Gina Guillaume-Joseph, Ph.D. Candidate
Software Project failures are costly and often result in an organization losing millions of dollars due to termination of a poor quality project (Jones, 2012). Software engineering is a risky endeavor whose outcome often cannot be predetermined. Software Testing is a critical component of mature software engineering; however, project complexities make it the most challenging and costly phase of the Systems Engineering Lifecycle (SELC) (Jones, 2012).
Predictive Analytics is a data driven technology used to predict and influence the future. We develop a Predictive Model that determines failure points in the SELC and relates them to specific causal factors of testing. Our work attempts to optimize project data and information to provide informed and real-time decisions that combat financial risks incurred with failed projects.
Ewusi-Mensah, 2003 offers an empirically grounded study on software failures and proposes a framework of abandonment factors that highlight risks and uncertainties present in the SELC phases of a software project. Takagi et al, 2005 analyzed the degree of confusion of several software projects using logistic regression analysis to construct a model to characterize confused projects.
This work introduces the Project Testing Confidence Metric (PtcM) and the corresponding Predictive Model. The Model developed from data of software project failures and successes is based on a framework that identifies significant influencing failure factors and impact on the four major phases of the SELC in table 1.0. The failure factors in the testing phase have the greatest impact on software project failure as demonstrated in figure 1.0. The variables in Table 1.0 and Figure 1.1 are used to develop the Model.
The Predictive Model leverages past project performance to predict outcomes of future work. The PtcM uses that data to determine the effectiveness of testing by correlating project failure with inadequate testing to isolate those areas for improvement. The Predictive Model and the resulting PtcM provide leadership insight into determining which projects to embark upon within the project portfolio as outlined in figure 1.2.
The Predictive Model and the PtcM will assist in maturing an organization’s testing and quality assurance capabilities by implementing institutional learning. By predicting the likelihood of project failure during the early planning phase, this work will promote a more successful project portfolio. Our work helps organizations answer the question, “What will happen in the future and how can we act on this insight?”