A pluralist systemic framework for the evaluation of factors affecting software development productivity

Abstract

pAst software development productivity research has taken two main directions: in the first, research has concentrated on determining the factors which have a significant effect on productivity; and in the second, the emphasis has been on determining the best way to measure productivity (Maxwell et al, 1996:706). Historically research was aiming mainly to find solutions in the second area while work in the first area was just a complementary one. The belief was that an analytical model may help resolve all problems associated with estimating software productivity. Recently there have been attempts for a new, holistic understanding of the problem. Such an example is the systems dynamics modelling approach for controlling software projects (Alide lHamid, 1990). Other studies that stress the importance of better understanding of various factors affecting programmer productivity without considering their role in a particular software cost estimation model are those by Kemayel et al (1991), Finnie et al. (1993), Maxwell et al. (1996). In most cases these factors have been analysed with respect to the tuning or validation of a particular metric for cost/effort estimation. These factors are referred to in a different way in different models: in Boehm's model (1981) they are called cost drivers; in Bailey and Basili (1981) -input to the model. Due to their large number, a certain classification of these factors is appropriate. Thus Kemayel et al. (1991) focus their attention on the controllable factors affecting software development productivity. According to them, controllable factors are those factors pertaining to the software process, that a typical software manager has the latitude to determine. They investigate 33 controllable factors placed in three groups: factors pertaining to personnel, factors pertaining to the process; and factors pertaining to the user community. Finnie et al. (1993) investigate 18 factors grouped in four groups: technical attributes, project attributes, developer attributes and user attributes. Benbasat and Vessey (1980) discuss 19 factors grouped in 7 classes: organisational operations characteristics, computer hardware characteristics, source language characteristics, programmer characteristics, programming problem characteristics. The brief look at the lists of the factors in various publications shows that a small number of them evolve with time and the development of technology. Thus it seems inappropriate these days to consider the effect of memory size or storage devices as have been done by Benbasat and Vessey (l980). Another conclusion that comes to mind is the fact that researchers focused their attention on quite different sets of factors and thus it is very difficult to compare results from various surveys. The purpose of this research is to define a holistic model of the interaction between factors affecting software development productivity. First the relationships between these factors will be discussed and then the methodological foundations and the steps of a framework for the evaluation of these factors within the environment of a particular software project will be examined.