Create Connected Process Flow

Table of Contents

One-Piece Flow Is the Ideal

Taiichi Ohno taught us that one-piece flow is the ideal. In school when you have the right answer for the test you get an A. The right answer is one-piece flow. So just go out and implement one-piece flow and you are doing lean. What could be easier? In fact, Ohno also taught that achieving one-piece flow is extremely difficult and, in fact, not always even practical; he said:

In 1947 we arranged machines in parallel lines or in an L-shape and tried having one worker operate three or four machines along the processing route. We encountered strong resistance among the production workers, however, even though there was no increase in work or hours. Our craftsmen did not like the new arrangement requiring them to function as multiskilled operators. . . .

Furthermore, our efforts revealed various problems. As these problems became clearer, they showed me the direction to continue moving in. Although young and eager to push, I decided not to press for quick, drastic changes, but to be patient.

Ohno learned to be patient and deliberate about reducing waste while moving in the direction of one-piece flow, also called “continuous flow.” Products that move continuously through the processing steps with minimal waiting time in between, and the shortest distance traveled, will be produced with the highest efficiency. Flowing reduces throughput time, which shortens the cost to cash cycle and can lead to quality improvements. But Ohno learned that one-piece flow is fragile.

Sustaining continuous flow also serves to surface any problem that would inhibit that flow. In essence, the creation of flow forces the correction of problems, resulting in reduced waste. We often use the analogy of a ship on a sea filled with dangerous rocks. As long as the rocks, like problems, are covered with water, like inventory, it’s smooth sailing. But if the water level is lowered, the ship can quickly be demolished by running into the rocks. In most operations there are boulders hovering just under the surface, so naturally we keep enough inventory to hide the problems.

Ohno discovered that if he reduced the inventory, the problems surfaced, and people were forced to solve them or the system was forced to stop producing. This was a good thing, as long as the damage was not too severe and the people had the capability to improve the process so that the problems did not recur. He also learned that the system needed some minimal level of stability, or the reduction of inventory would just result in a loss of production, as we saw in Chapter 4.

Connecting two or more processes into a continuous flow will increase the severity of any problems and necessitate their elimination. Connected flow across the enterprise means that production in the entire facility—and perhaps across multiple facilities—will be shut down if the problems are not corrected effectively. Imagine the importance of equipment readiness, manpower availability, and material supply when thousands of people all stop working if there is a failure! At Toyota this occurs from time to time. The entire operation is connected, and so within a few hours a problem with a main component will halt the entire facility.

Many organizations believe that this type of production stoppage is unacceptable. Stopping production is a sure ticket to the unemployment office. But Toyota sees it as an opportunity to identify a weakness within the system, to attack the weakness, and to strengthen the overall system. It is this counterintuitive thinking that perplexes bottom-line thinkers. The Toyota Way suggests that “failing” and correcting the shortcoming is a way to improve results for the long term.

Traditional thinking, in contrast, is that success is achieved by never allowing “failure” to affect the short-term result.

That said, the objective is not to entirely jeopardize performance. It is wise to prepare for flow by eliminating major issues, and to move with careful intent and understanding, beginning with planning, and developing the discipline for solving problems. As the process improves, and develops capability, the control parameters are compressed during the leveling phase to surface the next layer of issues in an ongoing cycle of continuous improvement.

Why Flow?

Most often the failure of implementation stems from a misguided belief that success is rooted in the application of lean tools (such as setting up the cell). We often tour clients through lean plants, in some cases Toyota plants, and it’s interesting to hear what they get out of the tour. They have overall impressions of cleanliness, orderliness, precision, and people engaged by their work. But their eyes light up when they see something they can directly apply in their plants.

One time, someone noted how a lean plant kept small cabinets of expendable materials by each work cell and the cell leader signed out materials as needed. A kanban system was used to replenish things like plastic gloves. The “industrial tourist” was excited about going back and setting up a similar system for expendable materials in his plant. Unfortunately, he had noticed only one specific tool, and failed to see the interconnectedness and interdependence of all the various elements. Successful creation of lean processes is derived from a deep understanding of how each tool is utilized to accomplish an end objective. A trained mechanic does not bring a wrench to the car and then find a nut to loosen. He first determines the nature of the problem, what will need to be done to correct it, and then selects the appropriate tools to complete the job.

Yet we often see organizations place the tool before the understanding. “We are going to implement visual control,” managers say, as if it were an individual piece of a jigsaw puzzle to be added. A key to long-term success is a combined effort that includes understanding the primary philosophy or concept, an effective strategy that necessitates the concept (it must become mandatory), a methodology for applying the concept, lean tools that support the method, and an effective way to measure the overall result.

We find it helpful to think about the relationship between one-piece flow and waste reduction in the context of a broader model as shown in Figure 5-1. Rather than leap into implementing tools for flow and pull, step back and understand the purpose. This model emphasizes the relationship between the primary principle of lean—the identification and elimination of waste—and the method for achieving that objective—reducing batch size to move toward continuous flow. The creation of continuous flow is often thought to be a primary objective when creating a lean process, but in reality, the creation of continuous flow is designed to drive waste from any operation: Waste elimination is the primary objective.

When material and information flow continuously, there is less waste in the operation. This is true by definition. If there were a lot of waste, material and information would not be flowing. However, there is something more profound happening here. Maintaining continuous flow between processes will create a linkage, making each process dependent on the other. This interdependency and the relatively small amount of buffering make any condition that interrupts the flow more critical.

Anyone who has attempted to implement one-piece flow (a difficult task indeed!) understands that heightening the level of problems can be of great benefit . . . or of great harm. If effective systems are not in place to support the operation, the severity of problems will surely spell doom. This is the time when lean tools must be applied to provide the necessary structure to ensure success rather than failure. The lean tools can help by providing both support systems and control methods to react appropriately to the problems that surface.

