John Dewey and Mike Rother: The Pragmatic Dreamers

In this article, I discuss Mike Rother’s book Toyota Kata and its implications.

1.

Mike Rother writes books about the Toyota Production System. Unlike his more notable peers, he did not attend the MIT Sloan School of Business. (John Roos was the founding director of the their International Automobile Program, for which Jim Womack was research director and John Krafcik was a researcher.) He never worked at Toyota. (Jeffrey Liker and John Shook did.) But he is a researcher at the University of Michigan College of Engineering, where Jeffrey Liker is a professor.  

He doesn’t write about why the Toyota Production System (also known as a lean production system) is more efficient than other production systems, or how to introduce them at your manufacturing facility, which seems strange since he consulted for manufacturers in the US and Germany for years. He writes about how people who continuously increase the leanness of their production system transform themselves into more objective, inquisitive, and team-oriented problem-solvers, people who test their assertions instead of arguing for them, who call solutions hypotheses and their implementation experiments. Scientists, basically.

Why is writing about Toyota so popular? It’s partly because from from 1937 to 2008, it operated at a profit,1 gained market share,2 and produced more reliable cars3  than American automobile manufacturers. It may also be because, for as long as researchers have been trying to learn, teach, and sell books about how to achieve a lean production system, American companies have been failing to do so.4

And they were really motivated to succeed. In 1988, The Sloan Management Review published an article5 that illustrates though some complicated math that the leanness of a facility significantly correlates with its productivity. The article, written by Jon Krafcik titled Triumph of the Lean Production System, describes Krafcik’s research of 50 facilities in the US, Europe, and Japan. It outlines the analysis that he performed.

Krafcik looked at an array of factors: location, management policies, dollar investment in automation, age of robots, age of the plant, variety of cars produced, rework area, and others. In the article, when contrasting lean from buffered facilities, he observes that the former had extra material on hand, extra personnel to tackle unexpected problems, and extra equipment on standby. He used linear regression analysis and found that facilities with less buffer were more productive. They were also more profitable.

He also found a problem with lean operations.


It is clear, too, that lean management policies have inherent risks that must be managed with a great deal of discipline and skill. From the experience of Japanese and Western producers, it appears that this risk can be largely neutralized by developing a well-trained, flexible workforce, product designs that are easy to build with high quality, and a supportive, high-performance supplier network.

The Triumph of the Lean Production System


Lean production systems have little buffer, so they are more likely to need to stop production when there is a problem. Buffered production systems keep going.

Imagine a shop floor with five work stations. The first station is upstream of the other four. The last station is downstream of the four. The center three have both upstream and downstream stations. The upstream stations feed their outputs downstream, where they are inputs. Each station takes an input, performs a process on it, and produces an output, which goes downstream. The input of the first station is raw material, delivered from another facility. The output of the final station is a deliverable, ready to be shipped.

At a buffered facility, if workstation three is having problems and producing below-average output, the upstream and downstream workstations can keep producing because at each station there is a buffer of input and adequate space to stack output. And since none of the operators is at capacity, one can can leave their station to help address the problem without their downstream process running out of input.

Now imagine a lean facility. The operators are at capacity, and there is no buffer between stations. If station 3 falls behind, station four runs out of input, and station 2 doesn’t have space to stack output, because the shop floor was designed without it. Production stops until the problem at station 3 is resolved.

This is an illustration of a lean technique called one piece flow, or 1×1. Each work station gets an input, processes it, and places it downstream, where the next workstation is ready to process it. It is also an example of a broader concept called Just-In-Time, or JIT.

(There are many techniques. Five that get discussed a lot are one piece flow, takt time, andon,  kanban, heijunka and jidoka.7  Just-In-Time or kanban is probably the most famous, though kanban is used in more than one way. It is used to refer to a visual management system using cards and columns.  The way I’m using it in this essay is similar but not quite. It does involve visual management, but a particular kind, called a visual pull system.

Imagine numerous upstream workstations that produce parts for numerous downstream workstations at a facility that produces a numerous variety of products from various combinations of these parts. For complex logistical reasons, the process isn’t able to achieve one piece flow, or even for the upstream workstations to know precisely the input requirements of the downstream stations for any given day.

The way that traditional facilities managed this was for a production scheduler to guess, and turn that guess into output requirements for each upstream workstation. Those stations would produce that amount and ‘push’ them to the downstream processes, where they stacked up if they weren’t ready for them.

The kanban technique has to parts:

  1. Build repositories [for example, bins] with level indicators that show when inventory is low and needs replenished, and when it is high, and no more should be deposited. The upstream workstations, instead of following the guessed-at schedules, produce output when the inventory in a bin drops to a low level, and they stop when they fill it back up.
  2. Observe the level of the bins. When the upstream is able to consistently maintain the level, without it getting too high or too low, move the lines down. Make it harder. Move it closer to one piece flow.

