PhD student Michael Unterkalmsteiner is eager to find out why requirements engineers and software testers don’t talk to each other. If communication between the two had been better, his research might not have been necessary. Michael is now helping four large Swedish organisations to improve their software development. He is also developing a system that will make it easier for Axis, a global surveillance camera developer, to select test cases ahead of launching new cameras.
It is rare in large companies that the people who write software specifications also carry out the necessary software tests. This alone does not pose a problem, but if those who write the specifications make changes to the software without informing the testers – then issues inevitably arise.
Michael Unterkalmsteiner’s research focuses on finding the problem areas in the communication between these two groups. While there are many technical solutions available for issues surrounding, for example, version control , Michael has found that communication difficulties are often related to a lack of resources.
”One of the issues is that people who work for software developers rarely have enough time to deal with the administrative tasks that are vital in order for the process between the requirements engineer and tester to flow smoothly.” Michael says.
In light of this, Michael and his colleagues have developed a framework called REST-bench, where RE stands for ‘Requirement Engineering’, ST stands for ‘Software Testing’, and bench stands for ‘benchmark’. The framework is used by companies who need help to identify bottlenecks in the software development process.
By interviewing a software tester and a requirements engineer who have worked together on the same project, Michael can create a so-called artefact map that shows how the process works. Following this, a workshop is held where the identified bottlenecks are presented, allowing the companies to see where improvements need to be made.
Michael has thus far carried out consulting research assignments for four different companies - Ericsson, Volvo, CompuGroup and Sony Mobile.
”Some of the companies are really eager to make the necessary improvements. I am also planning to revisit the same companies in a year or so to see if there have been any lasting improvements made.” says Michael.
He also says the method is applicable to practically any business that develops software, but that data to support the effectiveness of the method will not be available until next year.
As well as the REST-bench project, Michael is also working on another project together with surveillance camera developer Axis. When Axis develop a new camera they use the specifications from an earlier product as a base, adding the new features and changes the new product requires. The problem is that these changes can render the existing test cases for the camera irrelevant.
Michael has therefore developed a tool that serves as a link between source code and test case. Axis has approximately 1600 different test cases, and the tool assigns a ranking score to how well different test cases correspond to different camera features.
”The aim is that Axis will not need to choose different test cases at random, but instead use the ranking scores to establish which tests are most relevant. This saves both time and money.” says Michael.