-in-time system …. From the supermarket we got the idea of viewing the earlier process in a production line as a kind of store. 16
Supermarkets use a pull system, and for good reason. The obvious issue that requires supermarkets to use a pull system is spoilage. Food wholesalers earn 3%–4% earnings before interest, taxes, depreciation and amortization (EBITDA)/sales versus a general industrial average EBITDA/sales across all sectors of the economy of >15%. 17 Grocers must be much more conscious of waste than other industries, because they have less margin for error. Many food items cannot be held for very long in inventory, because they go bad. Were they to use a push system whereby they bought fresh produce or meat according to a schedule, their spoilage would increase dramatically, and they probably wouldn’t be in business long.
So, as Ohno observed, the modern American grocery stores use a pull method. By using a pull system in the grocery business, they can easily and quickly see their waste. Visibility is a major advantage to a pull system. Similarly, manufacturers who use pull can also identify wastes much more easily and quickly. In a push method in manufacturing, the waste is not as obvious as wilted lettuce and thus, a little more difficult to see.
The condition of the Toyota Template is to level pull the actual orders so that they arrive at their respective destinations along the way and at the customer when they are needed. Manufacturers make many parts, or sub-assemblies, that must sync up to move to an internal customer or to be sold to an external customer. If making cabinets, the doors, sides, tops, and bottoms must show up together at the assembly line for the cabinet to be assembled. This applies to most businesses. Even in businesses where many sub-assemblies are shipped to a worksite for assembly onsite, all parts must arrive in time for assembly to take place. Many times, in these circumstances, the manufacturer is unable to ship partial deliveries to a worksite, because the customer must assemble the product in the field. It can be costly to receive bits and pieces of the total product randomly because of a synchronization problem in manufacturing. Customers also don’t want to pay for a partial product in these circumstances, and they shouldn’t.
A substantial benefit gained when changing from a push to a pull system is the large reduction in inventory on the floor, or WIP. This is because the buffers between and within processes are now controlled. Each buffer has a predetermined quantity, and if a buffer is full, production upstream stops. Overproduction, in terms of WIP, is eliminated. What was once unsynchronized production in the push system becomes synchronized in the pull system. This reduction in WIP can be large, likely 50% or more. This means that money previously tied up in inventory is now cash in the bank.
The reduction of WIP in manufacturing is also beneficial to lead times. Little’s Law says that
|
If WIP is 100 units and you can process 25 units/day, then 100/25 = 4-day lead time. Now, let’s say that after implementing a pull system, the WIP is reduced to 50 units. Now, 50/25 = 2-day lead time.
When WIP is reduced, lead time decreases.
As for timing…You could only get the timing right if you conveyed the parts by having the following process pick them up. You’d screw up the timing if you simply pushed the parts onto the following process according to a production plan. 18
Pull production is the only way to get the timing right!
Predetermined buffers provide another significant benefit to pull: the ability to see and understand the abnormal condition. Buffers are a visual control. It’s very easy to determine whether the supplier is behind by observing the buffer. Conversely, if the buffer stays full, then the upstream process must stop, because there’s no place to put their product. This may be a signal that there’s a problem downstream. This visual control is very useful in understanding where problems occur, aiding in problem-solving.
An additional benefit, which may be less obvious, is headcount reduction. Typically, in a push system, scheduled volumes can fluctuate tremendously from day to day in many processes. One day, this process looks like it can’t keep up. The next day, it’s another process that appears to be behind. To combat this issue, many processes are staffed to absorb these fluctuations in demand. Over time, processes throughout the plant become overstaffed.
After pull is implemented, production is synchronized, and a pace, dictated by TAKT time, is established. This highlights the staffing situation. Because the pace is steady and established, most processes will be overstaffed. Remember, they were previously staffed to meet the extremes of the daily schedule in each area. Like WIP reduction, this can also result in a very substantial decrease in headcount and labor cost.
Finally, as mentioned previously, when the excess inventory disappears from the floor, good 5S becomes a goal that is much easier to attain.
On JIT, Ohno said, “once decided upon, it should be undertaken with a firm and determined mind.” 19
Culture: Pull systems do away with silo production associated with a push system. This silo mentality is the opposite of teamwork. In a pull environment, there is “tension in the line.” Processes are dependent on one another to get their work done. This tension creates a heightened sense of urgency throughout the manufacturing process. It ties internal suppliers and customers together at a pace that requires cooperation and teamwork. The effects of a pull system are very important contributors to a lean culture.