Figure 5-1. Waste reduction model

Less Is More: Reduce Waste by Controlling Overproduction

In a true one-piece flow, each operation only builds what the next operation needs. If the next operation gets backed up for some reason, then precedingoperations actually stop. It seems that nothing can be more uncomfortable in a traditional manufacturing operation than stopping. Yet the alternative to stopping is overproducing—producing more, sooner, or in greater quantity than the next operation requires. Toyota considers overproduction to be the worst of the seven types of waste because it leads to the other six types of waste (inventory, movement, handling, hidden defects, etc.). This is the key to understanding how less can be more (less means fewer parts produced in some individual steps in the process, more means getting more value-added activity done from the overall process). The case example below explains a typical situation of overproduction that reduced the ability to meet the customer requirement.

Case Example: Control Overproduction to Improve Operational Availability

While standing in the circle and observing a fabrication line, it was clear that overproduction was rampant. The line was filled with product, much of it stacked two and three layers deep. The workers were all busy, but we could see that the operators overproducing were engaged in “busy work” such as stacking and positioning the excess product.

Operators typically reached a point when no additional work would fit on the line, and then excess time was spent care-tending the overproduction (inventory). Cycle time comparisons to takt time revealed— no surprise—that these operations were below the takt time and had extra time available. Since they were not provided with additional valueadding tasks, the operators filled their extra time by overproducing and care tending.

Observation also showed that the process downstream of the overproduction (the customer) had to spend additional time moving and unstacking the product that was poorly presented in large batches. The cycle time of this operation was at takt time, but with the additional work required to move and unstack product, the total time actually exceeded the takt time. It could not achieve customer demand during scheduled work hours. In this case, the supplier process created the excess waste, but the negative effect was realized at the customer process.

We asked the operators at the initial operations to stop, and to stand doing nothing, rather than to continue producing when the next process had more than enough material to work with. It is, of course, very uncomfortable for operators to do nothing because they’ve been conditioned by management to “keep busy.” Toyota stresses the importance of this concept because it allows everyone to see and understand the amount of opportunity available. Everyone can see the idle time because it is not being clouded by busy work (overproduction).

By having these operators do less (make fewer parts), the customer operations also had less wasted time and were able to convert that time to more production. The total output of the entire operation increased significantly by simply controlling overproduction.

Of course, we were not satisfied to have operators standing around with idle time—the waste of waiting. The next step was to determine how to eliminate additional waste from these operations, and to combine operations and achieve “full work.” For this task standardized work analysis similar to the example described in Chapter 4 was used.

Case Example: Making Aircraft Repair Flow at Jacksonville Naval Air Depot

Repair operations have even more variability than manufacturing. Until you break into the equipment, you don’t exactly know what the problem is or how long it will take. So repair is often treated as a craft process: Get a team of expert repair persons to work on each piece of equipment. It is a return to the old days of the Model T, when a team of craftsmen stood around a stand and built the car in place.

The U.S. Department of Defense does a tremendous amount of repair and overhaul of ships, submarines, tanks, weapon systems, and aircraft. These are very large things. There is almost always urgency getting a plane out. A fighter plane being repaired in a hangar is one less plane available for combat.

The largest employer in Jacksonville, Florida, is a Naval Air Depot where aircraft is repaired for the Navy. Aircraft need to be completely overhauled at periodic intervals, and some aircraft have serious weaknesses that require specific repairs. Because of the urgency of getting planes overhauled, repaired, and back in service, when a plane comes in, it’s brought into a hanger, and skilled personnel attack it, taking it apart. Each plane sits in position and is dismantled, parts are repaired or replaced, everything is tested piece by piece, and it is finally reassembled and flown back into the field. Another motivation to get to work on the plane immediately is to get paid. The base gets paid based on charging hours for working on planes.

While the base had decades of experience repairing aircraft, the pressure to reduce the time aircraft spend on the ground was intense. In some cases aircraft are discontinued, and there are then a limited number available in service. If the planes spend too much time in the repair hangar, there won’t be enough to fly the scheduled missions. A program called

“Air Speed” was started at headquarters to speed up the process of repairing aircraft at NAVAIR facilities.

Two aircraft repaired at Jacksonville were the F18 and the P3 fighters, worked on in different hangers. Lean manufacturing experts were hired as consultants to lead internal lean teams and develop internal expertise. Independently, they analyzed the current situation for the P3 and F18. Their conclusions were the same:

  • Each plane was treated as a unique project, with craftsmen working in place, in no particular standardized
  • The work area around the plane was disorganized with tools and parts lying every which
  • Repair people spent an inordinate amount of time walking to get tools and parts and indirect
  • When the plane was disassembled, parts were tossed into boxes that were sent to storage (e.g., an automated storage and retrieval system), and then when the parts were brought out for reassembly, much time was spent sorting through boxes, looking for parts. Parts were often missing because they were "robbed" to work on another
  • Many planes were being worked on at once, and when they got stuck on one for some reason (e.g., needed key parts), they shifted to work on
  • There was a belief that the planes came in for repair unpredictably and that it was impossible to plan for a stable, leveled amount of

Value stream mapping revealed a huge amount of waste in the current processes. Future state maps were developed and similar solutions were presented for all the aircraft:

  • The process of disassembly, inspection, repair, and reassembly needed to be separated into distinct
  • A flow line needed to be set up with planes at different stations, and specific work done at each
  • The line then needed to be balanced to a takt Analysis of actual data showed the arrival of planes was far more stable than previously believed.
  • Standardized work needed to be developed at each
  • 5S was needed to stabilize the process and reduce much of the nonvalue-added walking and getting
  • A “hospital” position was needed so that if the workers got stuck on one of the planes (e.g., waiting for a long-lead-time part), the plane could be set aside in the hospital and the flow would not
  • Management needed to be educated in the process and stop the practice of bringing in additional aircraft whenever one arrived. They needed to control the work in process limiting aircraft to the number of stations in the flow lines (discussed later).

The work areas were laid out into workstations. There was a technical challenge in moving the plane from station to station. At some point the plane was taken apart and the center barrel and wings were removed, along with the wheels. The F18 was a new aircraft for the base, and they were able to purchase a system that held the plane together on a big fixture on wheels so it could be moved from position to position. This was not the case with the P3, so in its case a decision was made to use a “virtual flow line.” That is, teams of repair persons would come to each aircraft at fixed intervals of time to perform a stage of work.

This meant they would have to bring in the tools and materials needed for each phase of the process.

