In software technology today, several development methodologies such as extreme programming and open source development increasingly use feedback from customer testing. This makes the customer defect data become more readily available. This paper proposes an effective use of reliability models and defect data to help managers make software release decisions by applying a strategy for selecting a suitable reliability model, which best fits the customer defect data as testing progresses. We validate the proposed approach in an empirical study using a dataset of defect reports obtained from testing of three releases of a large medical system. The paper describes detailed results of our experiments and concludes with suggested guidelines on the usage of reliability models and defect data.