In a randomized controlled setting, medium-term implant migration and long-term clinical outcomes were compared for the Copeland and the Global C.A.P. humeral head resurfacing implants (HHRI). Thirty-two patients (mean age 63 years) were randomly allocated to a Copeland (n = 14) or Global C.A.P. (n = 18) HHRI. Patients were followed for 5 years with radiostereometry, Constant Shoulder Score, and the Western Ontario Osteoarthritis of the Shoulder Index (WOOS). WOOS and revision status were also obtained cross-sectionally at a mean 10-year follow-up. At the 5-year follow-up, total translation (TT) was 0.75 mm (95% confidence interval [CI]:​ 0.53–0.97) for the Copeland HHRIs and 1.15 mm (95% CI:​ 0.85–1.46) for the Global C.A.P. HHRIs (p = 0.04), but the clinical scores were similar at all follow-ups. The cumulative risks of revision at 5 and 10 years were 29% and 43% for Copeland and 35% and 41% for Global C.A.P HHRIs (p > 0.7). No implants were loose at revision, but HHRIs that were later revised followed an early offset-increasing migration pattern with medial translation and lift-off resulting in a mean 0.53 mm (95% CI:​ 0.18–0.88) higher TT at the 1-year follow-up compared to non-revised HHRIs. In conclusion, the Global C.A.P. HHRI had higher TT compared with the Copeland HHRI, but clinical scores and revision rates were similar. Nonetheless, revision rates were high and challenge the use of HHRIs. Interestingly, an early radiostereometry evaluated HHRI migration pattern with increased off-set predicted later implant revision.