Software reliability is an important part of software as it ensures both quality and stability. Because bugs accompany software during development, it is important to establish a method to check and evaluate software reliability. For companies, the software reliability growth model is a good solution because it can be used as an indicator of the number of potential failures [1]. Therefore, in the past few years, more and more companies tried applied software reliability growth models to ensure software quality [2][3][4][5][6]. Developing a model curve for the ratio that bugs occur will increase the efficiency of software development. Applying a growth model to the actual data, which includes bug information, is a useful method. Although numerous models have been developed, a universal model has yet to be established. In particular, some data may be poorly fitted by a single software reliability growth model. Herein we try to resolve the situation where all the actual data about information does not fit one growth model well. We use the multistage model as a new modeling method to fit defect information, improving the fit when some of the data is ill fitted using one software reliability growth model. Specifically, we address the following questions: RQ1: What is the most appropriate metric to evaluate the fit of a software reliability model? RQ2: Is the multistage model better suited for software reliability? What metric can be used to evaluate the suitability of the model? Both researchers and members of the company involved in the case study have a vested interested in this study. As researchers, we hope to find a way to ensure software reliability in software development and release, especially in the case where a traditional software reliability growth model cannot meet the required accuracy. The members of the software development company hope to discover a process that realizes a smoother and more economical development process using a predictive model.