Kaizen workshops were used to set up each piece of the overall system. There were 5S workshops to lay out the area, find places for everything, and label standard positions. There were material flow workshops to take parts off the plane and put them into “shadow boxes” or kits, so when they were brought back for reassembly they were organized.

Hazardous materials were set out on carts in kits. All the kits and parts and materials were set up on pull systems to be replenished as they were utilized. The slow and complex process of analyzing each procedure in detail to develop standardized work was started so that each station could be aligned with the takt time.

The P3 is an older plane soon to be retired. The Navy decided to reduce the available planes in the fleet by over 50, from 200 to 150, yet wanted a constant number in the field (about 120). This required less time tied up in maintenance to keep the planes needed in the fleet available. Due to some fuel tank and structural integrity problems associated with aging, additional stress testing and repair requirements were added, increasing the pressure—doing more in less time. In short, from the Navy’s perspective this was a crisis, and from a lean perspective an ideal opportunity to show the value of waste elimination.

Repairing these aircraft prior to the additional testing and repair requirements took 247 calendar days. To meet the 120 planes needed in the field at all times required a reduction in turnaround to 173 days, a 30 percent improvement.

In April 2004 the lean activities formally started under the direction of an experienced lean consultant.1 After value stream mapping and numerous kaizen events, significant results were already evident by February 2005, less than one year later, as can be seen in the table below.

Pre-Lean (4/04) Post Lean (2/05)
Planes in hanger (WIP) 10 planes 8 planes
Takt time Nonexistent 15 days
Lead time when takt achieved 120 days
Actual lead time (calendar days) 247 days 200 days (on track for target of 173 days)
Additional Results Reduced cost and manpower

Setting up the process was one thing. Managing it was another. It required a different approach to management than the current leaders were used to. While there were many different things to manage—5S, standardized work, problem resolution processes, etc.—one of the toughest challenges was fighting the urge to bring in more aircraft. The flow concept was based on a fixed amount of WIP (work in process).

That is, there were a certain number of positions and a hospital, and there should be no other aircraft in the hangar. When one plane was finished and taken out of the hanger, one more could be brought in.

This was counter to just about every instinct of the leaders and counter to the measurement system. First, they believed if they left a plane outside, it would take longer to get it fixed. The lean project in fact had shown the opposite—lead time could be reduced in a major way by working on a specific number of aircraft and leaving any additional outside of the hanger until there was a place opened up at the beginning of the line. Second, there were times when people were not busy working on the planes, since all the work that needed to get done was done on the aircraft in process. This was feared because the leaders were judged based on charging direct labor hours, which also justified having indirect labor in the hangar. At various times when a new plane came in, some higher level leader would at first order the plane to be taken into the shop. The lean consultants had to use their influence to get the plane taken back out. It was clearly a major cultural clash.

The results were quite astounding to the Navy. The Jacksonville base quickly became a preferred tour site for personnel from the Navy, Naval Air Depots, Air Force, and others to see real lean in action. Jacksonville was emerging as a benchmark. Perhaps most dramatic was to see planes being repaired in assembly-line fashion. Setting up a flow line with a takt time drove tremendous continuous improvement to eliminate waste and balance the line. Stability and control immediately began to replace chaos and disorganization.

Strategies to Create Connected Process Flow

Table 5-1, below, shows the strategies that guide the creation of connected process flow, as well as the primary and secondary lean tools often utilized. The same tools that were used during the stability phase may be used (continually refining the result), as well as additional tools, depending on the circumstances of the operation. The objectives and strategies, however, always apply.

Single-Piece Flow

This is the epitome of flow, and in fact the move toward single-piece flow has reached fad status, with many companies failing in their attempts to reach this level. Achieving single-piece flow is extremely difficult and requires a highly refined process and very specific conditions. It will not ever be possible in many situations, and in many others several iterations through the continuous improvement spiral would be required before attaining this level of capability.

Strategies Primary Lean Tools Secondary Lean Tools
• Continued elimination of waste

• Force problems to surface

• Make problems uncomfortable

• Establish connected processes to create interdependency

• Identify weak links in the flow and strengthen them

• Workplace/Cell design

• Pull techniques

• Clearly defined customer/supplier relationships

• Visual controls

• Kanban

• Kanban boards

• Supermarkets

• FIFO lanes

• Problem solving

Table 5-1. Strategies and Tools Used in Creating Connected Process Flow

As an analogy, imagine a bucket brigade line where the bucket is passed from person to person one at a time. The ultimate single-piece flow would allow the passing of a single piece from one member directly to the next. This would require perfect synchronicity between all members of the brigade. After handing off one bucket to the following member, a turn is made to the previous member to retrieve another bucket. Unless the timing between the two members is absolutely the same, one of the members will wait on the other, which is a form of waste. This level of precision would be exceptionally difficult, and only possible in cases where the cycle time balance is perfect. Any slight falter or misstep by one person on the line would throw off all the others, and the house could burn down in the meantime.

In most manufacturing operations utilizing one-piece flow, a single piece is placed between the workstations, allowing for minor variance in each worker’s cycle time without causing waiting time. Even at this level, the cycle time balance between operations needs to be exceptionally high. Additional pieces between each operation allow for greater variation in cycle times from operation to operation; however, this also increases the waste of overproduction. This is the conundrum. Decrease the buffer between operations to reduce overproduction, and increase the losses due to imbalanced work times.

There is a happy medium as you move forward with the creation of lean processes. That medium point will provide a certain degree of urgency for problems, so they’re not ignored, and also a degree of cushion until the capability of the operation is improved and a tighter level can be sustained. The continuous improvement spiral model outlined in this section moves this cycle forward.

The incremental leveling phase will require a reduction in buffer quantities throughout the flow stream, thus driving ever-smaller problems to the surface, where they demand attention. This will create new instability, and the cycle spirals toward a tighter level of performance.

When Is a Problem Not a Problem?

Within Toyota, leaders are conditioned to not only stop and fix problems, but also to continuously be on the lookout for problems before they occur. A well-established lean operation with continuous, connected flow provides signals, which give everyone an “early warning indicator” prior to complete system failure.