The operating method of the Toyota Production System is Kanban. 20
Kanban is a way to achieve just-in-time; its purpose is just-in-time. Kanban, in essence, becomes the autonomic nerve of the production line. 21
The Kanban links production in each process to the pace of production in the following process. 22
I must admit that the first time I encountered a kanban during my training, I was intrigued. My formal introduction to the kanban was during my training on the kanban pick-up route at Tsutsumi. I remember thinking that a kanban was like currency to the processes that used them. They couldn’t get more until they presented their “money.” One of the first things that aroused my curiosity was all the information on the card and what it meant. My trainer patiently explained each bit and why it was on the card. The information was carefully thought out, considering the information both Toyota and the supplier needed and no more.
It wasn’t until later, when I got back to Georgetown, that I became aware of the rules, how to determine the number in circulation, the different types and how they were used, and the importance of time. I also came to understand the reason for only two addresses. The kanban only works as a closed loop. When a third stop is entered into the equation, timing becomes a problem. If a kanban is used internally between two processes, the supplier and the user, it must only circulate between these two locations.
At Toyota, we used two types of kanban, each with a specific purpose. The parts withdrawal kanban was pulled at the user location for reordering another container from the supplier. We always pulled the kanban and dropped it in the kanban post when the first part was used out of the container. This was the rule. This differs from two-bin systems where the container acts as the kanban. In this system, the kanban is effectively dropped after the last part in the container is consumed. Of course, this reduces the lead time and may result in needing an additional container in the system. Also, with the parts withdrawal kanban, the minimum number of kanbans for any part is two. The reason for the two-card minimum is that if one is lost, the process will see that they’re running low and can take remedial action. This is something my trainer shared with me. I mention it because I’ve seen situations where only one parts withdrawal kanban is in circulation. When there’s only one kanban, the process may not know the kanban is lost until they’re completely out. The second type of kanban is the production instruction, or signal kanban. It’s used to signal the upstream producer that it’s time to make more. This kanban, which I’ll explain later, is used to connect batch production with one-piece flow processes.
Kanban is a tool for realizing just-in-time. For this tool to work fairly well, the production processes must be managed to flow as much as possible. This is really the basic condition. Other important conditions are leveling production as much as possible and always working in accordance with standard work methods. 23
A word of caution on the use of the kanban. Though it can be used, the kanban does not work well in a push system. This is because a pull system for a subset of parts has been introduced into a push environment. The effectiveness is minimal at best, and other problems are created in this situation. The timing problem, previously discussed, is amplified when kanban is introduced, because timing is critical in a kanban system. In the long run, this creates distrust of the kanban for replenishment, many times leading to work arounds and collapse. And in the big picture, when a lean tool doesn’t work well, there’s a tendency to think lean “won’t work here.”
In the TPS, overproduction is completely prevented by Kanban. 24
HEIJUNKA: PRODUCTION SMOOTHING
So, if you’re going to make different kinds of assemblies, you need to distribute the production of the different kinds evenly throughout the day … If you don’t level production, you’ll keep running out of parts. You level production by changing the order in which you make things. 25
The most important prerequisite of JIT production is production smoothing, or small lot production. 26
Generally, demand for products is highly variable. Customers don’t order the various products you sell in a level fashion. Orders are unpredictable both in volume and in the type of product. This unpredictability can wreak havoc in a manufacturing environment for many reasons. Toyota understood the importance of production smoothing after they began pulling.
Our biggest problem with this system [pull] was how to avoid throwing the earlier process into confusion when a later process picked up large quantities at a time. Eventually, after trial and error, we came up with production leveling. 27
Heijunka means level loading or production smoothing over time. In manufacturing, this means level loading of the production floor. There are two aspects. The first, and easier to achieve, is to level load production by volume. The second aspect of heijunka is to level load by product mix or type. This is the more difficult but also yields the greatest benefit.
The lack of production smoothing is a primary and continuous contributor to waste throughout the plant. Mura (unevenness) and muri (overburden) are greatly reduced in every process in the plant when heijunka is achieved.
When customers don’t order products in an order that leads to level loading of the production process, which is always, the orders arrive at customer service or sales in a random fashion. The question is how to meet customer demand.
