Wednesday, June 06, 2007

The Toyota Way - viewed as feedback control

Here I look at Jeffrey Liker's view of The Toyota Way, and twist it around to show that a hierarchical, nested control-loop model fits the data, which I will later show provides some insight that's not otherwise evident.

The first conclusion is the same theme I've been trying to work on - that personal health, social population health, and corporate health are not mortal enemies. They are, in fact, joined at the hip. The "loose coupling" is magnified by feedback loop compounding so much that there really is only one "mathematical" entity here on this planet, and it includes people, corporations, nations, and computer systems, as well as the supporting plant and animal life. We need to plan around that reality, or our plans will just keep on failing.

I plan to show that the Toyota Way is strong precisely because it meets that challenge and aligns the multiple levels and threads.

So, I start with Liker's diagram 1.1 from his book "The Toyota Way", above.



Then I invert it, because the rest of my model views this from the other side, and I see "causality" and "resting on" going downwards, not upwards.

That gives three axes - small to large, many to few, rapid to slow.
But, I want to swap people and processes in this diagram, because processes and problems are more related than that diagram shows.

And, it's not really problems, despite breaking the words-that-start-with-P rule, that I don't think is a natural constraint. STEPS make up Processes. I want to start with the way a HEALTHY organism looks and functions, not one with something that's broken. LATER we can deal with how "errors" at every scale, horizontal and vertical, high and low, can be detected, compared to the "me" and "not me" decision that "immune systems" make, and responded to gracefully at all levels simultaneously. (pulling the cord is just one way to do this.)


And, people make up Society. (This is consistent with Liker, and Philosophy, long-term-thinking, and social goals all sort of merge.)
So that looks somewhat better.
x
But the transitions between levels aren't symmetric. I like symmetric. I love symmetric, actually, and think Nature loves symmetric, because it's so much easier to learn and compute.
Symmetric across levels (scale-invariant) is a very popular design pattern on this planet, and it tends to work and be the easiest possible way to stack things.


so, we'll pull "people" out, for just a moment, and try "business" in that slot. Yes, that works much more smoothly up the chain. The transitions are all "composes" / "composed of" relationships now.
Then, I want to put people back in, but everywhere.


OK, Each step of this ladder or hierarchy now is something that has a local identity and local self-sustaining (homeostatic / allostatic) control loops in place. In the ball rolling on surface analogy, these each have locally defined "notches" or "grooves" where they try to keep the ball, and where it will roll "naturally" , given the dent in the terrain that the local control system has created.

But, this actually has "N-factorial" control loops. Not only does each level, taken one at a time, have a control system, but each pair of levels, (taken 2 at a time), has a control system to maintain their relationship:
x
If those control systems break, we call things "out of line" or "broken" or "Not working smoothly". We want steps to make up processes stably, and processes to work together to make up business entities, and business entities to work together (!) to form stable societies. (all of which requires people and affects people at every step - some other detailed loops I'm leaving out.)

Skipping the N=3 control loops, and N=4, etc, (departments, divisions, groups), we can get to the far end of that list, and look at the single loop that holds everything together, taken N at a time. This is the loop driven by "Pull" downward, which changes substrate and becomes "cars", which changes to "met needs" which changes to "social support and cash inflow" and return business and helpful regulatory climates and laws, which sustains both business and social and personal needs, and so is stable. We have found a "loop invariant" that works for people, teams, business units, business, societies, a nation, and a planet scale.

We can look for a moment at HOW this loop feedback is carried, and ponder the "visual" cards and "cords" used at Toyota. The key to any of it is "transparency". It should be obvious to anyone, inside or outside the unit, walking by, how they are doing. Not only are things not hidden, they are "worn on the sleeve." The internal health of the "cell" is reflected directly in what proteins are displayed proudly on its external surface for all passers by to see.


Now, if we look at the "people" aspect of this thingie, this multilevel living entity, we see that the people thread is actually multilevel too. We can see from DNA, making up cells, making up body systems,making up bodies, making up families and teams, making up communities and departments, making up businesses and societies and what public health calls "populations".
That thread has to be whole, and healthy, or this whole deal falls apart on that.

If the people are all sick, they are poor workers, poor consumers, and won't support the social structure continuing whatever it is that's wrong. If you destroy the ecosystem that supports the people, that's not a very stable long-term business model either.

Employees are people. They don't suddenly appear at 8 AM and disappear at 5 PM every day.
They have lives. If the lives don't "work", there goes your customer base and your workforce and your management team, all at once.

Finally, there are control systems at the non-DNA side of this double helix vine that grows up and around the core hierarchical trellice. That's the computer or IT side, including robots.
This has a similar ladder, from actuators to "arms" to process control systems to business control systems and customer relaiton systems to social control systems and regulatory legal systems. Again, for this to be adaptive and responsive and agile, it has to be healthy top to bottom, and all similarly chained up with a hierarchy of internal controls that reshape it, so that each of the individual components can have its own perceptions, self-control model, and action set that it can do -- and, if you add all these up, they are not only compatible, but the whole thing produces a product (cars), meets social needs, develops the people, develops the teams, develops the organization, and develops the society.
x

This last slide is an illustration that "you can't beat city hall". The long-range, large-scale "dent" in the world that higher parts or contexts create will compete and win over small-range, local "dents" that local control systems create.

