Heijunka is the leveling of production by both volume and product mix. It does not build products according to the actual flow of customer orders, which can swing up and down wildly, but takes the total volume of orders in a period and levels them out so the same amount and mix are being made each day. The approach of TPS from the beginning was to keep batch sizes small and build what the customer (external or internal) wants. In a true one-piece flow, you can build Products A and B in the actual production sequence of customer orders (e.g., A, A, B, A, B, B, B, A, B…). The problem with building to an actual production sequence is that it causes you to build parts irregularly. So if orders on Monday are twice those on Tuesday, you must pay your employees overtime on Monday and then send them home early on Tuesday. To smooth this out, you take the actual customer demand, determine the pattern of volume and mix, and build a level schedule every day. For example, you know you are making five A’s for every five B’s. Now you can create a level production sequence of ABABAB. This is called leveled, mixed-model production, because you are mixing up production but also leveling the customer demand to a predictable sequence, which spreads out the different product types and levels volume.
In this case, a production line makes three sizes of engines—small, medium, and large. The medium engines are the big sellers, so these are made early in the week—Monday through part of Wednesday. Then there is a several-hour changeover of the line to make small engines that are made the rest of Wednesday through Friday morning. Finally, the large engines—in smallest demand—are made Friday afternoon. There are four things wrong with this unleveled schedule:
- Customers usually do not buy products predictably. The customer is buying medium and large engines throughout the week. So if the customer unexpectedly decides to buy an unusually large number of large engines early in the week, the plant is in trouble. You can get around this by holding a lot of finished goods inventory of all engines, but this leads to a high cost of inventory, with all its attendant costs.
- There is a risk of unsold goods. If the plant does not sell all its medium engines built up Monday to Wednesday, it must keep them in inventory.
- The use of resources is unbalanced. Most likely, there are different labor requirements for these different-sized engines, with the largest engines taking the most labor time. So the plant needs a medium amount of labor early in the week, then less labor in the middle of the week, and then a lot of labor at the end of the week. There is potentially a lot of muda and mura.
- Placing an uneven demand on upstream processes. This is perhaps the most serious problem. Since the plant is purchasing different parts for the three types of engines, it will be asking its suppliers to send certain parts Monday through Wednesday and different parts for the rest of the week. Experience tells us that customer demand always changes and the engine plant will be unable to stick to the schedule anyway. Most likely there will be some big shifts in the model mix, e.g., unexpected rush order of large engines and the need to focus on making those for a whole week. The supplier will need to be prepared for the worst possible scenario and will need to keep at least one week’s worth of all parts for all three engine types. And something called the “bullwhip effect” will multiply this behavior backward through the supply chain. Think of the small force in your wrist creating a huge and destructive force at the end of the whip. Similarly, a small change in the schedule of the engine assembly plant will result in ever-increasing inventory banks at each stage of the supply chain as you move backward from the end customer.
In a batch-processing mode, the goal is to achieve economies of scale for each individual piece of equipment. Changing over tools to alternate between making product A and product B seems wasteful because parts are not being produced during the changeover time. You are also paying the equipment operator while the machine is being changed over. So the logical solution is to build large batches of product A before changing over to product B. But this approach does not allow for heijunka.
In the case of the motors, the plant did a careful analysis and discovered the long time to changeover the line was due to moving in and out parts and tools for the larger engine and moving in and out new parts and tools for the smaller engine. There were also different-sized pallets for the different engines. The solution was to bring a small amount of all the parts on flow racks to the operator on the line. The tools needed for all three engines were mounted over the production line. It was also necessary to create a flexible pallet that could hold any size engine. This eliminated the equipment changeover completely, allowing the plant to build the engines in any order it wanted on a mixed-model assembly line. It could then make a repeating (level) sequence of all three engine sizes, so it matched the mix of parts ordered by the customer (see Figure 10-3). There were four benefits of leveling the schedule:
- Reduced risk of unsold goods. If the plant makes only what the customer orders, it doesn’t have to worry about eating the costs of owning and storing inventory.
- Balanced use of labor and machines. The plant can create standardized work and level out production by taking into account that some engines will require less work and others will require more work. As long as a big engine that takes extra work is not followed by another big engine, the workers can handle it. Once the plant takes this into account and keeps the schedule level, it can have a balanced and manageable workload over the day.
- Smoothed demand on upstream processes and the plant’s suppliers. If the plant uses a just-in-time system for upstream processes and the suppliers deliver multiple times in a day, the suppliers will get a stable and level set of orders. This will allow them to reduce inventory and then pass some savings on to the customer so that everyone gets the benefits of leveling.
- None of this would have been possible if the plant hadn’t found a way to eliminate the setup time for changeover.
Though it may seem unrealistic that you could do this in every circumstance, several decades ago Shigeo Shingo proved in his time studies that this was exactly what you had to do. Shingo was not a Toyota employee, but worked closely with Toyota. He was a meticulous industrial engineer who paid attention to every microscopic reach and grasp of the worker. In the Toyota style, he thoroughly analyzed the setup process for large stamping presses and discovered that most of the work performed fell into one of two categories: it was muda or it was something that could be done while the press was still making parts. He called the second category “external setup,” as opposed to “internal setup,” which was work that had to be done while the press was shut down.
In traditional mass production, the first thing the setup teams did when they performed the changeover of a production line from one model to another was to shut down the press. Shingo wondered how much of the changeover he could perform while the press was still running, so he organized an operator’s workplace for that purpose and made other technical improvements until there was no more setup the operator could do while the press was running. Things like getting the next die and tools, preheating the die, and setting it in place beside the press were external and could be done while the press was making parts. When he finally shut down the press, all that was left to do was basically to swap the dies and start it up again. Amazingly, these several-hundred-ton presses that previously took many hours to change over could, it turned out, be changed over in minutes. Think of it like a racing pit crew that quickly services and gets the car back on the track, often in less than a minute.
Over the years changeover has become a kind of a sport in Japan, a manufacturing equivalent of an American rodeo. On one trip I took to Japan in the 1980s, I visited a Mazda supplier of stamped door panels whose team had recently won a prize in a national competition for changing over a several-hundred-ton press in 52 seconds.