Ted Goranson - Personal Blog

When we were running SEMATECH in its original charter, a key realization was that semiconductor companies were not in the product business they thought they were. They were in the business of making semiconductor factories (they call fabs). These cost a billion dollars a piece then ($2B now), with a huge technical challenge of getting it all integrated and the yields up. Every piece of equipment and many of the processes are wholly new, and you essentially throw it all away in 3-4 years.

They made their money selling fine products, but their business was in engineering the processes. Once we moved from product feature-centric to process feature-centric, we were able to save the industry. There are other examples of apparent different businesses.

• Google provides a dynamically generated index to information as their product, but they actually sell dynamically generated indices of consumers.
• Boeing sells planes but is in the safe, cheap transport business basically counting on others to merely operate the planes.
• GE sells aircraft engines, but they see their core competency as thermodynamic engineering and material science.
• Apple sells delightful, profoundly useful products that happen to be disruptive, but their real business is in reinventing the production system. They’ve made some improvements in production management that give them an advantage over competitors, allowing them to stream new ideas, technologies and features more quickly than the competition. Compared to what they might do (and what we promote) the changes are small, significant only because the competition is so profoundly broken.

So Horace is right, I believe when he says that cars are ready for disruption. And he is right to focus on the production system.

One great advance in production systems was the ‘modularization’ of product components so that each component could be designed and produced separately and then assembled. This notion is well over a century old. A second advance was in farming out the components to subcontractors. While this is quite old as well, it matured into what we have today only about 25 years ago.

Dividing the product into components supplied by others has two profound effects.

• Capital investment and associated risks are distributed out of the central enterprise. The effect of this cannot be understated; each supplier has to go get their own capital.
• Every supplier’s place in the enterprise is conditional based on a variety of competitive factors. This introduction of market forces into the equation is huge; in theory, every cycle thousands of companies are trying to innovate their way into hundreds of enterprise slots. So if the prime contractor has a good way of communicating value, many innovative solutions should emerge.

In practice, the enterprise has yet to fully benefit from competitive innovation and decentralized financing from this modularization. The capital financing system is decentralized but the reward system and metrics stay centralized. For example, a steering wheel supplier like ZF TRW hadsto build its production line for Ford, and hope that Ford’s cars are successful. TRW has minute control over that success and the lenders factor this in. Japanese keiretsu get around these extra capital costs by keeping the supply chain constellation and funding stable; the same investors ‘own’ Toyota and their steering wheel supplier. This also makes process integration far easier. But innovation is effectively killed.

Apple uses their capital power to fund innovation in competitive suppliers, giving huge benefits especially in materials processing and fabrication where process and equipment are tightly linked. But this is very tricky, being the funder but forcing the supplier to innovate on their own. We got some insight into this with the collapse of their sapphire film supplier.

Types of Modularization

When we think of modularization, we have several notions:

The product itself is divided into parts. An old style PeeCee was an extreme example of this. End users could literally buy components from absolutely discrete categories and put them together themselves, and could then over the life of the device insert upgraded components at will. Military aircraft are another extreme example. One reason is that the whole system is just too complicated to manage, so breaking the aircraft into components puts constraints on the dynamism of the product and process features. This ensures that the aircraft will be poorly optimized, many have concluded. (Another reason for dividing up complex military systems is that many of these systems are so profoundly expensive that to get them funded requires having components made in key Congressional districts, bringing ‘jobs.’)

The process workers are divided into companies and teams. This is usual. Different companies are arranged in a ‘supply chain’ to provide discrete parts and services. These are assembled on the way to shipping to a customer. We’ve already mentioned the potential but unrealized advantages of involving market forces and distributing capital risk.

The enterprise functions are modularized. This is quite a bit more prevalent today than in the past as specialized abstractions and computerized tools are adopted. For example strategic planners live in a different infrastructure and culture than, say, the manufacturing engineers and operations managers. Typical divisions include financial, sales, strategy, marketing, design, legal and production. The rationale is that each is essential and expected to play their role well. The divisions have hard boundaries.

Capital investment is modular. In general, money is attracted either by discrete project or discrete vehicle. An example of the former is money borrowed to finance a factory. The latter is more complex because various intermediaries can be involved because of value-sucking holes in the law. Thus, it is advantageous for a company to borrow money to buy a commercial aircraft and lease it for the life of the plane to a carrier. Another example is how loans may be bundled so that many investors are investing in many projects. This mitigates their risk but creates a long, inefficient pipeline of value monitoring. Notwithstanding, where the money touches the enterprise, there is someone fencing capital to production.

Each of these four penetrate the way things are done today, whether in manufacturing, military operations, health care, education or disaster response.

Congruence of Modularity

In today’s environment, much of this modularization is congruent, for example the breakdown of the product corresponds exactly in most cases with the breakdown in process groups (as suppliers) and an many cases, the modularization of capital. Financing internally is effectively modular and linked to process components. With the popularization of activity based costing, financing can be as fine grained as the analysts wish.

In each case and especially in cases where the modularization continues over categories, the enterprise and what it can do is constrained.

This is bad. Any dimension of the enterprise has less intrinsic agility. It also has structural inefficiencies, in part because it has to maintain so many ontologically distinct functions in each of the entities, and worry about how they interface. Modules mean interfaces.

There cannot be a global story that everyone understands and how they can add value to the story. These inefficiencies are very serious. We looked at military systems for example (because we could) and the numbers are astonishing.

An ideal would be a situation where enterprise strategists could decide when and where to have boundaries so as to pick the tradeoffs. And then once things got going, the strategist or manager or collaborative could shift, soften and harden these boundaries as they learned more about what they are doing.

If we were going to innovate in this space, we don’t necessarily have to defeat this modularization. I think we can in ways that matter by employing ‘value features,’ but that is the subject of another post. All we have to do is loosen the boundaries among the modules a bit and uncouple the modularization classes.

For example, there is no reason other than tradition that the management of capital is tied to the production metrics of the supply chain. We know how to adapt the technical and management infrastructure to relax this binding.

A more general vision would allow the decoupling of these modularizations one to another. This would make it more possible to think in whole system terms, to allow strategic managers greater flexibility in how they do things and set the stage for a new approach to value. In particular, we need a way to model it and incidentally measure it.

But this turns out to be hard, hard, hard. We spent boatloads of money at DARPA on things at the fringe of this.

However, things are different now. I think with these new foundations in geometric logics and situation theory we finally have what we need to disrupt production systems by relaxing these constraints. I’ll make it happen somehow.

Back to Dediu

He’s right that automobile manufacturing and sales is ripe for disruption, and that Tesla isn’t the one to do it. Also, I think he is dead on that the enabler here is the production system. Except for some relatively small innovations in the supply chain this hasn’t changed in 100 years. But he is on the wrong track with the solution being modularization. It is the problem, not the solution. His misdirection is understandable because you really have to have been working in this area to even see what it is. So much of it is invisible to the consumer.

The disruptive production systems of the future will have universal semantics that mitigate divisions.