Uncertainty of Software Requirements
Thomas Fehlmann, Luca Santillo · 2007
Uncertainty is a measurable notion in statistics and measurement theory. Also, uncertainty typically propagates from one stage to the other in a measurement or estimation process. Software Requirements constitute an event space following standard distributions. Thus statistical concepts such as residuals, standard distribution, and uncertainty – and uncertainty propagation – are applicable from statistics and measurement theory. This work presents preliminary results that allow evaluating the benefits of pursuing such a research direction for Six Sigma practice. We present the results of a field study where we compare the life cycles of new requirements to Personalised Customer Communications Software with the calculated uncertainties of these requirements. A general approach to uncertainty propagation is applied in this case to quantify to what extent the uncertainty in input variables (requirements definition, for instance) can affect the outcomes of software analysis and development processes (project size, for instance), and finally our software estimation true capability. Uncertainty propagation is also known as “error propagation” in science (not to be confused with defects in software engineering).