A strategy for a credible and auditable estimation process using the ISBSG international Data repository
Alain Abran, Reiner R. Dumke, Jean‐Marc Desharnais, I. Ndiaye, Christian Kolbe · 2002
The software estimation process must provide credible input for business decisionmaking. Most of the time, business managers must rely on incomplete information to make decisions: there is almost always some information lacking or too expensive to gather within the time frame of the decision--making process. Within their field of expertise, decision-makers can make valuable expert judgments on the missing or incomplete information. However, many decision-makers have not mastered the information technology domain and need expert support to fill in the gaps in their knowledge of the subject. For this reason, the software estimation process must provide decision-makers not only with estimates (the “the numbers”), but also information on the quality and confidence level of that estimate. The key assumptions and the key uncertainties inherent in the estimation process must be conveyed to business managers to help them make informed business decisions on the basis of the estimates provided to them. For example, decisionmakers should have a feel for the quality and accuracy of the inputs, as well as for the estimation models used for deriving the estimates. The estimation process must be credible from a business perspective, and the outcomes of the estimation process must include statements on the credibility of its various components. Furthermore, since major investment decisions are made based on these estimates, the full estimation process should be auditable. This paper highlights the key elements that make an estimation process both credible and auditable. This includes a discussion on the quality of the input measures (products, processes and resources), the reliability of the productivity models built into an estimation process, the other inputs to the estimation process (key assumptions, constraints and expert judgments) and the type of decisions that should be taken on the basis of the confidence level of the outcomes of the estimation process. This is illustrated with examples from the multi-organizational project repository of the International Software Benchmarking Standards Group – ISBSG.