Software project managers’ decisions on reviews and tests are difficult. This paper describes a cost-benefit model for specific decisions on quality assurance. The quantitative model is based on single relationships and is quantified with historical data. Its results are shown and are compared with cost estimations. The model is able to reflect reported results of process improvement. Data collected in student projects is used to evaluate the model. Project averages and single projects are considered. Furthermore, results of a cross-validation are shown.