Binary decision diagram (BDD) is an efficient algorithm applied in many domains. When used in the area of system reliability and safety, the BDD technology can improve both qualitative and quantitative analysis of system failures by solving large fault trees. However, the size of a BDD depends critically on the decision variable (basic event of fault trees) ordering. This paper proposed a novel ordering method, named as progressive neighbor first (PNF) ordering, which emphasizes the logical relationship of the basic event variables in a fault tree. The ordering priority gives to the neighbor events of an ordered variable. It is a progressive ordering method, i.e., on each BDD branch, it allows different variable ordering. Compared with current existing ordering heuristics, our proposed PNF ordering method shows a significant advantage on the performance of BDD size.