One way would be to simply make products in a first‑in, first‑out (FIFO) sequence in which the orders arrive in customer service. Let’s look at a simple example. Let’s suppose that you make four different products (A, B, C, and D) that share many of the same processes. Time to produce each varies by product type. A takes 10 minutes, B takes 15 minutes, C requires 20 minutes, and D requires 30 minutes to produce. In this scenario, both mura and muri would be prevalent, because production is at the mercy of chance. Also, each process must be staffed, have materials available, and enough equipment to meet peak production in every process.
An example of heijunka in Body Weld was in the Roof and Cowl area. Customers don’t buy the same number of cars with regular roofs as those that have sunroofs. Though it can vary, let’s say that one in seven cars sold had sunroofs. The cycle time to build a sunroof is considerably longer than the cycle time for a regular roof. Toyota used a product indicator in the process (a light board to indicate body number and option, in this case which roof) that told the team member which roof to build. In this example, the member would build six roofs, one sunroof, six roofs, one sunroof, and so on. The process layout, machine cycle times, and walk time were set up to enable one person to build both types. After the roofs were built, they went to buffer pallets for each type. A carrier, again responding to body number and type, picked up a roof from the correct pallet, either the sunroof pallet or the regular roof pallet, and delivered it to the Framing Body Line to be mated up with the underbody and side members.
For these sub-assemblies to sync up together on the correct car body, heijunka was established. Production Control arranges them in the build order required to achieve heijunka for Body Weld considering options such as the roof types. Without heijunka, more employees would be required to perform the same work. If the orders were built in a random fashion, the Roof and Cowl area would need to be staffed to handle three or four sunroofs consecutively, since this could happen. In this case, the sunroof members would be very busy, while the regular roof member would be waiting.
Heijunka can be a difficult concept to understand in a traditional manufacturing environment. It’s hard for leaders to believe in something when the benefits are less obvious or easy to see. Typically, leaders want to see direct benefits in anything they undertake. How is smoothing production throughout the plant measured? What’s the Key Productivity Indicator? There isn’t one.
The fascination with tools, the “point kaizen” emphasis, and a lack of problem-solving have led to a focus on symptoms. The lack of production smoothing ensures the introduction of random variation in the form of mura and muri, from the very beginning, into the entire manufacturing system. This variation results in costs that are not easily seen. Delays at various places along the production process result in waiting elsewhere.
The lack of level loading can affect direct costs too. Let’s look at an example that illustrates the effect of not level loading.
There was a manufacturer that made items that were boxed up at the end of each assembly line. This was a manual process performed by a worker. As the items came down the line, the boxer would choose the box based on the size of the product. The company had to stock several different box sizes. There was always some danger of running out of one size or another. The company decided to invest in a corrugated box-making machine that cut boxes to custom sizes. Making boxes the correct size for each product would save some costs.
The plant had two production lines that made the same products. They ran both lines on first shift and neither line on second shift. On third shift, the first line ran for 4 hours; then, the crew would move to the second line and run it for 4 hours. The assembly lines ran:
First shift = 16 hours
Second shift = 0 hours
Third shift = 8 hours
Total = 24 hours
The lines were simple conveyor belts on which the product was loaded and assembled as it moved down the line. Their production volume was maxed out due to environmental reasons. They couldn’t make more product.
So, when it was decided to buy corrugated box-making machines—because the lines weren’t leveled by volume on each shift—two machines were needed, one for each line. If the 24 hours of production had been spread evenly over three shifts, there would only have been the need for one machine.
In addition to needing two machines instead of one, the assembly area sat empty for an entire shift. Again, if production had been level, half of the assembly area floor space would have been available for another use. Finally, because no production was run on second shift, the buffer from the upstream supplier had to be much larger than needed. This was because the line ran twice as fast on first shift as on third shift. So, the buffer had to be twice what it would have been if production had run evenly on all three shifts.
To achieve heijunka, there are some cases where we must employ the use of buffers. At Toyota, we had buffers in many areas of the plant: between processes, between groups and automated lines within departments, and between departments. One such buffer was critical to heijunka in the largest department in the plant, Assembly.
The Paint Department has a buffer area called the selectivity bank. This area held a standard number of cars. It was used to shuffle the cars that came from body into an order that achieved the heijunka requirements for Assembly. This was required because the heijunka requirements for Body Weld were different from the requirements for Assembly. In many manufacturing environments, the requirements from area to area for level loading differ. Buffers can help in heijunka.
