PROCESS MATURITY AND SOFTWARE QUALITY: A FIELD STUDY

Quality has emerged as a key issue in the development and deployment of software products (Haag et al. 1996; Prahalad and Krishnan 1999; Yourdon 1992). As software products play an increasingly critical role in supporting strategic business initiatives, it is important that these products function correctly and according to users’ specifications. The costs of poor software quality (in terms of reduced productivity, downtime, customer dissatisfaction, and injury) can be enormous. For example, the Help Desk Institute, an industry group based in Denver, estimates that in 1999, Americans spent 65 million minutes on “hold” waiting for help from software vendors in debugging software problems (Minasi 2000). An unresolved issue is how software quality can be improved. On the one hand, some software researchers and experts argue that quality can be tested into software products. That is, defective software products can eventually become “bug-free” through rigorous testing (Anthes 1997; Hanna 1995). However, on the other hand, there is a notion that quality must be designed or built into software products from the start (Fenton and Neil 1999). That is, quality in design predicts quality in later stages of the product life cycle. There is some empirical evidence to support this notion from Japanese software factories (Cusumano 1991), the NASA Space Shuttle program (Keller 1992), and a variety of projects at IBM (Buck and Robbins 1984). If the level of quality persists throughout a product’s life cycle, how can quality be designed into the product? In manufacturing, Bohn (1995) found evidence that process maturity (i.e., the sophistication, consistency, and effectiveness of manufacturing processes) was positively associated with product quality. This relationship is believed to exist because as a process becomes more mature and less variable, the outputs of the process (i.e., products) have a higher level of quality (Fenton and Neil 1999; Ryan 2000; Zahran 1998). In the context of software production, this implies that maturity of the software development process is essential to reducing process variability and thus improving the quality of software products (Humphrey 1988). There is some empirical support linking process maturity to software quality. For example, Diaz and Sligo (1997) found initial evidence of a positive relationship between process maturity and software quality at Motorola. Herbsleb et al. (1997) found additional anecdotal support of this relationship, but suggest that further research is needed to understand more precisely how process maturity and software quality are related. Determining whether process maturity is linked to software quality is important because, in practice, many managers still emphasize testing at the end of the development cycle instead of building in quality through better processes (Anthes 1997; Hanna 1995). Thus, this study has been designed to address the following question that is central to these issues: What is the relationship between process maturity and software quality over the product life cycle? We develop a conceptual framework (Figure 1) for assessing the relationship between process maturity and software quality at different stages of the product life cycle: development, implementation, and production. Our models are empirically evaluated using archival data collected on software products developed over 12 years by the systems integration division of an information technology company. Based upon our analysis, we identify the direct and indirect marginal effects of improved process maturity on software quality at the different stages of the software life cycle. Our results also provide insight into the question of whether quality is a persistent characteristic of software.