For Rother’s clients (and he has had hundreds over the years), due to long lists of stubborn and often obscure obstacles in their operations, their culture, and their equipment,  this was exasperatingly difficult. First management installs a bunch of unnecessary bins, when things worked just fine when they just stacked output on pallets on the shop floor, then, just when they start getting used to the new system (and maybe even start to enjoy the organization of the yellow bins) some dick from management comes out and fucks with the levels.  Imagine the meetings. Imagine the union complaints. After failure and frustration, they decided the technique wasn’t for them, and they went back to their traditional methods. And this is just one example of one technique. All of the lean techniques are challenging, and, as Krafcik observed, risky. But it’s not just that, they are unnecessarily challenging. It’s as if they make it harder just to make it harder.)

So, why is the lean process more productive than the buffered? And why should American manufacturers try to implement such exasperatingly difficult techniques, when it’s not even clear that they are necessary? Krafcik’s article showed correlation but not causation, and many executives in America in the latter part of the 20th century believed that either Toyota was only doing better because of macroeconomic reasons (examples: currency exchange rates, weakness of the yen against the dollar) or because of the culture, not of Toyota, but of Japan. (Krafcik addresses this last misperception at the outset of his article, and dispelling it is a theme.) Despite these misgivings, the answer, for a few decades, that researchers and books offered went something like this: they are more productive because they are more efficient. They waste less. Wasting less makes you more productive and more profitable.

And there were good reasons to believe this. Taiichi Ohno, the architect of the what would later be called the Toyota Production System was famously obsessed with waste. Besides being brilliant, he was autocratic, and he gave marching orders: discover waste, and eliminate it. He had a list of the seven wastes (muda in Japanese). One would be forgiven for perceiving the lean techniques, developed by Ohno, the waste-hater, as methods intended solely for expunging waste.

Rother’s explanation is different. He came to it piece by piece while consulting for manufacturers in the US and Germany. Over and over, he found that when manufacturers tried to run lean processes, they failed spectacularly. Production came to a halt. People got angry, and the lesson that many organizations came to was, ‘lean is not for us.’ Who knows why it works for Toyota. Maybe they’re lying. But it doesn’t work here. After hearing this enough times, and with the help of conversations with Toyota employees, Rother began to have a revelation. What makes lean facilities more productive isn’t precisely that they operate using lean techniques; they become more productive in their struggle to do so. Failure is part of the purpose of lean techniques. Not the sole purpose. The techniques are also models of simplicity, elegance and order. (This may have been partly why it was so hard to see their hidden purpose.) But the purpose that makes them worth pursuing. 

The promise that the early books seem to offer is, if you can reach or come close to this ideal state, your company will be profitable. 

Rother offers a new deal: if you persistently and intelligently work together to move towards this vision by setting a series of more and more challenging target conditions, your employees will develop a level of knowledge and creative flexibility, a mastery of their environment, that they would not have otherwise. 

It isn’t that implementing lean techniques makes a facility more productive. It is the changes in the characteristics of the team that does it. What Rother discovered after so many had failed in their search was that Toyota is playing internal games, for which it sets its own rules. It competes with itself. It is setting difficult, even impossible challenges, so that it can learn more about itself, discover and overcome its weaknesses, weaknesses which the games reveal. The obstacles that keep a part of the process (a set of work stations) from achieving the target condition (one piece flow, kanban, or one of the many other techniques, most of which involve some kind of metric that can always be tightened until it is unattainable ) are the obstacles that the team needs to work on next. Whenever a Toyota process actually achieves a challenge it has set, it sets the bar higher. Otherwise, the game would be over, and the learning would stop. 


We should think of and use the pull system as a tool to establish target conditions in our effort to keep improving toward the ideal state condition. Each state we achieve is simply the prelude to another.

This last point was made clear by remarks from two Toyota people. The first was: “The purpose of kanban is to eliminate the kanban.” While I was still pondering that one, I heard another Toyota person say: “We don’t know how you make progress without kanban.”

Toyota Kata, page 99


2.

The Triumph of Fragility
In Figure 5 of his article, perhaps the most important figure in the article because it shows the correlation between the leanness of a production system and its productivity, Krafcik makes an interesting choice. Instead of using the terms lean and buffered, as he had throughout the article, he uses the terms fragile and robust.

Why would a fragile system be more productive than a robust one?

If a manufacturing facility is a laboratory, then a fragile one would respond clearly to experimental changes. A robust laboratory would not. It would run just about the same unless you made significant effort to trip it up.

An organization needs to be a laboratory because people learn through experience. While the obvious reason to experiment on a system is to find ways to improve it, a more subtle benefit is there: the operator is learning about the nature of the system. She is learning, not just how to improve the system, but what the system is like. This gives her power that the operator of a robust system does not have. Maybe more importantly, in ingrains in her the habit and discipline of objective, scientific-thinking. (This is the primary premise of Toyota Kata.)

