The importance and complexity of recent software is increasing more and more because the software now must provide a considerable number of functions as well as a high level of performance. Therefore, the development of highly reliable software is important. To develop reliable software, it is necessary to manage the software reliability from the early development phases. However, most reliability estimation models are used during the late test phases such as system testing or operational testing. In order to develop highly reliable software, it is necessary to manage software reliability at the early test phases such as unit and integration testing based on the characteristics of the phases in which developers and testers are not separated and in which developers perform tests and debug activities together. Therefore, the creation of a new reliability estimation model considering the test and debug time together is necessary. This thesis proposes a new reliability estimation model to manage the reliability of individual units from the early test phases. The model is also fitted to actual data, demonstrating its usefulness. In this study industrial data are collected used in an experiment.