Guide to Advanced Empirical

136 MM ller and D. Pfahl step, but also because it triggers workforce learning and product maturation. A repetition of the verification (and, as a consequence, the rework) step has multiple effects. First of all, it increases project duration. On the other hand, it speeds up the development (more precisely rework) rate due to learning. Similarly, due to learning, it reduces the defect injection (per size unit) during rework. Finally, it also decreases the defect detection rate during the subsequent verification step due to product maturation, because most of the defects have already been detected, and there are only a few defects still contained in the artefact which are harder to detect. The last two effects mentioned have a damping effect on the number of rework (and re-verification) cycles, since they both make it more probable that the number of defects detected during re-verification are below the value of model parameter Defect Threshold This is an example of negative feedback.
It should be pointed out that the causal network in Fig. 5 is only a subset of the base mechanisms that typically drive the behaviour of a software project. For example, normally one would expect an influence on development rate from defect detection (per size unit. This, and possibly other base mechanisms, have been omitted to keep the example simple and compact. For the same reason, base mechanisms related to project effort consumption have been omitted.

Download 1.5 Mb.

Share with your friends: