Kinematic motor variability is extensively studied in occupational, clinical and sports biomechanics, but the consistency of most motor variability metrics have never been reported. In this study, fourteen subjects performed a repetitive pipetting task on three separate days. Movements of hand, arm and pipette tip were recorded in 3D and used to compute shoulder elevation, elbow flexion and shoulder-arm coordination angles, as well as pipette-tip endpoint precision. Cycle-to-cycle motor variability was quantified using linear dispersion measures of standard kinematics properties such as peak velocity, range of motion, and inter-segmental relative phase. Between- and within-subject consistencies of these variability metrics were quantified by variance components estimated using a nested random effects model. For most metrics, the variance between subjects was larger than that between days and cycles. Entering the variance components in statistical power equations showed that for most metrics, a total of 80-100 subjects will be required to detect a 20% difference between two groups with sufficient power, while this difference can typically be detected  in repeated-measures (paired) designs using 25 subjects. The reported between- and within-subject variance components can be used as a data base to facilitate efficient designs of future studies of kinematic motor variability.