“Heijunka is a pre-requisite for Just-in-time delivery.” 28 Conversely, JIT delivery is not possible without production smoothing.
Culture: Heijunka built on this culture by smoothing out production for everyone. This was good for employees, because the work was better paced, and it solved the problems caused by erratic production numbers.
SINGLE-MINUTE EXCHANGE OF DIES (SMED): SET-UP TIME REDUCTION
In production leveling, batches are made as small as possible in contrast to traditional mass production, where bigger is considered better. 29
[W]hile producing item A in quantity, the process may not meet the need for item B. Consequently, shortening setup time and reducing lot sizes becomes necessary. 30
Rapid changeovers are an absolute requirement for the Toyota Production System. 31 31
One of the challenges to production leveling is the need for batch building processes to make smaller quantities of each part. This is because the downstream processes will no longer be batch building as a result of production smoothing. Many companies are attempting to make things using single-piece flow on assembly lines nowadays. Sub-assemblies are made and passed to the next process one at a time based on some signal. However, as we move upstream to suppliers of the assembly lines, or sub-assemblies feeding the assembly lines, we encounter a problem with one-piece flow. Many of these suppliers, such as stamping presses or injection molders, must make parts in batches or lots.
Some machines or processes must batch, because they make large quantities of multiple part numbers using different materials, tools, and machine settings. Typically, machine operators seek to maximize the use of their machines by building large quantities to reduce the number of changeovers. It wouldn’t be practical or make sense for these machines to try to make parts one at a time. A large chunk of production time would be eaten up in changeover time from part to part.
In the auto industry, this is true of the Stamping Department. Stamping presses are very large machines—as large as a 1500 square foot house! A press may make 40–50 different part numbers. In this scenario, one-piece flow is not an option because of the number of different parts and the die change time. This circumstance also exists in industries outside of automobile manufacturing, anywhere batch building must be done. The only way for a press to make all the different part numbers and keep the downstream processes supplied is to build to a store based on actual usage. At Toyota, this is accomplished through use of a Triangle Kanban.
The Triangle Kanban serves the same purpose as the parts withdrawal kanban, except that it’s a signal kanban. The kanban is placed on the container of a particular part that represents the trigger point. Before this container is taken to the line in the Body Shop, the kanban is removed and placed on a kanban post in the Stamping area. These kanbans are retrieved by Stamping employees periodically and returned to the press where the parts are to be made. They’re generally kept in FIFO order at the press. By this method, all Stamping production is virtually self-scheduled. Of course, there are occasional exceptions, such as scrap. This abnormal situation will accelerate the movement of the particular kanban for that part, so that it is run sooner than normal. In this way, the kanban reacts to the scrap reflexively, in effect, automatically rescheduling the part production sooner.
The Triangle Kanban system accomplishes linkage between batch building and one-piece flow by systematically minimizing batch sizes, which increases changeovers. This is the opposite of what each operator was attempting to do with their individual machines in a push system. The difference is that the Triangle Kanban system considers all parts on each machine that builds in batches in an area. This is a very important point.
The Triangle Kanban moves between the store and the press. This works best when heijunka exists in the plant, because lot sizes and inventories of the various parts can be minimized. This method can be used without heijunka. However, if level loading isn’t present, the trigger point of every part in the store must be increased to account for the uneven pull from downstream, and lot sizes will also increase.
Another benefit of this kanban is that only one kanban is needed per part number. This is because the kanban is placed at the trigger point for the part. As I stated previously, when the trigger point is reached, the kanban goes back to the process where the part is produced. No need for a kanban on every container.
Most of us have used a signal kanban before. Think about the reordering of checks. They’re used in numerical order. When we arrive at the last 50 or so checks, we have a reorder form. It has all the pertinent information on it: our routing number, account number, address, and so on. The only thing we need to do is select the quantity and drop it off at the bank. This is a signal kanban too.
The Triangle Kanban is the best pull method to move from batch building based on a schedule to batch building based on consumption. This addresses the synchronization problem associated with schedules and batch building.
Figure 9.1 shows a picture of a Triangle Kanban with all the required information needed to make the part.
FIGURE 9.1
Example of a Triangle Kanban.
Batch building must be linked with one-piece flow in other industries. There are many plants that have machines that must batch build. Many times, these processes are producing to a schedule. They attempt to somehow match their batch production schedule to their customer’s one-piece flow schedule. The process may be able to make all the parts needed, but timing is the problem.
