MULTAN-multisolution tangent formula procedure is based on the probability distribution of three phase invariants. Based on this distribution, we analyzed how a small starting set, the main weakness of MULTAN, effects the process of phase development and refinement by the tangent formula. The reason of MULTAN failure and the approach to overcome the disadvantage of small starting set are discussed here. The principle and advantage of RANTAN-random MULTAN is elucidated briefly and compared with MULTAN in fourexamples.