Research on technology that is to be adopted in an industrial setting must give evidence of relevance to the industry. For this, case studies are important in that they give the opportunity to test technology in realistic surroundings with all the affecting factors.
This thesis is a systematic review of 50 randomly selected articles that report case studies. The objective of the investigation is first of all to get an overview of the state of the art regarding the use of case studies in empirical software engineering. Secondly, the investigation should identify important characteristics of case studies for researchers to give careful considerations when conducting case studies.
The data collected during analysis of these 50 articles, was used to address the following issues: the extent of case studies in empirical software engineering, the quality of reporting case studies, the specification of the case study research method, what researchers call a case study, the affiliation of authors, confusion regarding research methods, and the extent of the use of multiple case studies.
The main findings of this review are:· Close to twelve percent of the 427 papers searched, use case study as the research method.· There are great variances in the way of reporting case study results. The general impression is that information is not clearly reported.· Researchers are not very likely to explicitly state what kind of research method that has been used.· Case studies are mainly used for two purposes, namely evaluative and demonstrative purposes:1) Typical characteristics for articles with an evaluative nature are rather high response rates for the six questions in the survey, the reporting of observations of use, and most likely the use of professionals as subjects.2) Typical characteristics for articles with a demonstrative nature arerelatively low response rates for the six questions in the survey, thereporting of technology outcome, and most likely the use of authors of the articles as subjects.· The majority of the articles with authors affiliated in research communities appear to report technology data.· The lack of observations of use may be reminiscent of the assertion method.· The extent of multiple case studies is 22 percent.
The following criteria for case studies in empirical software engineering were suggested: First of all, the author should specify that the research method used is the case study method. The focus in the case study should be use/evaluation of a software technology. Furthermore, the case study should test a technology in an industrial setting. Finally, the technology must be used by others than the researchers themselves (because of no manipulation), preferably by professionals.
There is a need for a specified definition of case studies in empirical software engineering. Additionally, in order to produce results that are easy for reviewers and industry to relate to, there is a need for standards for how to conduct case studies. Use of guidelines would help researchers ensure the quality of the results. Hence, guidelines for assistance through the case study process will be an important device for improving future use of this research method.
The main contribution of this review is in presenting the state of affairs and a characterization of case studies as used in empirical software engineering. Such an overview should be useful for researchers in the work of improving the case study research method. Ultimately, this review should contribute to the work of improving the use of the case study research method in empirical software engineering.
It is hoped that the findings of this research will prove valuable to empirical software engineering, whose main interest is that of investigating the interaction between technology and developers.