Measuring Effort Estimation Uncertainty to Improve Client Confidence

  • Authors:

  • John Moses

  • Affiliations:

  • University of Sunderland, School of Computing, Engineering and Technology, St. Peter's Campus, Sunderland, SR6 0DD, U.K. john.moses@sund.ac.uk

  • Venue:
  • Software Quality Control
  • Year:
  • 2002

Quantified Score

Hi-index 0.01

Visualization

Abstract

A description and a theoretical discussion are given of an approach that it is proposed will improve client confidence in estimates produced using algorithmic software estimation methods, such as Function Point or COCOMO estimation. The approach utilises Bayes Theorem and Bayesian inference. The underlying theory has been successfully used in other arenas of subjective measurement to help improve measurement consistency. It is also proposed that the approach can be used to improve estimation consistency during the estimation of the effort required to develop software development artefacts (e.g. project effort, milestone effort, requirements changes). Software developers will also be able to measure uncertainty in their estimates of artefacts using Bayesian inference. Outsourcers can use the approach to provide statements for client companies about the confidence they have in their estimates. The statements of confidence can then be used to assist outsourcers and clients during project negotiations. Examples are provided to show how the method can be used to measure estimate uncertainty, how estimators can be supported during their estimation procedures, and what kinds of statement can be made to aid project negotiations.