I was commissioned as a United States Marine Corps officer and later graduated as a flight officer from the Navy's flight school in Pensacola, Florida. I went on to obtain additional training and flew over 2,000 hours in the Phantom F-4 aircraft, both reconnaissance and fighter/attack models, over eight years.

The F4 was an extremely complex aircraft with tens of thousands of separate parts making up the airframe and its electrical and electronic, pneumatic, hydraulic, jet propulsion, fuel, and weapon systems. Everything in the plane was designed to support its combat or reconnaissance missions. Because of the aircraft's complexity, any number of individual component failures could negatively impact the aircraft's mission and safety. Systems design and redundancy helped minimize those issues.

I also obtained an MBA while I was in the Marine Corps, with an emphasis on finance, statistics, and management principles. This was my educational and technical background when I took my first civilian job to work in a high-tech manufacturing firm. Specifically, my assignment was to serve as the project manager overseeing the development of an advanced manufacturing facility to build then state-of-the-art printed circuit boards for a Department of Defense client. As it turned out, the diversity of my education and work with complex systems, such as the F4, became instrumental in my ability to assess the critical issues facing this start-up manufacturing facility.

One of the issues with through-hole technology is that the holes that are drilled through the circuit board take up a lot of space in the interior layers of the boards, and also constrain the location of connections points. For example, if layers one and five in an eight-layer board need a connect point, then the through-hole forming the electronic connection takes up space on the top and bottom of the board—limiting the number of electronic components that can be mounted on a circuit board, and through the other layers—which limits the space available to etch the circuitry.

To eliminate the through-holes, our engineers defined processes to build circuit boards, layer by layer, more analogous to the fabrication approach to building integrated circuits. This approach allowed the drilling and plating of holes between each discreet layer, and not through the board. Our engineers also used lasers to design and etch much smaller circuit patterns on each layer—enabling more complex circuitry—and to discretely drill much smaller holes between each new layer added to a circuit board. This strategy provided more space on the surface of each board to attach surface-mounted components.

For example, let's suppose a through-hole board has a theoretical failure rate of, say, 1 defect per 1,000 manufacturing steps—to keep the math simple—and that each board has 100 manufacturing steps. In this scenario, we should expect an average of 1 failure for every 10 boards produced; however, if our discrete-hole boards have the same number of manufacturing steps per layer, times 8 layers, then our average defect rate is now 1.25 boards per defect.

I spent my first two months working in the shop with the fabricators, across all shifts, helping to build the circuit boards. I wanted to understand the processes intrinsically. Within the first month, I began to get a very unsettled feeling that we would not be able to deliver the volume of boards required under our contract, nor within the approved budget constraints. It was clear to me that the engineers were optimizing their workflow around their processes, but were not looking at the operations of the manufacturing system as a whole. They were all very bright individuals, but their primary direction was to focus on developing the individual equipment and processes within their area of expertise.

I completed some simple queuing theory and defect-per-unit calculations, and the results were not good. We were not going to make our product numbers, and we would be late in delivery and be over budget. I would like to say that my concerns were immediately addressed, but I was the new kid on the block, and nobody wanted to hear this news from me.

Local optimization simply means that each engineer was evaluating the needs of their equipment and processes in isolation, as if they had no effect on the other elements that made up the manufacturing system as a whole. And it's a complex problem to solve. For example, my first attempt was to use relatively simple queuing theories to evaluate the flow across the development activities. Those models gave me an approximation of the issues we would face when the plant went live, but they were only approximations, and I knew the situations would, in fact, be much worse—and they were.

The problem was that the plating line could simultaneously plate many more boards than the plasma etchers could handle on delivery. Plus, the plasma etching process took longer than the plating and was often bottlenecked by other downstream processes. The initial fix was as simple as installing an oven between the two processes to keep the boards dry while sitting in-queue. We also had to add additional capacity in the plasma etchers to better match flow rates across the combined manufacturing processes.

There were other process bottlenecks that occurred as boards in the process cycled back though the same equipment that was busy working on other boards in production. The cyclical nature of the discrete-hole development process, building each circuit board one layer at a time, meant that the boards recycled continuously back through the same equipment.

The required changes were much more challenging to implement in production than it would have been had we addressed the plant design issues as a whole before we designed, purchased, and installed all the capital equipment. For example, we had to offload some of the work in our low-volume, high-tech facility to the company's high-volume shop to make room to move all our existing equipment plus new equipment to support a better flow and capacity. We had to purchase additional equipment to align capacities across the fabrication process, causing further delays in production. We had to move the equipment. We had to redesign equipment and tooling to support the flow of work further. We also had to redesign the product's substrate to use a less dangerous and less costly material.

In the end, our budget ballooned to more than three times the original planned costs to build the advanced interconnected facility. On the other hand, the efficiencies gained provided significantly more capacity than was needed to support our original client's demand, and our per-board costs were less than half of our original budgeted cost estimates.