Appreciation for a system is the first component of the system-of-profound-knowledge and the concept that most clearly marks Deming's evolution from statistician to systems thinker. Deming defined a system as a network of interdependent components that work together to accomplish a common aim. The key word is "interdependent" — the components of a system interact, and the behavior of the whole cannot be understood by analyzing the parts in isolation. This means that optimizing individual components does not optimize the system; in fact, sub-optimization of the parts often degrades the performance of the whole. Management's job, in Deming's formulation, is to optimize the system — to manage the interactions, not just the components. In his provocative late-career paper does-anybody-give-a-hoot-about-profit, Deming applied this systems perspective to argue that profit is an emergent property of a well-managed system, not a goal to be pursued directly.
This insight directly challenges the dominant mode of American management, which is organized around functional departments, each measured on its own metrics, each competing for resources and credit. When the purchasing department optimizes for lowest unit cost, it may destroy value for manufacturing (through variable quality) and for customer service (through inconsistent product). When the sales department optimizes for volume, it may create demand the production system cannot satisfy without sacrificing quality. Deming's Point 9 of the-14-points-for-management — "break down barriers between departments" — follows directly from appreciation for a system. Departments are components of the system; barriers between them prevent optimization of the whole.
The concept of a system's "aim" is central and often overlooked. Deming insisted that a system must have an aim, and that the aim must be stated by management. Without a stated aim, there is no system — only a collection of components. The aim transforms a collection into a system by providing the criterion against which the interactions among components can be evaluated and optimized. This connects to Point 1 of the 14 Points — "create constancy of purpose" — which is essentially the requirement to define and maintain the system's aim. It also connects to theory-of-knowledge: the aim provides the theoretical framework within which observations about the system's behavior become interpretable. Without an aim, data about the system is just noise.
Deming's systems thinking connects to a rich intellectual tradition. Russell Ackoff, a systems theorist at Wharton, was a friend and intellectual ally — both men emphasized the primacy of the system over its parts and the dangers of analytic reductionism. Jay Forrester's system dynamics work at MIT, which influenced Peter Senge's "The Fifth Discipline" and the learning organization concept, shares Deming's emphasis on feedback loops, unintended consequences, and the counterintuitive behavior of complex systems. Stafford Beer's viable system model and cybernetic management theory pursue similar themes from a different starting point. Deming's contribution to this tradition is the integration of systems thinking with statistical-process-control-and-variation-theory — the insistence that you cannot understand a system without understanding the variation in its processes, and that common-cause-vs-special-cause-variation provides the analytical framework for determining whether the system or a specific factor is responsible for observed outcomes.
In the lean manufacturing lineage, appreciation for a system manifests as value stream thinking — the practice of mapping and optimizing the entire flow of value from raw material to customer, rather than optimizing individual process steps. Taiichi Ohno's genius at Toyota was precisely this systems perspective: he saw that optimizing individual machine utilization (a component metric) led to overproduction (a system-level waste). The lean startup methodology extends this into product development: Steve Blank's customer development process and Eric Ries's Build-Measure-Learn cycle both treat the entire business model as a system to be understood and optimized, not a collection of independent functions to be managed separately. Boyd's OODA loop, similarly, treats military operations as a system where speed and coherence of the whole cycle matter more than optimization of any individual step.