Moving most of the employees formerly in marketing, engineering, and production groups into dedicated teams for specific products does create problems for the functional needs of each firm along the value stream, a point we will address in Part III . Similarly, the need to include employees of key component and material supply firms as dedicated members of the product team raises difficult questions of where one firm stops and the next begins, the second major topic of Part III .
ORDER -TAKING
The historic practice in the bicycle industry has been to task the Sales Department with obtaining orders from retailers. In the United States, these range from the giant mass-marketers like Wal-Mart at one extreme to thousands of tiny independent bicycle shops at the other. When the orders are fully processed—to make sure that they are internally consistent and that the buyer is credit-worthy—they are sent to the Scheduling Department in Operations or Manufacturing to work into the complex production algorithm for a firm’s many products. A shipment date is then set for communication back to Sales and on to the customer.
To check on the progress of orders, particularly in the event of late delivery, the customer calls Sales, which then calls Scheduling. When orders are really late and important customers threaten to cancel, Sales and Scheduling undertake some form of expediting by going directly into the physical production system in both the assembler firm and the supply base to move laggard orders forward. This is done by jumping them to the head of each queue in physical production.
Under the influence of the reengineering movement in the early 1990s, a number of firms integrated Sales and Scheduling into a single department so that the orders themselves can be processed much more quickly—often by one person tied in to the firm’s electronic information management system so that orders never need to be handed off, placed in waiting lines, or put down. (They now flow.) As a result, orders can be scheduled for production in a few minutes rather than the days or even weeks previously required; at the same time, order information can be transmitted electronically to suppliers. Similarly, expediting procedures are tightened up to eliminate the confusion which often arose between Sales and Scheduling.
These innovations certainly helped, but a fully implemented lean approach can go much further. In the lean enterprise, Sales and Production Scheduling are core members of the product team, in a position to plan the sales campaign as the product design is being developed and to sell with a clear eye to the capabilities of the production system so that both orders and the product can flow smoothly from sale to delivery. And because there are no stoppages in the production system and products are built to order, with only a few hours elapsed between the first operation on raw materials and shipment of the finished item, orders can be sought and accepted with a clear and precise knowledge of the system’s capabilities. There is no expediting.
A key technique in implementing this approach is the concept of takt time, 3 which precisely synchronizes the rate of production to the rate of sales to customers. For example, for a bicycle firm’s high-end titanium-framed bike, let’s assume that customers are placing orders at the rate of forty-eight per day. Let’s also assume that the bike factory works a single eight-hour shift. Dividing the number of bikes by the available hours of production tells the production time per bicycle, the takt time, which is ten minutes. (Sixty minutes in an hour divided by demand of six bikes per hour.) Obviously, the aggregate volume of orders may increase or decrease over time and takt time will need to be adjusted so that production is always precisely synchronized with demand. The point is always to define takt time precisely at a given point in time in relation to demand and to run the whole production sequence precisely to takt time.
In the lean enterprise, the production slots created by the takt time calculation—perhaps ten per hour for high-end bicycles (for a takt time of six minutes) and one per minute for low-end models (for a takt time of sixty seconds)—are clearly posted. This can be done with a simple whiteboard in the product team area at the final assembler but will probably also involve electronic displays (often called andon boards) in the assembler firm and electronic transmission for display in supplier and customer facilities as well. Complete display, so everyone can see where production stands at every moment, is an excellent example of another critical lean technique, transparency or visual control. 4 Transparency facilitates consistently producing to takt time and alerts the whole team immediately to the need either for additional orders or to think of ways to remove waste if takt time needs to be reduced to accommodate an increase in orders. 5
Raising awareness of the tight connection between sales and production also helps guard against one of the great evils of traditional selling and order-taking systems, namely the resort to bonus systems to motivate a sales force working with no real knowledge of or concern about the capabilities of the production system. These methods produce periodic surges in orders at the end of each bonus period (even though underlying demand hasn’t changed) and an occasional “order of the century” drummed up by a bonus-hungry sales staff, which the production system can’t possibly accommodate. Both lead to late deliveries and bad will from the customer. In other words, they magically generate muda.
PRODUCTION
The historic practice in the bicycle industry was to differentiate production activities by type and to create departments for each type of activity: tube cutting, tube bending, mitering, welding, washing and painting for the frame and handle bars, and final assembly of the complete bike. Over time, higher-speed machines with higher levels of automation were developed for tasks ranging from cutting and bending to welding and painting. Assembly lines were also installed to assemble a mix of high-volume models in dedicated assembly halls.