The ability to find problems before they occur allows leaders to take preemptive corrective action, thus averting the failure.

Note: Within Toyota, “failure” is not considered to be a “bad”thing. In fact, lack of failure is considered to be an indication that the system has too much waste. Not knowing when and where the failure will occur is an indication of a poorly designed system.

Key Criteria for Achieving Flow

As we discussed in the last chapter, foundational elements are necessary for achieving smooth flow. These key criteria are generally met during the stability phase, but bear repeating here.

  • Ensure consistent capability, which is the primary intent of the stability phase. At the very least, the level of capability should be on a daily basis. During each day the operation must be capable of fulfilling the requirements of the customer.
  • Consistent capability requires consistent application and availability of resources—people, materials, and The inconsistent availability of these resources is the primary reason that flow is unsuccessful. Methods must be put in place to ensure availability of resources (not by simply adding resources, which is added cost).
  • Reliability of processes and equipment is imperative. Initially this would encompass the larger issues such as downtime, or changeover, but as the process is refined it would include lesser issues such as ease and simplicity of use.
  • Operation cycle times must be balanced (equal) to the takt time. Uneven work times will create waiting time and overproduction.

The Risk of One-Piece Flow Before Its Time

We have seen companies coming back from training classes excited about one-piece flow immediately create a cell, discover the cell is shut down most of the time, and conclude that lean does not work in the real world. They are suffering from a problem known as “rolled throughput yield.” Take the case where five machines are linked together in a one-piece flow and each machine independently breaks down 10 percent of the time—that is 90 percent uptime. In this case the uptime of the cell will be:

.95 = .9 x .9 x .9 x .9 x .9 = 59 percent uptime of the cell!

The solution: Keeping a few pieces of WIP between operations in carefully selected locations can increase this to 90 percent.

Case Study: The Danger of Single-Piece Flow for Short Cycle-Time Jobs

The move to making material flow from traditional “batch and queue” methods has become somewhat of a fad. As with most fads, they can be taken to an extreme, and negative consequences ensue. The single-piece flow “fad” has, in many cases created reduced performance results.

Single-piece flow may not be the most efficient method for short cycletime operations (30 seconds or less).

A kaizen workshop was held with the objective of establishing singlepiece flow capability in the assembly operation. The product was an assembled fitting requiring 13 seconds to complete. The takt time was determined to be 5 seconds, based on the customer demand. The work was divided among three operators, and a work cell (another fad) was created to facilitate the passing of product between operators, which is necessary for flow.

Several months later this work area was struggling to meet the customer demand, and operators had returned to batching product between operations. Observation revealed two major issues. First, as the cycle balance chart in Figure 5-2 shows, the cycle times for the operators were not well balanced.

This imbalance in work cycle times is a major reason operators begin to deviate from the “no batching” rule. When operators deviate from the original plan, it’s a strong indication that there is a flaw in the plan. Unfortunately, a struggle usually ensues as management attempts to enforce the rules of flow rather than to stop and consider where the process is flawed. Learn to see operator deviation as a positive! Stop and observe and find the real cause, which if corrected will yield a stronger process.

Figure 5-2. Original cycle balance chart for fitting assembly

If the cycle times were properly balanced and smooth flow achieved, there is another less noticeable problem. Attempting single-piece flow when the work cycle time is very short creates a high ratio of waste to value-added. Here’s why: During any work process there is inherently some amount of necessary waste, such as picking up the part and setting the part down for the next operation.

This waste can be minimized, but in the best-case scenario will still require one-half to one second for each motion (pick up, and put down). Assuming the best case, this would require a total of one second per work cycle—a half second to pick up, a half second to put down—of motion waste. If the work cycle time is five seconds total, one second for handling amounts to 20 percent of the total time! This comes to over 30 percent on a threesecond operation. That is a huge amount of inevitable waste. Yet this waste is often overlooked because of the assumption that if the material is flowing and the operators are moving continuously, it is “lean.” As we see here, that is simply not the case.

This operation would be improved by having two operators pick up a part and complete it entirely, rather than breaking the operation into multiple jobs in an attempt to create “flow.” The time would be reduced by two seconds, and the result is 11 seconds to complete (Figure 5-3). The net time per piece is 5.5 seconds (two people working simultaneously produce two parts every 11 seconds and 11 seconds divided by 2 pieces = 5.5 seconds per piece), which is 0.5 seconds over takt. The next step would be to reduce other waste and simplify the operation so it can be completed in 10 seconds or less, resulting in a net time per piece below takt time (5 seconds).

Figure 5-3. Cycle balance chart for improved fitting assembly

In this example, the creation of flow actually reduced performance by 33 percent (three operations rather than two). Also, in the scope of the entire value stream, this operation was a very small portion of the total material flow. There were much greater opportunities to create flow and reduce the throughput time in other areas by connecting operations utilizing the pull methods described below.


The terms “pull” or “pull system” are often used interchangeably with flow. It should be understood that, like flow, pull is a concept, and the two are linked, but not the same. Flow defines that state of material as it moves from process to process. Pull dictates when material is moved and who (the customer) determines that it is to be moved.

Many people are confused about the difference between the “push” method and the “pull” method. Some erroneously think they are “pulling” because the material continues to move or flow. It is possible to flow without having pull. There are three primary elements of pull that distinguish it from push:

  1. Defined. A defined agreement with specified limits pertaining to volume of product, model mix, and the sequence of model mix between the two parties (supplier and customer).
  2. Dedicated. Items that are shared between the two parties must be dedicated to them. This includes resources, locations, storage, containers, and so forth, and a common reference time (takt time).
  3. Controlled. Simple control methods, which are visually apparent and physically constraining, maintain the defined agreement.

In a push system there is no defined agreement between the supplier and the customer regarding the quantity of work to be supplied and when. The supplier works at his own pace and completes work according to his own schedule. This material is then delivered to the customer whether the customer requested it or not. Locations are not defined and dedicated, and material is placed where there is an opening. Since there is no definition, or dedication, there is no clear way to understand what to control or how to control it.

