Basic Estimation Strategies

Abstract

One of the essential decisions during estimation is the abstraction level on which we estimate. At one extreme, we may predict effort for a complete project. At the other extreme, we may predict effort for individual work packages or activities. Dissonance between abstraction levels on which we are able to estimate and the level on which we need estimates is a common problem of effort prediction. We may, for example, have past experiences regarding complete projects and thus be able to predict effort for complete projects; yet, in order to plan work activities, we would need effort per project activity. Two basic estimation “strategies” exist for handling this issue: bottom-up and top-down estimation.

Further Reading

• R.D. Stutzke (2005), Estimating Software-Intensive Systems: Projects, Products, and Processes, Addison-Wesley Professional.

  In Chaps. 11 and 12 of his book, the author discusses bottom-up and top-down estimation strategies, respectively. Example techniques for implementing both estimation strategies are described. The author also discusses threats in making effort-time trade-offs and schedule compression.

• Project Management Institute (2006), Practice Standard for Works Breakdown Structures. 2nd Edition. Project Management Institute, Inc. Pennsylvania, USA.

  The practice standard provides a summary of best practices for defining work breakdown structures. It provides an overview of the basic WBS process, criteria for evaluating the quality of WBS, and typical considerations needed when defining WBS. Finally, the standard provides a number of example work breakdown structures from various domains, including software development and process improvement.

• R.T. Futrell, D.F. Shafer, and L. I. Shafer (2002), Quality Software Project Management, Prentice Hall.

  In their book, the authors discuss approaches for creating project work breakdown structures and identifying project activities in the context of project management. In Chap. 8, they present top-down and bottom-up strategies for creating project-oriented WBS and show different WBS approaches that implement these strategies. In Chap. 9, the authors show how to populate a WBS to identify project activities and tasks relevant for effective project planning.

• IEEE Std 1074 (2006), IEEE Standard for Developing a Software Project Life Cycle Process, New York, NY, USA. IEEE Computer Society.

  This standard provides a process for creating a process for governing software development and maintenance. It lists common software development life cycle phases, activities, and tasks. The standard does not imply or presume any specific life cycle model.

• ISO/IEC 12207 (2008), International standard for Systems and Software Engineering - Software Life Cycle Processes, International Organization for Standardization and International Electrotechnical Commission (ISO/IEC), and IEEE Computer Society.

  Similar to the IEEE 1074 standard, this international standard aims at specifying a framework for software life cycle processes. Yet, it comprises a wider range of activities regarding life cycle of software/system product and services. It spans from acquisition, through supply and development, to operation, maintenance, and disposal. Similar to IEEE 1074, this standard also does not prescribe any particular life cycle model within which proposed phases and activities would be sequenced.

• M. Jørgensen (2004), “Top-down and bottom-up expert estimation of software development effort.” Information and Software Technology, vol. 46, no. 1, pp. 3–16.

  The author investigates strengths and weaknesses of top-down and bottom-up strategies in the context of expert-based effort estimation. In his industrial study, the author asked seven teams of estimators to predict effort for two software projects, where one project was to be estimated using a top-down strategy and the other using a bottom-up strategy. In the conclusion, the author suggests applying a bottom-up strategy unless estimators have experience from, or access to, very similar historical projects.

• S. Fraser, B.W. Boehm, H. Erdogmus, M. Jørgensen, S. Rifkin, and M.A. Ross (2009), “The Role of Judgment in Software Estimation,” Panel of the 31st International Conference on Software Engineering, Vancouver, Canada, IEEE Computer Society.

  This short article documents a panel discussion regarding the role of expert judgment and data analysis in software effort estimation. Panelists underline the thesis that “there is nothing like judgment-free estimation” but also stress the importance of quantitative historical information for software estimation.

• M. Jørgensen (2007), “Forecasting of software development work effort: Evidence on expert judgment and formal models,” International Journal of Forecasting, vol. 23, no. 3, pp. 449–462.

  This article provides a comprehensive review of published evidence on the use of expert judgment, formal methods, and hybrid methods for the purpose of software project effort estimation. The final conclusion is that combining model- and expert-based approaches works principally better than either one alone.

• M. Jørgensen (2004), “A Review of Studies on Expert Estimation of Software Development Effort,” Journal of Systems and Software, vol. 70, no. 1–2, pp. 37–60.

  In Sect. 5.2 of his article, the author provides an overview of various approaches for combining multiple estimates provided by human experts. In addition, he discusses typical factors on which benefits of combining multiple expert estimates depend.

• M. Jørgensen and K. J. Moløkken (2002), “Combination of software development effort prediction intervals: why, when and how?” Proceedings of the 14th International Conference on Software engineering and knowledge Engineering, Ischia, Italy, pp. 425–428.

  The authors investigate an empirical study of three strategies for combining multiple, interval estimates: (1) average of the individual minimum and maximum estimates, (2) maximum and minimum of the individual maximum and minimum estimates, and (3) group discussion of estimation intervals. Results of the study suggest that a combination of prediction intervals based on group discussion provides better estimates than “mechanical” combinations. Yet, the authors warn that there is no generally best strategy for combining prediction intervals.