Essays in empirical software engineering processes, contracts and metrics programs

This dissertation consists of three independent essays in empirical software engineering. The first essay examines the effects of communication mechanisms and software processes on project performance in offshore development projects. Offshore software development is a relatively new form of software outsourcing in which the software vendor and client are in different countries. The actual software is developed in the vendor country and then shipped to the client organization. We study the impact of communication mechanisms between the two parties such as project status meetings on project performance. We also examine the influence of software processes instituted in the vendor organization on project performance. Performance is measured by three variables—effort, cycle time and rework. The analysis is carried out on 34 projects collected from two offshore developers in India. The results show that processes and communication are strong predictors of performance. We also show evidence for a triangular relationship between the three performance variables studied. The second essay studies the choice of contracts in offshore development projects. The contract choice is hypothesized to be a function of information known to the contracting parties. We use prior theory from incomplete contracts and co-operative game theory to formulate hypotheses about the effect of this information on contract choice. Contracts in offshore development are of two types—Fixed Price and Time and Materials. We test out the hypotheses on data on 93 projects collected from an offshore developer. Subsequently, we also test for the effect of contract choice and other software engineering variables such as duration and employee attrition on project profits. The results support our model of contract choice and also indicate that the contract choice is not efficient with respect to information known during contracting. The third essay examines the success of metrics programs in software organizations. Software metrics programs constitute an important part of process-based improvement activities in software organizations. Anecdotal evidence shows two out of three programs fail in two years. We propose and estimate a structural equation model to study the impacts of organizational factors on the use of metrics programs in decision-making and subsequent organizational performance. Moreover, metrics programs and process-based improvement activities are influenced strongly by institutional forces in the software industry. We also model the influences of these institutional forces on software metrics programs. The data for the paper was collected through a survey conducted over the World Wide Web. The analysis was carried out using LISREL. Results show support for the proposed model. In particular, the institutional forces have a strong influence on metrics programs. We also show that the use of metrics information in decision-making leads to improved organizational performance.