All bike makers produced a wide range of models using the same production equipment, and part fabrication tools typically ran at much higher speeds (expressed as pieces per minute) than the final assembly line. Because changing over part fabrication tools to make a different part was typically quite time-consuming, it made sense to make large batches of each part before changing over to run the next part. The typical final assembly plant layout and materials flow looked as shown in Figure 3.1 .
F IGURE 3.1: B ICYCLE P LANT L AYOUT AND F LOW
As batches of parts were created, an obvious problem arose: how to keep track of the inventory and make sure that the right parts were sent to the right operation at the right time. In the early days of the bicycle industry—an activity dating back to the 1880s and a key precursor to the auto industry—scheduling was handled by means of a master schedule and daily handwritten orders to each department to make the parts final assembly would need.
After nearly a hundred years, these manual scheduling methods were replaced in the 1970s by computerized Material Requirements Planning systems, or MRPs. A good MRP system was at least 99 percent accurate in keeping track of inventory, ordering materials, and sending instructions to each department on what to make next. As a group, these systems were a clear improvement on older manual systems for controlling batch-and-queue operations and became progressively more complex over time. Eventually capacity planning tools were added to evaluate the capacity of machines at every step in the production process and to guard against the emergence of bottlenecks and capacity constraints.
MRP, however, had a number of problems. If even one part was not properly logged into the system as it proceeded from one production stage to the next, errors began to accumulate that played havoc with the reorder “triggers” telling a department when to switch over to the next type of part. As a result, downstream manufacturing operations often had too many parts (the muda of overproduction) or too few parts to meet the production schedule (producing the muda of waiting).
A worse problem was that total lead times in batch-and-queue systems were usually quite lengthy—typically a few weeks to a few months between the point in time when the earliest upstream part was produced and the moment when a bike containing that part was shippe
d to the retailer. This would have been fine if orders had been perfectly smooth, but in fact orders received by the bike manufacturer changed all the time, partly due to the bonus-driven selling system, partly due to the substantial inventories in the retail channel, and partly due to seasonal demand patterns, particularly for low-end bikes. What’s more, there were often engineering changes in bicycle designs, even for mature products, meaning that a considerable fraction of the parts piled up alongside the value stream were suddenly either completely obsolete or in need of rework. 6
MRP systems which were very simple in concept therefore became exceedingly complex in practice. In the bicycle industry, every firm’s MRP system was supplemented by a backup system of expediters moving through the production system to move parts in urgent shortage downstream to the head of the queue in every department and at every machine. Their efforts, while essential to avoiding cancellations or large penalties on overdue orders, played havoc with the internal logic of the MRP system—often causing it to generate absurd orders—and with inventory accuracy as well. In the end, most MRP applications were better than manual systems, but they operated day to day at a level of performance far below what was theoretically possible and what had been widely expected when MRP was first introduced.
Just-in-Time, an innovation pioneered at Toyota in the 1950s and first embraced by Western firms in the early 1980s, was designed to deal with many of these problems. This technique was envisioned by Taiichi Ohno as a method for facilitating smooth flow, but JIT can only work effectively if machine changeovers are dramatically slashed so that upstream manufacturing operations produce tiny amounts of each part and then produce another tiny amount as soon as the amount already produced is summoned by the next process downstream. JIT is also helpless unless downstream production steps practice level scheduling (heijunka in Toyota-speak) to smooth out the perturbations in day-to-day order flow unrelated to actual customer demand. Otherwise, bottlenecks will quickly emerge upstream and buffers (“safety stocks”) will be introduced everywhere to prevent them .
The actual application of JIT in the bicycle industry largely ignored the need to reduce setup times and smooth the schedule. Instead, it concentrated on suppliers, making sure that they only delivered parts to the final assemblers “just in time” to meet the erratic production schedule. In practice, most suppliers did this by shipping small amounts daily or even several times a day from a vast inventory of finished goods they kept near their shipping docks. Some final assemblers even specified the existence of these safety stocks and periodically sent around their purchasing staffs to inspect them. In the end, “just in time” was little more than a once-and-for-all shift of massive amounts of work-in-process from the final assembler to the first-tier supplier and, in turn, from first-tier supplier to firms farther upstream.
To get manufactured goods to flow, the lean enterprise takes the critical concepts of JIT and level scheduling and carries them all the way to their logical conclusion by putting products into continuous flow wherever possible. For example, in the case of the bicycle plant shown in Figure 3.1 , flow thinking calls for the creation of production areas by product family, which includes every fabrication and assembly step. (Product families can be defined in various ways, but in this industry they would logically be defined by the base material used for the frame, specifically titanium, aluminum, steel, or carbon-fiber. This classification makes sense because the fabrication steps and processing techniques are quite different in each case.)
