Let's look at any process. An employee is at his workplace performing his tasks. Workorders and reports arrive at his desk, and in the correct way he gets through them.
If one watches what he is doing, one can see things that have to do with the aims of his job, and one can see things that have to do with the way he has organised his job. Or with the 'why' and 'when' of what he's doing. This is, the work in progression in a process knows different aspects: primary aspects (resulting from the aims), organising aspects (such as the means needed for the job), and controlling aspects.

Seen from within the process, any of these aspects is a world of it's own. This can be seen very well by watching the objects relevant to the process.
A file contains all documents related to the task, and will be delegated from one employee to the other. Stock is replenished, or something is taken from a stock. Information is used and new data (facts) are registered.
As a matter of fact, any process is a co-ordinated contruct of all these kinds of object-related functions.

The distinguished aspects can be teared off from the processes, and then be grouped together into separate clusters. These clusters can then provide the processes with the needed aspects, by rendering the relevant services. Only the co-ordination of these services remains to the processes.
The clustering is best done using methods that optimise the joining of functionally strong bound objects. In this way one can realise clusters that are functionally fully independent of each other.