Abstract
Quality assurance effort is often a major cost factor during software development. Especially testing effort can consume more than 50 percent of the overall development effort. Furthermore, it is often unclear how efficient existing quality assurance techniques are and what the potential for savings might be. Currently, companies often conduct different quality assurance activities, such as inspections and testing, in order to find as many defects as possible before software products are delivered. However, most often such techniques are applied in isolation and do not exploit synergy effects from systematically combining them, such as reduced effort or higher defect detection rates. Moreover, the relations between different static and dynamic quality assurance techniques are widely unclear. In addition, testing activities often have a broad scope and are rarely applied in a focused manner, which results in high costs. This thesis presents the In2Test approach, which systematically combines inspection and testing processes for focusing testing activities. The main ideas of this integrated approach are (a) to use early inspection results to prioritize testing on parts of a product or on defect types that are expected to be most defect-prone, (b) to consider product metrics and historical data in order to further improve the test focus, (c) to guide the prioritization of system parts and defect types by using rules that are based on explicitly defined assumptions about the relationships between inspection results and remaining defects.
The approach was validated in two case studies: The validation was aimed at (a) showing that In2Test allows for effort reduction during test execution while keeping a comparable level of detected defects during testing, i.e., In2Test allows for improved efficiency compared to nonintegrated approaches; (b) revealing underlying assumptions for the prioritization of system parts and defect types for testing based on inspection results; and (c) showing that In2Test is mature for use in industrial applications.
The validation (a) showed an effort reduction in the case studies of between 6% and 34% at a comparable level of detected defects depending on the concrete assumptions and selection rules applied, leading to an efficiency improvement of between 7% and 52%; (b) revealed a set of initial selection rules and assumptions with promising results; (c) showed the applicability of the approach.