How to Avoid Drastic Software Process Change (using Stochastic Stability)

That is, using internal changes, the benefits of drastic change can be achieved without project disruption. For example, Figure 2 compares defect/KLOC predictions from SEESAW and the drastic changes of Figure 1. SEESAW’s
recommendations, comprising only internal changes, resulted in the smallest defect predictions (see line 1). The rest of this paper explains how Figure 2 was generated. After some preliminaries, we discuss the options available within the current structure of four sample projects. Figure 1’s drastic changes will then be implemented as operators on those projects. Next we describe SEESAW and the models it operates on: the COCOMO effort and time estimator [7, p29-57]; and the COQUALMO defect predictor [7, p254-268]. In the results that follow we show that, for our case studies, SEESAW’s search through the space of internal changes usually out-performs the drastic changes of Figure 1.

The contribution of this paper is a demonstration that managers have more options than they may realize. Stochastic stability is an insightful method for discovering useful project reconfigurations that improve a project while avoiding requiring drastic and disruptive project change.