Of course, some element of control does happen through expediting, changing the schedule, and moving people, but this only leads to additional waste and variation. It could be argued as well that the agreement is defined based on the schedule. All processes are working to the “same” schedule. In fact they may be on the same schedule, but they are not on the same page.

A “pull system” is an aggregation of several elements that support the process of pulling. The kanban “sign” is one of the tools used as part of a pull system. The kanban is simply the communication method and could be a card, an empty space, a cart, or any other signaling method for the customer to say, “I am ready for more.” There are many other elements as well, including visual control and standardized work. If the three elements of pull are properly installed, a “connection” is formed between the supplier and customer processes. The three elements dictate the parameters of the connection and its relative strength and “tightness.”

The case example below illustrates the three distinct requirements for pull. Single-piece flow is the easiest to explain and understand, but the same principles apply for any variation whatever the situation. For example, the same principles apply to high-mix, low-volume operations, and to batching operations where the quantities between processes may be much larger. This following example is the easiest to understand, but the principles can be applied to any situation.

Case Example: Creating One-Piece Flow

Operation A supplies parts to Operation B, which supplies parts to Operation C.

Is the agreement defined and specified?

Yes. We said it was single-piece flow, so in this case the defined quantity is

implied in the name. (As we will see, implied definition is not sufficient). What is the specified agreement?

Provide one piece at a time. When is the piece provided?

When the next operation takes the previous piece (remember the bucket brigade).

Upon observation, we can determine whether the agreement is being followed. In this case we see in Figure 5-4 that Operation B is not following the agreement and has exceeded the defined limit of one piece.

How do we know this is a violation of the agreement?

Figure 5-4. Flow that is not defined

Figure 5-5. Single-piece flow with visually defined agreement

It is implied in the term “single-piece flow” that only one piece will be between operations. THIS IS NOT GOOD ENOUGH! The agreement needs to be distinct and visible to everyone.

If it is not distinct and visible, what will happen?

The agreement will not be followed, which is a deviation (creates variation) from the agreed-upon standard (we see that in establish ing pull we begin to create a structure to support the next phase— standardization).

How do we make it visual so that it is easily controlled?

Define and dedicate the space for one piece. The space is outlined with tape or paint to show that only one piece is permitted, and a sign or label is added to further clarify this (a taped square on the table is not completely clear, so a sign is added for clarification of what the square means), as shown in Figure 5-5.

In addition to the visual markings, the space could be physically limited (controlled) by allowing only enough room for a single piece. This technique is especially effective when the parts are oriented vertically and can be placed into a slot, thus controlling the quantity.

One of the primary benefits of creating flow and establishing defined agreements is that the effect of problems can now be seen easily. In the example above, if consistent deviation from the agreement occurs and the visual controls are in place, there is another problem.

When deviation is occurring, this is a clear message of an underlying problem that needs to be addressed. In this situation managers often state, “They know what they’re supposed to do, but we can’t get them to do it.” Many managers make the mistake of blaming the operator for not following the rules, and in fact the operator is compensating for a problem that needs to be corrected. Stop, and “stand in the circle” to identify what the operator is compensating for.

There are generally two reasons for this condition. The first thing to evaluate is whether the agreement is visual and easily understood by everyone; the

second is to look for additional problems that the operators feel compelled to “work around.”

The primary causes of deviation by operators are:

  1. Imbalanced work cycle times that may be due to normal variation in work content, operator skill, or machine cycle times. Typically, the person with extra time will deviate.
  2. Intermittent work stoppages due to lack of parts or (the fear of) operators leaving the work area to perform additional tasks—such as retrieving parts or performing quality checks—machine failures, or correction of defects.
  3. Intermittent work delays due to struggles with machines or fixtures, or overly difficult or complex tasks.
  4. Miscellaneous issues such as “building ahead” to “buy time” for changeover, an operator leaving the line for some reason, or to stagger break or lunchtimes, or such.

In some situations the correct course of action would be to adjust the defined quantity of WIP between operations. Single-piece flow requires perfect operation time balance, which is extremely difficult to achieve. Consider an operation that will incur natural variations in the work cycle time, such as deflashing an injectionmolded part.

The cycle time will vary slightly each time because this is largely a manual task, and no one can complete work cycles with exact precision (Olympic athletes, after all, do not run every race in the exact time every race). These minor variations may cause intermittent interruption in the flow. Operators do not like to wait with nothing to do, so they will naturally add buffer to compensate. The addition of buffer is the logical choice to compensate for minor time variation; however, the quantity to add needs to be defined as the standard. Perhaps the defined buffer to allow for the minor time variations should be two or at most three pieces.

The Value of Outside Eyes

The problem with communication is that it is hard to understand why others misunderstand what we clearly understand. The point of an agreement on a standard is for everyone to have the same understanding. One simple way to test this is to find someone who is not familiar with the work area, show her the standard, and ask her to explain the agreement. You may be surprised to discover how challenging it is to clearly communicate agreements visually!

Complex Flow Situations

If we consider a different example with a higher degree of complexity, we can see that it is a derivation of the same concepts. In this example, there are three different models of product to produce—–Models 1, 2, and 3—and we need the flexibility to produce any of the models at any time, one at a time. The layout is shown below in Figure 5-6.

Suppose Operation C is required to produce Model 2. They would remove the single piece from the defined location between Operation B and Operation C.

This provides a signal to Operation B in accordance with the agreement—an empty space serves as a signal, and the agreement is that when the customer pulls a part, it is replaced—to produce a Model 2 part. The layout would now look like Figure 5-7.

Operation B then removes part 2 between himself and Operation A, causing Operation A to respond by beginning a Model 2 part. When completed, Operation B will replenish the defined location between himself and Operation C. The layout would now look like Figure 5-8.

Again, this is a simplistic model; however, the three required conditions exist and are supported by visual methods. This basic model works well for producing high-volume or low-variety products, or for stock items. The primary advantage is the flexibility to produce any of the models at any time and to change between the models quickly.

