The diversity and complexity of 5G network services make the Poisson distribution-based reliability test confidence analysis method of traditional products no longer applicable. In order to solve this problem, this paper proposes a confidence analysis method based on application failure distribution statistics. In the early stage of the reliability test, it is designed to conduct a pre-test. Two methods for statistically fitting the pre-tested application failure data are described to obtain the failure distribution law of the 5G network applications. The confidence analysis is then applied in order to get the required duration of the test and the confidence interval of the test result. Case verification shows that this method can help 5G network test designers to obtain credible test results.