Better yet, if noise problems can be managed, the lean enterprise groups the product manager, the parts buyer, the manufacturing engineer, and the production scheduler in the team area immediately next to the actual production equipment and in close contact with the product and tool engineers in the nearby design area dedicated to that product family. The old-fashioned and destructive distinction between the office (where people work with their minds) and the plant (where people work with their hands) is eliminated.
(We’re often struck that in the old world of mass production, the factory workforce really had no need to talk to each other. They were supposed to keep their heads down and keep working and professionals rarely went near the scene of the action. So production machinery could make a lot of noise. The isolated workers simply donned their ear protection and shut out the world. In the lean enterprise, however, the workforce on the plant floor needs to talk constantly to solve production problems and implement improvements in the process. What’s more, they need to have their professional support staff right by their side and everyone needs to be able to see the status of the entire production system. Many machine builders are still oblivious to the fact that a lean machine needs to be a quiet machine. )
In the continuous-flow layout, the production steps are arranged in a sequence, usually within a single cell, and the product moves from one step to the next, one bike at a time, with no buffer of work-in-process in between, using a range of techniques generically labeled “single-piece flow.” To achieve single-piece flow in the normal situation when each product family includes many product variants—in this case, touring and mountain bike designs in a wide range of sizes—it is essential that each machine can be converted almost instantly from one product specification to the next. It’s also essential that many traditionally massive machines—paint systems being the most critical in the bike case—be “right-sized” to fit directly into the production process. This, in turn, often means using machines which are simpler, less automated, and slower (but perhaps even more accurate and “repeatable”) than traditional designs. We will look in detail in Chapter 8 at the Pratt & Whitney example of simplified blade grinding machinery that we mentioned in the Introduction.
This approach seems completely backward to traditional managers who have been told all their lives that competitive advantage in manufacture is obtained from automating, linking, and speeding up massive machinery to increase throughput and remove direct labor. It also seems like common sense that good production management involves keeping every employee busy and every machine fully utilized, to justify the capital invested in the expensive machines. What traditional managers fail to grasp is the cost of maintaining and coordinating a complicated network of high-speed machines making batches. This is the muda of complexity.
Because conventional “standard-cost” accounting systems make machine utilization and employee utilization their key performance measures while treating in-process inventories as an asset—even if no one will ever want them—it’s not surprising that managers also fail to grasp that machines rapidly making unwanted parts during 100 percent of their available hours and employees earnestly performing unneeded tasks during every available minute are only producing muda.
To get continuous-flow systems to flow for more than a minute or two at a time, every machine and every worker must be completely “capable.” That is, they must always be in proper condition to run precisely when needed and every part made must be exactly right. By design, flow systems have an everything-works-or-nothing-works quality which must be respected and anticipated. This means that the production team must be cross-skilled in every task (in case someone is absent or needed for another task) and that the machinery must be made 100 percent available and accurate through a series of techniques called Total Productive Maintenance (TPM). It also means that work must be rigorously standardized (by the work team, not by some remote industrial engineering group) and that employees and machines must be taught to monitor their own work through a series of techniques commonly called poka-yoke, or mistake-proofing, which make it impossible for even one defective part to be sent ahead to the next step. 7
A simple example of a poka-yoke is installing photo cells across the opening of each parts bin at a workstation. When a product of a given description enters the area the worker must reach into the boxes to get parts, breaking the light beam from the photo cells on each box. If the worker attempts to move the product on to the next station without obtaining the r
ight parts, a light flashes to indicate that a part has been left out.
These techniques need to be coupled with visual controls, as mentioned earlier, ranging from the 5Ss 8 (where all debris and unnecessary items are removed and every tool has a clearly marked storage place visible from the work area) to status indicators (often in the form of andon boards), and from clearly posted, up-to-date standard work charts to displays of key measurables and financial information on the costs of the process. The precise techniques will vary with the application, but the key principle does not: Everyone involved must be able to see and must understand every aspect of the operation and its status at all times.
Once the commitment is made to convert to a flow system, striking progress can be made very quickly in the initial kaikaku exercise. However, some tools (for example, massive paint booths with elaborate emission control equipment) will be unsuited for continuous-flow production and won’t be easy to modify quickly. It will be necessary to operate them for an extended period in a batch mode, with intermediate buffers of parts between the previous and the next production step. The key technique here is to think through tool changes to reduce changeover times and batch sizes to the absolute minimum that existing machinery will permit. 9 This typically can be done very quickly and almost never requires major capital investments. Indeed, if you think you need to spend large sums to convert equipment from large batches to small batches or single pieces, you don’t yet understand lean thinking.
Lean Thinking Page 7