Overcoming the challenges in cost estimation for distributed software projects

  • Authors:

  • Narayan Ramasubbu;Rajesh Krishna Balan

  • Affiliations:

  • Singapore Management University, Singapore;Singapore Management University, Singapore

  • Venue:
  • Proceedings of the 34th International Conference on Software Engineering
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

We describe how we studied, in-situ, the operational processes of three large high process maturity distributed software development companies and discovered three common problems they faced with respect to early stage project cost estimation. We found that project managers faced significant challenges to accurately estimate project costs because the standard metrics-based estimation tools they used (a) did not effectively incorporate diverse distributed project configurations and characteristics, (b) required comprehensive data that was not fully available for all starting projects, and (c) required significant domain experience to derive accurate estimates. To address these challenges, we collaborated with practitioners at the three firms and developed a new learningoriented and semi-automated early-stage cost estimation solution that was specifically designed for globally distributed software projects. The key idea of our solution was to augment the existing metrics-driven estimation methods with a case repository that stratified past incidents related to project effort estimation issues from the historical project databases at the firms into several generalizable categories. This repository allowed project managers to quickly and effectively “benchmark” their new projects to all past projects across the firms, and thereby learn from them. We deployed our solution at each of our three research sites for real-world field-testing over a period of six months. Project managers of 219 new large globally distributed projects used both our method to estimate the cost of their projects as well as the established metricsbased estimation approaches they were used to. Our approach achieved significantly reduced estimation errors (of up to 60%). This resulted in more than 20% net cost savings, on average, per project – a massive total cost savings across all projects at the three firms!