SMED can be done whether you push or pull, because it involves reducing the changeover on one machine at a time. Whether you push or pull, it would be a good thing to do. When implementing a pull system, changeover reduction should be done prior to implementing pull in batch build processes. The ability to changeover quickly and optimize lot sizes is critical in pull systems.
The Pattern: The pattern of the Toyota Template is continuous flow, pull production, and heijunka.
Culture: Continuous flow, pull production, and heijunka are critical concepts in building a lean culture. Continuous flow requires the need for multi-skilled employees. This makes the work safer and more interesting for the employees. Pull production to TAKT creates tension in the line, which in turn, increases the sense of urgency in production processes. This sets a consistent work pace for the employees. Heijunka addresses unevenness and overburden by providing a steady and smooth work environment, which is respectful of employees and physically and mentally beneficial. All three contribute to the lean culture by showing respect for the employees.
1. Liker, Jeffrey K. 2004. The Toyota Way: 4 Management Principles from the World’s Greatest Manufacturer, p. 27. New York, NY: McGraw-Hill.
2. Ohno, Taiichi. 1988. Just-In-Time for Today and Tomorrow, p. 18. Cambridge, MA: Productivity Press.
3. Ibid., p. 17.
4. Shimokawa, Koichi and Fujimoto, Takahiro. 2009. The Birth of Lean, p. 79. Cambridge, MA: The Lean Enterprise Institute.
5. Ohno, Taiichi. 1988. Just-In-Time for Today and Tomorrow, p. 9. Cambridge, MA: Productivity Press.
6. Shimokawa, Koichi and Fujimoto, Takahiro. 2009. The Birth of Lean, p. 79. Cambridge, MA: The Lean Enterprise Institute.
7. Ohno, Taiichi. 1988. Toyota Production System: Beyond Large-Scale Production, p. 122. New York, NY: Productivity Press.
8. Ibid.
9. Ohno, Taiichi. 2013. Taiichi Ohno’s Workplace Management: Special 100th Birthday Edition, p. 103. New York, NY: McGraw-Hill.
10. Ohno, Taiichi. 1988. Toyota Production System: Beyond Large-Scale Production, p. 4. New York, NY: Productivity Press.
11. Ibid., p. 17.
12. Ibid., pp. 41–42.
13. Ohno, Taiichi. 2013. Taiichi Ohno’s Workplace Management: Special 100th Birthday Edition, p. 64. New York, NY: McGraw-Hill.
14. Ohno, Taiichi. 1988. Toyota Production System: Beyond Large-Scale Production, p. 44. New York, NY: Productivity Press.
15. Ibid., p. 4.
16. Ibid., p.26.
17. Damodaran, Aswath. 2017. Margins by sector (US). New York University Stern School of Business. http://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/margin.html.
18. Shimokawa, Koichi and Fujimoto, Takahiro. 2009. The Birth of Lean, p. 84. Cambridge, MA: The Lean Enterprise Institute.
19. Ohno, Taiichi. 1988. Toyota Production System: Beyond Large-Scale Production, p. 32. New York, NY: Productivity Press.
20. Ibid., p. 27.
21. Ibid., p. 29.
22. Shimokawa, Koichi and Fujimoto, Takahiro. 2009. The Birth of Lean, p. 75. Cambridge, MA: The Lean Enterprise Institute.
23. Ohno, Taiichi. 1988. Toyota Production System: Beyond Large-Scale Production, p. 33. New York, NY: Productivity Press.
24. Ibid., p. 29.
25. Shimokawa, Koichi and Fujimoto, Takahiro. 2009. The Birth of Lean, p. 83. Cambridge, MA: The Lean Enterprise Institute.
26. Monden, Yasuhiro. 2012. Toyota Production System: An Integrated Approach to Just-in-Time, 4th edn, p. 74. Boca Raton, FL: Productivity Press.
27. Ohno, Taiichi. 1988. Toyota Production System: Beyond Large-Scale Production, p. 27. New York, NY: Productivity Press.
28. Toyota Motor Manufacturing. Toyota production system terms. Georgetown, KY: Toyota. http://toyotaky.com/terms.asp.
29. Ohno, Taiichi. 1988. Toyota Production System: Beyond Large-Scale Production, p. 127. New York, NY: Productivity Press.
30. Ibid., p. 31.
31. Ibid., p. 96.