Modeling Software Process Variability: An Empirical Study

Variability in software process models justifies tailoring them to meet the specific goals and characteristics of organizations and projects. At the present time, processes are modeled using different notations which typically do not have constructs which are appropriate for expressing process variability. To fill this gap, we have extended SPEM to vSPEM, by adding new variability constructs (such as variants and variation points). This article presents an empirical validation to check whether the variability constructs supported through vSPEM are more appropriate for modeling variant-rich processes than SPEM v.2.0, in terms of understandability of the notation, as well as of their variability mechanisms. The results show that the notation influences the understandability, both of process diagrams and of the variation mechanisms. The results also indicate that the understandability of the vSPEM variability mechanisms is higher than for SPEM. On the other hand, the understandability of process diagrams is lower with vSPEM than with SPEM. If we compare the relative results obtained with regard to understandability of diagrams and understandability of variation mechanisms of both vSPEM and SPEM, it can be observed that there is a 126.99% improvement in the understandability of vSPEM variability mechanisms. This means that, although there is a decrease in diagram understandability (34.87%), the enhancement just This is the author's version of an IEEE-copyrighted article. The definite version was published in IET Software, 2011, Vol. 5, Iss. 2, pp. 172–187 doi: 10.1049/iet-sen.2010.0020. (available at http://ieeexplore.ieee.org) 2 mentioned is 3.64 times that fall. Taking into account that variability in software processes needs support for variability in understandability such as vSPEM, it seems to be a case of a tradeoff between diagram understandability and that of variability mechanisms. The results indicate accepting that a slight decrease in understandability of the diagrams might lead to a large increase in understandability when using variability mechanisms.