A tolerance relation based consensus approach is proposed to solve multi-attribute group decision making consensus problems, in which the evaluation information is incomplete. In this consensus approach, the difficulty of calculating the group consensus degree is also avoided in multi-attribute group decision making with incomplete evaluation information. Moreover, the unknown information is not filled up, so the difficulty of calculating the unknown information is avoided for that the tolerance relation is constructed by the known evaluation information. Finally, a case study is given to illustrate the feasibility and effectiveness of the proposed method.