Abstract Leak detection by measuring the pressure increment in a vacuum has found wide application on pressure vessels, pipes and accessories. However, the reproducibility of the method and the effect of some secondary factors on the results of leak detection have not yet been satisfactorily studied and explained. In this paper, the leak detection method is discussed and the possibility of measuring small leakage rates outlined. Some factors influencing the results, especially the virtual leakage rate and test temperature, which may affect the accurate measurement of small leakage rates, are investigated in detail. In order to improve measurement accuracy, it is suggested that the evacuation time is prolonged and, if possible, the testing system is pre-heated, to eliminate the influence of the virtual leakage rate as far as possible. The research results indicate that both the variation of the ambient temperature and the inconsistency of the temperature of the testing system with that of the gas leaking into the system will introduce errors. Methods for correcting measured results are proposed.