Letís imagine youíve been presented with a problem about a system of some kind. For example, maybe how to fix a broken lawnmower (where the lawnmower is being viewed as a network of relationships between mechanical components and electricity), how to categorise the information on a website (a network of relationships between memes), or how to stop acid rain (a network of relationships between rain, energy consumption, sulphur emissions, global climate, politics, and so forth).
How do you solve the problem?
In thinking about a solution, most people "run home to Descartes" and adopt some form of analysis. That is, they will attempt to break the problem they are facing up into smaller (and by implication) easier to understand fragments. A designer, for example, might separate the problem of how to communicate a client brand into smaller parts of identity, typeface, logo, palette and persona. A programmer on the other hand - a natural born analyst if ever there was such a thing - might process model. A salesman might resort to marget segmentation techniques based upon known demographics.
In this Cartesian-Newtonian world, the aspiration of the "analyst" - a mantle that all these three people are taking on here - is to render a system, a complex relationship between things (e.g. a company's commercial success, or the subtleties of meterological cycles) into an arbitrary set of simpler parts (in the case of a company, those simpler parts may be measures of its profitability, its operating costs, its sales activity and so on), which is then supposed to gain the analyst a better understanding of that complex system.
Do you see the problem? The key failing in any kind of analysis is that the very thing that defines a system, and the very thing that the analyst seeks to understand - and that's any system, be it a business process, a social group or a biological cycle - is the relationships between the parts. And by disassembling a system into its parts - to gain an understanding of the system - you destroy the relationships between those parts, and hence the very thing you're trying to understand.
Remember, the great shock of twentieth century science is that systems cannot be understood by analysis.