An analysis of trends in productivity and cost drivers over years

TLDR: The empirical analysis on how changes in software engineering practices are reflected in COCOMO cost drivers and how software productivity has evolved over the years provides empirical evidence that the productivity trends can be characterized by the improvements in software tools, processes, and platforms among other factors.

Abstract: Background: Software engineering practices have evolved considerably over the last four decades, changing the way software systems are developed and delivered. Such evolvement may result in improvements in software productivity and changes in factors that affect productivity.Aims: This paper reports our empirical analysis on how changes in software engineering practices are reflected in COCOMO cost drivers and how software productivity has evolved over the years.Method: The analysis is based on the COCOMO data set of 341 software projects developed between 1970 and 2009. We analyze the productivity trends over the years, comparing productivity of different types and countries. To explain the overall impact of cost drivers on productivity and explain its trends, we propose a measure named Difficulty which is based on the COCOMO model and its cost drivers.Results: The results of our analysis indicate that the overall productivity of the projects in the data set has increased noticeably over the last 40 years. Our analysis also shows that the productivity trends and productivity variability can be explained by using the proposed Difficulty measure.Conclusions: Our analysis provides empirical evidence that the productivity trends can be characterized by the improvements in software tools, processes, and platforms among other factors. The Difficulty measure can be used to justify and compare productivity among projects of different characteristics, e.g., different domains, platforms, complexity, and personnel experience. Although we define the measure using the COCOMO cost drivers, it may not fully represent the most important factors influencing productivity. One direction for our future work is to analyze the effectiveness of the measure using more cost drivers on more data points.