Jeffrey Liker, Professor of Industrial and Operations Engineering at the University of Michigan in the USA, comments on Bill Bellows’ article on Dr Deming and systems thinking.
I greatly appreciated Bill Bellows’ both engaging and informative article about systems thinking and the role of Dr Deming. At his heart, Deming was a statistician. And it is extremely difficult to model and predict everything that will happen in a complex system.
Those who do try to model the system to predict and control have to make very unrealistic simplifying assumptions – the history of economics has much greater success at providing compelling explanations of what has already happened then predictions of what is occurring in the future. If weather systems are hard to predict, multiply the uncertainty many times when dealing with socio-technical systems.
As Bellows points out, systems can be regarded as closed or open. If we see the system as closed, that is, draw artificial boundaries and ignore what is outside, then we can make glorious predictions. When we realise systems are open and the boundaries are arbitrary, then “systems will be understood to be open, with the most important numbers both ‘unknown and unknowable’.”
This is scary stuff indeed. Normal science assumes all is knowable if only we can select and measure the right variables and model them correctly.
What evolved in Japan, which was referred to as the Deming Wheel, is plan-do-check-act (PDCA); there is debate on how it evolved. The original Shewhart cycle that Deming drew on was specification-production-inspection. Deming elaborated this to four steps: design, production, sales, research. Both models were drawn as circles, thus there was a process of continual learning.
It is said that the Japanese elaborated this to PDCA as we know it today. What I believe is most important about PDCA, however it evolved between Deming and the Japanese, is the underlying assumption that the world is unknown and unknowable. PDCA is a philosophy of learning by trying, not one of confirming carefully developed theories. The action is at least as important as the depth of analysis in the planning stage. Often we will hear Japanese sensei say something like: “Please develop your best idea and try it.” The trying, testing and reflecting is where the real learning takes place.
That brings us to systems and system boundaries. I agree that somehow, maybe because of Deming, the best Japanese companies, like Toyota, become system thinkers. They saw value streams, they saw the whole, and they saw the complex interactions with the human systems. But for the sake of action and learning by doing it is necessary to draw boundaries to try something. Kaizen can focus on an individual process, a portion of a production line, end-to-end value streams within the factory, or processes that cut across the company. From a strict systems perspective we can never know whether we are “sub-optimising” unless we consider the whole system and all of its interactions. But this is impossible to date. We lack the brainpower or modeling power to represent all these interactions.
I just happened to visit Denso’s Nishio plant in Japan where they perform die casting. They looked mainly at the technical system of going from metal to cast and then machined parts with a focus on what their chairman called 1/N machines where N had to be an integer. The improvements were expected to be ½ reduction or 1/3 or ¼, but not 3 or 4%. The core concept was to thoroughly remove muda based on equipment downsizing. The existing production process consisted of separate large machines building large batches with cranes and forklifts moving material between machines. The final result of hundreds of improvements through kaizen were synchronised lines with one-piece flow or very small batches (e.g., die cast, heat treat, machining). The mantra was compact, simple, and slim building only what is necessary. The results were dramatic, for example in one case they used one-sixth of the space, one-third of the investment, and 30% less cost.
Through relentless PDCA performed by a team that included production engineers and shop floor workers, Denso broke through the limits of conventional technology. Their equipment vendors who only knew how to build big and complex individual machines could not have accomplished this.
They were dealing with a system, though a highly bounded system that included all the major production processes for one product insider the plant. They moved line by line coming up every time with 1/N improvements.
If they had thought too hard about all the complex interactions I believe they would have stalled. It was the spirit of passionately striving to achieve a vision through relentless kaizen by the right team – highly developed production engineers and highly developed team members – that led to accomplishing the goal time after time.
I very much appreciate system thinking and the Denso case was an example of system thinking beyond what vendors of individual pieces of equipment were doing. But it still had very practical boundaries and a learning-by-doing approach that was not paralysed by the recognition of all the complex interactions that could not possibly have been modeled.