This paper presents the concept of completeness for kinematic identification of robot manipulators. Completeness is defined as the ability to map joint positions into tool positions for all arbitrary manipulators. It is suggested that complete models must contain a certain number of independent parameters. Furthermore it is suggested (and shown by practical examples) that the required number of independent kinematic parameters is easy to determine a-priori. This enables one to check a model for completeness. Although the basic idea behind kinematic identification may have been considered well known, several identification algorithms in the recent literature are incomplete. Two examples are included in this paper. For this reason, this paper presents the topic and includes the conditions for a complete and viable identification algorithm.