Figure 5-6. Layout for single-piece flow with three distinct models

Figure 5-7. Layout showing pull by Operation C and signal to produce Model 2

Figure 5-8. Layout showing replenishment of part, and pull from customer

Pull in a Custom Manufacturing Environment

Because of the simple model (see Figure 5-8), which is based upon the production of the same three models of parts again and again, many people believe that pull in a high-variety or custom production environment is not possible. This is based on the incorrect assumption that when Operation C produces a specific model, they will send a “pull signal” to the preceding operation (B) to make a replacement for that same model. Operation C uses a “1” and Operation B makes a replacement version of “1.”

What if you have thousands of possible items and some may be used only once per month? In a high-variety, high-mix, or custom production situation the instruction on what to produce next (the custom order) would be given to Operation A rather than C. After completion, Operation A passes the part to Operation B. Then Operation B would work on this part, complete it, and pass it to Operation C. In this manner the work “flows through” the subsequent operations. Remember that flow and pull are not the same thing. The common assumption is that the work must be pushed to Operation B and Operation C if the instruction to produce is provided to the beginning of the line (Operation A).

Look back at the distinctions between push and pull. The first element is a defined agreement between the two parties. Is there a defined agreement between Operation A and Operation B in a custom production situation? Yes, it is still one piece of work in process. The second element requires that the location be defined in accordance with the agreement and then dedicated. The space is dedicated just as in the previous example. The third element requires a method to control the production to satisfy the agreement (the standard). How is the production controlled? It is controlled the same way—visually.

What is the difference? The only difference is in the agreement of “what the customer wants.” In this case, the quantity is the same, but what about the model? The customer processes (B and C) do not dictate the specific model produced by their supplier. The agreement is that each operation produces the next product in the same sequence presented by the preceding operation. This is referred to as “sequenced pull” or “sequenced flow.”

Figure 5-9, below, shows sequenced flow production for a high product variety situation. Operation A receives the schedule, and has previously produced a Model 2, Model 1, and another Model 2; and the next item on the schedule is Model 3. Since there is an open space between Operation A and Operation B, A has permission to produce the next item on the schedule. The rules of pull are still followed in that Operation A would not produce if the space were full.

The rule states that an operation can complete the part in process if the customer space is full, but will not pass the part to the space. The part will remain in the workstation. In effect, Operation B still dictates what to do (build the next item on the schedule) and when to do it (when the space is empty). If Operation B completes the part before the signal space for Operation C is empty, the operator will hold it in the workstation and wait for a signal from Operation C to replenish the space.

Figure 5-9. Sequenced flow for high product variety production

In a high model-mix environment, the level of flexibility is limited by the lead time from the point-of-schedule introduction to the completion of the product. This is dictated by the number of operations that must be “flowed through.” Instant changes to the schedule will not yield instant changes in the output because of the flow-through time delay.

For this type of flow to work well, each operator must have the capability to produce any model that comes at any time. Often the greatest challenge in establishing sequenced flow in a custom environment is achieving a balance of operation times. Refer to the case study in the previous chapter for an example of reducing the high degree of variation often found in a custom production facility, and how better balance is achieved by defining the time requirements more narrowly.

What if there is not a perfect balance in cycle times across Operations A, B, and C? First, ask: “Can each operation consistently perform the task in less than the customer requirement time—the takt?” Second, if on average the answer is yes but because of variability, the takt time is often missed, we need to put in some buffer. The buffer does not have to be an unmanaged push system. It can be defined with a specific visual arrangement showing the number of pieces allowed, e.g., three between stations. And the principle of first in-first out (FIFO) should be used to prevent a particular part from “cutting in line.”

So we see that flow and pull work hand in hand. Establishing the three elements necessary for pull then creates defined connections between operations. These connections are important to surface and highlight problems. They create a singular process in which all operations are interdependent. This step will significantly increase the level of urgency to resolve any interruptions to flow. If a problem occurs in any operation, it will quickly affect all other operations. Working around the problem by shifting manpower or machinery, or changing the schedule, will cause additional problems throughout the entire system because all operations are linked.

Creating Pull Between Separate Operations

From this understanding of the basics of pull it is possible to design a system that will be effective in any situation. The single-piece flow model above is specifically for lineor cell-type operations where the workers pass the product down the line.

How are the basics applied in operations that are separated physically, or for operations that produce parts in batches? First of all, it is important to understand the inherent nature of an operation. Someone well trained in TPS will understand that at the current time some operations are not conducive to single-piece flow for some reason. It may be the size of the part (very large or small), a resource that is shared (has multiple suppliers and/or customers), or has a limitation in the process, such as changeover times.

For example, the stamping operations at Toyota are not currently capable of producing one fender, then changing to a hood, and then back to a fender one piece at a time. The stamping operation has multiple constraints preventing single-piece flow, and the parts are produced in “lot size” quantities. First, the size of the equipment prohibits placement next to the customer operation (the body welding department). Second, the machine (“shared resource”) produces multiple part models that are required by different customers (the fender is installed at a different location than the hood), so it is not possible to place the equipment in proximity to all customers. Also the changeover time, while it is very good, still limits the ability to make one piece, change over, make another, and change over again.

How do the basic concepts of define, dedicate, and control apply in this situation? Start with an understanding of the agreement between the supplier and the customers. Supply the correct material when requested. All operations must adhere to the basic rule: “Always satisfy the customer,” or put another way, “Never short the customer.” This is Rule 1. Always follow Rule 1! (Note the paradox of this statement. While it is the goal to always satisfy the customer we have stated previously that a process that never stops a customer operation is likely to have excessive waste built in.)

Is the agreement defined? The first step is to establish the correct amount of WIP between supplier and customer to buffer the time requirement of the supplier to changeover and also to supply the second customer. Many operations currently have loosely defined (not visual and controlled) agreements that are a good starting point for the quantity needed.

Are the locations for the storage of WIP defined? Are they dedicated, and are they clearly marked? This should include information defining the maximum allowable amount, and the minimum. The maximum serves as a visual indication that overproduction has occurred, and the minimum serves as an “early warning indicator” that a problem with supply may occur and should be investigated (find the potential problem early, before it becomes a problem). Are the containers used to transport material dedicated? In our stamping example the containers are specifically designed to hold a certain part. A fender will not fit in a hood container.

