The problem of the design of a system must be stated strictly in terms of its requirements, not in terms of a solution or a class of solutions. - A. Wayne Wymore

During this period I was able to carry out only one project of real interest to me. Pure Oil was a fully integrated oil company in the sense that they engaged in exploration for oil, construction of oil wells, production of crude oil, transportation to refineries, distribution of refinery products to company owned storage facilities and to gas stations. We developed a linear programming model of the whole network hoping to discover the optimum allocation to and from connected nodes in order to meet required deliveries at minimum cost. Then we collected all the data needed by our model and ran the model. The computations took several days and the results were disappointing: The optimum allocations were not significantly different from their current practices. Is it possible that we had discovered a system of human beings in which everyone knew the payoff functions and constraints, and, over time, had evolved behavior patterns that enabled them to achieve near optimum performance? Or was it possible that our model was not detailed enough as it was based so closely on current practice that no new behavior could emerge? Or was our data incorrect? This was disappointing but great experience.

Only if mathematical rigor is adhered to, can systems problems be dealt with effectively, and so it is that the systems engineer must, at least, develop an appreciation for mathematical rigor if not also considerable mathematical competence.

[The process of system design is]... consisting of the development of a sequence of mathematical models of systems, each one more detailed than the last.