The GEOTRACES program requires the analysis of large numbers of seawater samples for Th-232, Th-230, and Pa-231. During the GEOTRACES international intercalibration exercise, we encountered unexpected difficulties with recovery and contamination of these isotopes, Th-232 in particular. Experiments were carried out to identify the source of these issues, leading to a more streamlined and efficient procedure. The two particular problems that we identified and corrected were (1) frits in columns supplied by Bio-Rad Laboratories caused loss of Th during column chemistry and (2) new batches of AG1-X8 resin supplied by Bio-Rad Laboratories released more than 100 pg of Th-232 during elution of sample. To improve yields and blanks, we implemented a series of changes including switching to Eichrom anion exchange resin (100-200 mu m mesh) and Environmental Express columns. All Th and Pa samples were analyzed on a Neptune multi-collector inductively-coupled-plasma mass spectrometer (MC-ICP-MS) using peak hopping of Th-230 and Th-229 on the central SEM, with either Th-232, U-236 (or both) used to monitor for beam intensity. We used in-house laboratory standards to check for machine reproducibility, and the GEOTRACES intercalibration standard to check for accuracy. Over a 1-y period, the 2 s.d. reproducibility on the GEOTRACES SW STD 20101 was 2.5% for Th-230, 1.8% for Th-232, and 4% for Pa-231. The lessons learned during this intercalibration process will be of value to those analyzing U-Th-Pa and rare earth elements as part of the GEOTRACES program as well as those using U-series elements in other applications that require high yields and low blanks, such as geochronology.