In decision making it has been generally assumed that the decision maker possesses the enough knowledge about the whole problem and she or he is able to distinguish the degree up to which an option is better than other one. However, this may be impractical in real world situations, which can involve many options to select from information sources that can be dynamic and conflicting. Therefore, different procedures for estimating incomplete information have been developed in the literature, which principally deal with fuzzy preference relations. In this research, and different from the existing procedures that deal with fuzzy preference relations, we introduce a procedure based on granular computing to estimate missing values in intuitionistic reciprocal preference relations. In particular, we describe how the granular computing can be used to estimate missing information so that the estimated values lead to complete intuitionistic reciprocal preference relations with consistency levels as high as possible. The results show that the procedure based on granular computing is able to increase the consistency level achieved by the existing procedures dealing with missing values in intuitionistic reciprocal preference relations.