The final piece is visual awareness of the needs of the customer. If the customer process is not within visual sight distance, a mechanism must be used to provide visual awareness of the customer needs and status. The visual mechanism used to provide a signal from customer to supplier is the kanban. Traditionally when dealing with suppliers that are physically separated but close enough to send truckloads throughout the day, Toyota used a physical card as the kanban. A kanban that has been returned from the customer represents the consumption of material, and as kanbans are accumulated at the supplier, they are a visual representation of the WIP agreement. The kanbans represent an inverse of the WIP quantity. More kanbans at the supplier equals less WIP at the customer.

We do not intend to completely explain the workings of kanban here, but the principles are easily understood. The kanban is a control mechanism. It can be a space on the floor if two operations are near each other. If customer and supplier are separated by line of sight, it can be a card, or return of an empty rack, or an electronic signal. The kanban must contain information relevant to the agreement, such as the supplier and customer locations, machinery utilized, material, and of course quantity and model.

Refer back to the single-piece flow example above. How did Operation B know that Operation C needed another Model 1? Operation C removed the part, and the empty space signaled Operation B of the need to replace it. The space serves as a kanban, with the pertinent information regarding quantity and model specified by visual indicators. Any kanban system is simply a derivative of this basic concept.

An automotive seat supplier had a very elaborate “phase-gate process” for developing new products. Each phase in the process of developing a vehicle had been defined in detail. The criteria for predefined “gates” for the product design was clear, and if upon review the design did not meet all those criteria, it would not pass to the next step in the process. This process was taught to everyone so they knew what to do in the process and when to do it.

One of our associates worked with them as a consultant to develop a value stream map of the current process and discovered that it did not match the phase-gate process on paper very well (a common finding). There were constant delays causing backups in the system and no good flow. A future state vision was developed, and they went to work stabilizing subprocesses and then, somewhat crudely, connecting them together.

One of the bottlenecks in the current state was the process of producing and testing prototypes. Seats were designed, parts were ordered, and hundreds of prototypes built and tested.

When that process was mapped, it became clear that this was a classic case of batch and scheduled push (see Figure 5-10). All the seats were completely designed, including heated, not heated, bench, captain’s chairs, power, and so on. Based on the designs, parts were ordered. The parts arrived at various times from suppliers. The prototype group waited as long as they could for all the parts they needed and then started building whatever seats they could with the parts they had.

Then they released lots of seats to testing. Seats that failed testing had to be redesigned to correct the problems.

A future state map was developed. It became clear that the fundamental problem was batching. Each step in the process developed large batches and pushed to the next process. The inventory triangles in the current state diagram show the result—inventory. In the case of seat designs, it was an inventory of information—the designs—accumulating in front of parts ordering. The solution: Create a sequenced pull system. But how do you do this with an information process like engineering, where each design is unique?

Figure 5-10. Current state map of prototyping process

The answer was to schedule the releases at each step based on a “staggered release.” Don’t wait to design all the different seat varieties. Design one and release it to parts ordering so they can get started ordering parts. Get all the parts for that seat to build prototypes, and get prototypes for that one seat to the test department so they can give feedback as quickly as possible to the seat engineers.

A key tool to enable this was something they called a “pull board.” It was a simple visual management tool: a white board with key information about each of the seats in process. Each department had one. So parts ordering could see when they ordered parts, when the parts were due to arrive, whether they came in on time, as well as when the next seat design could be expected. If they started to get backed up with seat designs before getting the parts in, they could inform engineering of this. If they were ready for more, they could inform engineering of this as well.

The result was significant time reductions for this process. It was no longer a bottleneck, and feedback was faster and improved the quality of designs. Suddenly the process gained some semblance of control.

Case Example: Creating Flow in Order Processing

The creation of flow is an effective method that will provide benefits to any operation that produces a “product.” (We often think in terms of a manufactured product, but these concepts apply to anything that moves from person to person as it is being processed. It could be a purchase order, an insurance policy, or a sandwich being prepared at Subway.) In this case, the “product” was a customer order that required data entry into the computer system, modifications to the order for special customization, ordering of materials for custom work, CAD drawing work to design custom elements, and a review process.

Similar to typical manufacturing operations, each of these functions was separated into different departments, each with a specific task. The order would move from department to department, each time landing in an “in basket” pile. Elaborate systems had been developed to track dates and to ensure that orders were processed FIFO, but in reality this was not the case. Some orders were more complex, requiring more time, and others were simpler jobs and “finish up” jobs that needed to move more quickly because they were related to completing jobs that had already been shipped to customers. The result was long lead times for order processing, which left little time for manufacturing and also left the stress of dealing with the complexity.

As with any situation where flow is attempted, the balance of work time and content was a major challenge. Any particular job might take longer to complete in order entry than in CAD, or vice versa. Bottlenecks shifted continuously, and as a result the lead time through the process varied considerably. This problem was compounded when associates were absent from work (especially if the current order mix was such that more time was required in the department with an absent associate).

The process was first mapped, and the product was evaluated for separation into product families (value streams). The decision to split the product into families was necessitated by the need to isolate variation, as described in Chapter 4. The product was divided into three value streams based on the complexity and time required to process each order. The most complex orders with the greatest degree of variation became one value stream, and the simpler finish-up jobs with the least amount of variation became another. The final value stream (the majority of orders) included orders that were more “standard” in terms of complexity and time required.

The group realized that the associates could be aligned in “teams” to create work cells dedicated to the particular product value stream. The office was rearranged so members in the teams were sitting next to each other. This facilitated the flow of orders. The separation of orders according to complexity and time required also allowed for a defined standard number of associates for each value stream. When this is defined it is often discovered that there are “extra” people in the process. In fact they are not “extra” per se, since the time is utilized to “cover” for any variation, including absences. It is preferable to define the correct number of associates required (based on takt time and work content) for standardized work and the desired flow. If each position is standardized, it is essential that it be filled continuously! In this case the “extra” associates become team leaders, and fulfill many important functions that will be described in Chapter 10, including filling in for associates who are absent.

As explained earlier in the chapter, it was necessary to define the agreement between operations for flow, to dedicate the resources, space, and method, and to develop a control mechanism so that each value stream could achieve connected flow. An important aspect of these elements is the visual awareness of status within each cell.

