An effective and scalable metaheuristic algorithm termed Phase Based Optimization (PBO) for solving big optimization problems is proposed. In the natural system, the individuals with three phases which are gas phase, liquid phase and solid phase have completely different motional characteristics. PBO mimics the above three kinds of motional characteristics of individuals, and three corresponding operators, diffusion operator of gas individuals, flowing operator of liquid individuals and perturbation operator of solid individuals are devised. The diffusion operator and the flowing operator are utilized to perform the task of divergence and convergence respectively, and the perturbation operator plays a role of fine-tune search. Despite its algorithmic simplicity, PBO can effectively find a very better solution even in a high dimensional search space. The experimental results demonstrate that PBO can provide much better accuracy on optimized solutions and lower time complexity than the other state-of-the-art optimization algorithms.