The test data of large-scale complex product usually presents small sample and multi-phase characteristic.Therefore, to reduce the evaluation error caused by small sample of one phase, a constraint condition is introduced to products reliability evaluation, thus the multi-phase test data can be used. Under such constraint condition, the reliability parameter is considered as a random variable and the reliability parameter distribution function is given firstly. Then an approximate calculation method for mission reliability and mean time between failure (MTBF) is presented. Finally, the result based on actual data is shown. Since the constraint condition is brief and the reliability parameter distribution function can be achieved, this method is more applicable for engineering than traditional AMSAA-Model (army materiel systems analysis activity model), and the reliability interval estimation can be gained which can not be obtained from Duane-model.