Article Preview
TopIntroduction
The announcement of the Agile Manifesto (Fowler & Highsmith, 2001) signaled a paradigm shift (Kuhn, 2012) in software development. The resulting community response to agile software development was quite favorable, and while researchers and practitioners raised or discussed issues on quality (Ambler, 2005), balance between agility and discipline (Boehm & Turner, 2004; Henderson-Sellers & Serour, 2005), project management (Turk, France, & Rumpe, 2005), novelty (Merisalo-Rantanen, Tuunanen, & Rossi, 2005), and adaptability (Aydin, Harmsen, Van Slooten, & Stagwee, 2005), others asked for theory-based, rigorous research to evaluate the practices (Batra, VanderMeer, & Dutta, 2011; Erickson, Lyytinen, & Siau, 2005). Based on a review of empirical studies, Dyba and Dingsøyr (2008) reported that agile development was being used only for small projects. Leffingwell (2007) provided a basic framework for scaling agile for larger projects. Although it took some time before practitioners addressed the scaling issue, in recent years, large-scale agile methodology (LSAM) has become commonplace in part because of the increasing popularity of methods such as Scaled Agile Framework or SAFe (Knaster & Leffingwell, 2017) and Large-scale Scrum or LeSS (Larman & Vodde, 2016). The rising demand of LSAM (VersionOne, 2018) provides preliminary evidence that we may have found an acceptable balance between agility and discipline as commended by (Boehm & Turner, 2004).
This research commentary suggests five research challenges when conducting quantitative studies on LSAM. First, the LSAM empirical literature needs to be expanded, given that it mainly characterized by qualitative studies that primarily focus on coordination or related issues. Future research should evaluate and validate research models proposed in the qualitative studies by using quantitative techniques. Second, the notion of “large” in LSAM needs to be clarified. Third, the popular methods list too many maxims, which may result in difficulty in operationalizing the LSAM dependent variable, particularly for acceptance studies. Fourth, the researcher may encounter difficulty in choosing independent variables from a lengthy list of LSAM constructs, especially for project success studies. Finally, given the recent criticisms of LSAM practices, researchers should take a cost-benefit approach when evaluating LSAM because large projects invariably have inherent problems that can be mitigated but not eliminated by software development methods. Researchers may mistakenly focus on shortcomings that may not result solely from the use of LSAM; in fact, practitioners may rate a project as a success despite noting some deficiencies. For a more accurate assessment, researchers should compare the performance of LSAM with that of a conventional methodology.