This was the insight of John Dewey and the Pragmatist philosophers: people learn when they test their understanding of the world by making a prediction, acting in some way that could falsify that prediction, and observing the results. If their prediction bears out, then maybe their understanding isn’t wrong. If it doesn’t, they know they need to revise it.

I acknowledge that another blogger has made basically the same argument.

3.

The Triumph of Science


To see the outcome is to know in what direction the present experience is moving, provided it move normally and soundly. The far-away point, which is of no significance to us simply as far away, becomes of huge importance the moment we take it as defining a present direction of movement. Taken in this way it is no remote and distant result to be achieved, but a guiding method in dealing with the present.

John Dewey, The Child and the Curriculum 


The philosophy of Pragmatism is inspired by the scientific method and promotes scientific thinking, but writing this today, that might be a confusing statement. I notice a tendency among my peers to treat science as the corpus of peer-reviewed research papers. The problem with that is, even if every published research article had a sound design of experiment and was reproducible, the fact that it exists and you can read it doesn’t mean that your doing so improves your understanding of the world. Acting in the belief that this new knowledge is true, and observing the results, does.


The map is not a substitute for a personal experience. The map does not take the place of an actual journey. The logically formulated material of a science or branch of learning, of a study, is no substitute for the having of individual experiences. The mathematical formula for a falling body does not take the place of personal contact and immediate individual experience with the falling thing. But the map, a summary, an arranged and orderly view of previous experiences, serves as a guide to future experience; it gives direction; it facilitates control; it economizes effort, preventing useless wandering, and pointing out the paths which lead most quickly and most certainly to a desired result. 

John Dewey, The Child and the Curriculum


A buffered environment is like a coddling parent, while a lean one not only lets you make mistakes, but arranges is so that you receive high fidelity feedback promptly.

It’s a cliche` that we learn by doing. Actually, we learn by testing our understanding. We learn by seeing if we’re right. A lean environment is conducive to this learning; a buffered one is not.

4

Rother’s genius is to take this wisdom and systematize it. In his books, he is drawing maps by which we can chart our own education. He says in a lecture (he has many available on youtube) that his primary interest is not in helping corporations, but in influencing primary education. You see these interests on his website .

He is doing both amazingly well. By drilling down into this subject, he has created material that anyone can use to change their life, because he takes the psychology of the scientific method and combines it with the psychology of habit-formation. His online training material and his books break down the concepts into practice routines we can use to develop the habits of scientific thinking.

In his book (quoted above) John Dewey presents the paradox: one needs a pattern to follow, some behavior to duplicate, so that they can walk out on their own and make their own discoveries. Pure inculcation won’t work, nor will random trial and error. Or at least, neither is efficient. Rother’s method is to train coaches, who can train ‘learners’ in the basic steps of the scientific method, and then let them (the learners) attempt to employ what they have learned.

This is all based on the Deming Cycle.

  1. You make a plan, which is a hypothesis about the way things are, the way things should be, what you should do to achieve this desired state, and a prediction about what will happen when you try.
  2. Then you run your experiment.
  3. Then you study the results.
  4. Then you reflect. You ask, what did those results tell me? Should I behave similarly in the future, or differently?

By training learners to perform what he calls Improvement Kata (a series of behaviors that promote scientific learning) Rother both teaches and doesn’t. He teaches the learner how to learn for themselves. That may sound like a cliche` too, but ask yourself, how many times have you seen it done successfully? It is painstaking to teach someone to learn for themselves, because you have to help them develop each behavior through diligent practice. You can’t just tell them how to do it. This is why Rother emphasizes the use of coaches and learners. He doesn’t claim that, by reading his books, you will be successful. He claims that, if you read his books, and engage in a coach/learner relationship in a supportive group, you might.

So, rather, he doesn’t teach how to learn. He teaches how to put yourself in a situation that will promote your drilling on the behaviors of learning. He provides not only the map, but also instructions on how use it, and instructions on how to draw your own map.

He teaches you to create your own kanban, and lower your own levels.

 

http://www.nytimes.com/2008/12/23/business/worldbusiness/23toyota.html 

2 https://www.brookings.edu/wp-content/uploads/1991/01/1991_bpeamicro_mannering.pdf

3 https://www.consumerreports.org/cro/2012/01/our-own-reliability-history/index.htm

4 https://www.lean.org/Search/Documents/352.pdf

5https://www.lean.org/downloads/MITSloan.pdf

6Buffering is used in manufacturing to compensate for variations in the production process. Changes in supply and demand would be an example of these variations. Think of buffering as a means to ensure that production lines continue running smoothly despite unforeseen factors, such as machine breakdowns, coming into play.’

7 Toyota Kata, page viii