**5.3 Dependencies of metrics**

It is beneficial to categorize system metrics into dependent and independent metrics. There should be clear precedence relationships among the dependent metrics in terms of their logical ordering of causes, events or development activities. If necessary, such dependencies can be artificially manipulated by associating one or more attributes such as priority level, task complexity, urgency, service queue re-ordering, etc. For example, priority of a certain activity can be put to a halt if a new higher priority activity arrives in the service queue. In general, we can establish a set of primary measures from which other attributes could be derived. That is, having recorded the primary measures associated with a given process, one could reconstruct evolutionary behavior depicted by them and by secondary measures where the later could be obtained by some combinations of the primary measures (Ramil & Lehman, 1999).

Dynamics of System Evolution 37

powerful decision tool. Naturally, real data will have variations. Efforts need to be given to understand the causes of variations in the collected data sets. Since any positive process improvement changes deliver better positive results the objective here is to improve the

Newly added desirable features requested by system users or added by the developer team should nominally add value to the system. Therefore, such activities raise the level of overall satisfaction in user expectations. Evaluation of the track history of development activities within a system usage cycle can be captured through the changing rates of accumulating task completions compared with the cumulative system values. These cumulative system values may a composite estimation that is based on the accumulation of changes in

Fig. 4. Identifying possible state transitions in various systems from intersecting points

within a yearly cycle

process and thereby maintain system value at higher level.

identified USIs, superimposed with system access rates, for example