If you add up the slopes, you end up with the bottom diagram, where the local dent is now tipped and "won't hold water" anymore. Nothing that happens locally can overcome that problem.

If the higher context is aligned with society and good things, this misalignment is a good thing, and will put pressure on the local structure to move its goal over to where everyone else is.

If the higher context is misaligned and the local context is good, well, that's bad news. Again, sooner or later the larger context (society) will win out, but locally the misdirected efforts will appear to be winning, even though they are continually defeating local efforts to align with the global good. Nothing can survive, long term, if the contexts above and below and around it are pulling it some other way, because, if we move up a scale, it now is the "local" entity and Society is the "global" entity. Things will just keep breaking. That is not a "solution" to the problem that will be stable, grow, thrive, and be a good investment.

Moral - Either we solve EVERYTHING or we solved nothing.
Toyota tackled everything.
Trying to pick up some part of that scheme and apply it just to manufacturing, while ignoring the rest of the "Toyota Way philosophy" won't actually ever work, according to that model.

5 comments:

Wade said...

It seems to follow from that model that "competition" is not a long-range permanent feature of our lives, any more than the cells of our body spend all day "competing" with each other for resources or power.

If, in fact, the "loose coupling" is actually "tight coupling" because we left the compounding effect of feedback loops out of the equation, then there really only is ONE living thing on this planet, planetary sized.

Also, that means that there is no such thing as a "win-lose" result.
That's like being overjoyed because you just shot a hole in the bottom of your competitions end of the lifeboat you're both on.

"Ha ha! You're sinking!" will be a short-range joy. Sooner or later, the effects will affect you.

You cannot build a stable company out of people and divisions that spend all day putting each other down and exulting in each other's "defeats" and losses. That's the Toyota Way lesson, rephrased.

If we don't let go of THAT part of our old culture, it won't matter how many kanban cords and colored cards we put up. It won't work. It won't fly. The global philosophy's tilt of the playing field will undo all the best local efforts.

Corporate and human and society and investor-wealth-building are not resting on "competition" so much as they are on a cooperative matrix of suppliers and buyers helping each other do business successfully.

Well. That's news.

Wade said...

On a mathematical level, if there are huge efforts by people to "get" or "destroy" other people, then there will not be open-ness to admitting something in one's own area is wrong, and transparency will become opacity.

The world will go dark, and everyone will be operating in the dark. This will "break" every vertical control loop, as well as defeating all the horizontal peer-to-peer stabilizing and reality-check loops.

You can't get humans to behave that way if they feel psychologically or physically endangered by admitting something around them isn't working correctly.

Golly - that gets us right back to questions of honesty, integrity, compassion, concern for others, and all the other "virtues" that I described in earlier posts as being the stuff successful corporations are made from.

Without them, workers on the bottom won't do their jobs. If they don't apply vertically to all levels, we get more Enron's and CEO's destroying entire fortunes to steal some loot for themselves.

Virtue directly affects the bottom line, long term. There's no way around it. You can motivate it with religion, or you can motivate it with science (maybe), but you cannot build a sustainable business or society or place to invest your money and life without it.

'Nuff said.

Wade said...

So, back to my T.S. Eliot favorite quote:

They dream of building systems so perfect that no one will need to be good. But the man that is will shadow the man that pretends to be."

Wade said...

Also - ecosystems may have local "competition" but they have more global cooperation between species that dominates in effect than the local competition between individual animals.

Darwin's theory did not say "Winner take all". Reducing an ecology to a single player is always fatal, sooner or later, when the environment changes and the player can no longer adapt because they have homogenized everything into some huge BORG like matrix and removed all the precious diversity.

Cooperation does not require uniformity. E pluribus unum has room for a lot of diversity locally, so long as there is agreement on the fact that we're not fighting so some states will be victorious and other states will be defeated and pillaged.

Wade said...

An immediate conclusion from this model is that a good rule of thumb would be for every intelligent agent, at any level, to do two things:
1) align itself, so far as possible, with "the big picture". The higher up in the chain you go, the stabler the GOAL will be, and you want to build your investment on solid rock, not a swamp or sand.

2) Within that constraint, so far as possible, synchronize yourself with your neighbors - to the sides, above, and below you.

Hmmm. This turns out to be exactly the algorithm I used at Parke-Davis to align biomedical images in order to build larger montages stably. If you leave out the "align with the big picture" step, you end up getting misalignments like the MC Escher Waterfall. If you keep both rules in, you get a good montage even with noisy or sparse data.

Hmm. I believe also I recently quoted a section from the Christian Bible, Matthew, where Jesus said something about there were two commandments to focus on -- Love God, and Love your neighbot as youself. And all the rest of the Law followed from that.

So, I think both the Science of nested control loops, the Business of Toyota, and the Abrahamic religions (Judaism, Christianity, Islam) should all be capable of getting on board and agreeing that these basic principles are good, at least for people.

My argument is that these laws are scale-invariant, and they also apply to teams, departments, divisions, companies, societies, cultures, and nation-states.

We have very very few models that span such a range of scale, and have looked good for over 2000 years.

They deserve closer inspection.

Wade