Abstract
This paper considers the approach taken by the ESPRIT-funded REQUEST project to measuring, modelling and predicting software quality. The paper describes previous work on defining software quality and indicates how the work has been used by REQUEST to begin the formulation of a constructive quality model (COQUAMO). The paper concludes that the original goal of the project to produce a predictive quality model was unrealistic, but that a quality system incorporating prediction, analysis and advice is feasible.