After receipt, each order was identified, placed in a colored folder according to the designated value stream, and put in a queue rack. The leader was able to see the workload and make adjustments as necessary to shift some work to other value streams if the “backlog” exceeded the agreed-upon limit (the standard). Agreements were established (standards) regarding the allowance of shifting work (e.g., the simple jobs could shift to the medium value stream, but the complex jobs could not shift to the simple stream). Also, clear rules were established regarding who was allowed to make the shift. If all teams fell behind based on the defined limits, overtime was used to support the workload.

Within each team, the elements of flow were established between each operation. Because of the inherent variation in time required from order to order, a connection mechanism was needed to buffer the variation in work times, but also to support flow and surface problems. Singlepiece flow was not possible. In this case a sequenced queue (sometimes referred to as a FIFO lane) was utilized. The queue rack had a defined number of spaces to indicate the status of flow and balance between operations. The team leader monitored the queue levels and made minor adjustments within the cell (e.g., completing an order “off line” and reinserting it) to support balance. As always, these adjustments were only made when the condition exceeded the defined agreement, and after careful assessment of the situation.

For example, if the defined agreement was a maximum of five orders between team members, and the maximum level was reached, the team leader would be notified by the team member to evaluate the situation. If the team leader determined that the subsequent orders were simpler for the downstream operator (the one who is “behind”), he or she might decide to take no action. The imbalance could be temporary and correct itself on the following orders. If the product mix had complex orders downstream at the bottleneck, an automatic correction was not likely, and the team leader would make adjustments.

In addition to improving flow, the teams realized that separating orders according to complexity and difficulty provided an opportunity to train new associates on simpler jobs before progressing to more complex work. Associates from different departments became part of one team, and cross training was done to facilitate flexibility within the team. Locating operations in close proximity facilitated quicker feedback on problems as well, and the “rework” required was reduced significantly.

This group was able to create a dramatic reduction in the lead time for orders, especially the crucial “finish-up” jobs. As the business grew, the order-processing group consistently processed a much greater number of orders without the addition of associates or the need for overtime.

Traditional Batch & Queue

Ideal State of Lean

Push or


Schedule each process and push to the next

Supermarket Pull (Kanban)

Upstream process replenishes what downstream customer took


Sequenced Pull


Pull from a feeder in sequence




Defined lane with defined standard WIP

between unlinked processes in FIFO


Continuous Flow

(1 pc Flow)

Physically link process steps with no inventory between

Figure 5-11. Continuum of flow

Flow, Pull, and Eliminate Waste

The most common perception of lean is that it is about “just in time”—the right part, the right amount, the right time, the right place. As we see, there is a lot more to it. The key to eliminating waste is creating flow, and the principles of pull require the production in a “just in time” manner.

It is best to think of flow on a continuum, as shown in figure 5-11. Even the dreaded schedules create some degree of flow. At the other extreme is a onepiece flow process with no inventory between operations. Between, you can have a supermarket that is being replenished, you can pull parts in sequence from one process to the next, or you can flow through a lane with a defined amount of inventory without breaking the FIFO order. Notice that the famous kanban system in which a supermarket is replenished is not the preferred choice, but the next worst choice besides scheduling. Kanban is an admission that inventory is needed and must be managed. Waste is designed into the system. Both sequenced pull and FIFO generally require less inventory than supermarket systems and have better flow.

The main point is not that you either use one-piece flow or you’re not lean. The point is that the focus should be on waste elimination. If you have a supermarket replenishment process, take out a kanban and stress the system. If you have a FIFO lane, reduce the lane by one piece and it will force continuous improvement.

Reflect and Learn from the Process

  1. Using your current state map as a guide, walk the path of the material flow once During this walk identify processes that are inherently inflexible where continuous flow is not currently possible. Do not attempt this exercise from your office. You must see each of these processes to understand the cause and effect relationships that prevent flow.
    • Identify the inflexible processes on your map.
    • List the cause of the inflexibility for these processes, such as long setup times or shared resources that supply multiple parts or processes.
  2. Evaluate each customer-supplier relationship in the value
    • Determine whether each connection will utilize a FIFO-style connection (if flow-through is possible), or a supermarket style connection.
    • Develop a plan to define each connection in terms of what will be included, how many (define the unit of measure), and where the material will be.
    • Determine whether the space needs to be dedicated, whether containers or carts are dedicated, and whether the resources are dedicated to this connection.
    • Identify the control mechanism for each connection and how you plan to make it visual and easy to verify adherence.
  3. Good flow depends upon good balancing of cycle times along the value
    • Measure the cycle times of each operation in your value stream, and create a cycle balance chart to determine the current operation balance.
    • As you walk the value stream, identify the physical signs of work imbalance (such as waiting, or inventory accumulation), and highlight these on the current state map.
  1. The following questions apply to any operation that is a lowvolume, high-variety (custom, semicustom, or make-to-order) Your objective is the same as any other company— to create the best possible flow. In relative terms your flow may never be perfectly balanced or smooth, but it can be improved.
    • Evaluate the grouping of your product into “families” based on the work content time required at each operation (short time, medium time, and long time).
    • Is it possible to achieve a better work flow by controlling the product mix introduced into the value stream (to even out the work content time)?
    • Graph the part numbers by the quantity ordered in a year from highest to lowest volume (P-Q chart), and identify product families based on volume and frequency of Highand medium-volume parts are candidates for cells. You may also be able to use these to level the schedule (see Chapter 7).
    • In a custom production environment the defined agreement is based on an agreed-upon “unit.” What will your defined unit be? (It may be part-by-part, or order-by-order, an increment of time, or another common )
  2. The following questions pertain to nonmanufacturing processes. The result of your work may be less tangible than a manufactured product, but work is being done and there is an end The end result is your “product.”
    • Define the Identify and map the flow of the product through the various processes.
    • In nonmanufacturing processes the product is often not easy to see as it moves through the operations. It may be paperwork or information in a computer. These create unique challenges in trying to make the process visual.
      • Do you have visual awareness of the product flow (product that is “in the system” or stacked in an inbasket is not visual)?
      • If the product itself is not visual, how can you create visual awareness of its progress?