A selective-dry-cow therapy algorithm was evaluated for microbiological cure risk, new infection risk, culling and occurrence of clinical mastitis before 30 DIM, and 1st-test milk yield and linear score (LS) in a randomized on-farm clinical trial including 612 dairy cows. An algorithm using DC305 and test-day data was used to identify cows as “low risk” (cows that likely will not benefit from dry cow antibiotics) or “high risk” (cows that will benefit). Low risk cows were those that had all of: < 200k SCC at last test, an average SCC <200k on the last 3 tests, no signs of mastitis at dry-off, and have not had more than 1 clinical mastitis event in the current lactation. Low risk cows were randomly assigned to receive either intramammary antibiotics and external teat sealant (DCT) or teat sealant only (TS). Quarter milk samples were obtained from cows at dry-off and 1–7 DIM to determine cure and new infection at the quarter level. Samples from high risk cows were used to determine positive and negative predictive values (PPV, NPV) of the algorithm. Mastitis events, milk production, LS, and culling data were retrieved from DC305. Data analysis was performed in SAS 9.4: categorical outcomes were analyzed using Fisher’s exact tests while continuous outcomes were compared with t-tests. PPV and NPV were each 70%. Of cultures eligible for cure analysis (n = 157), 91% of DCT cured, while 83% of TS did (RR of non-cure TS:DCT = 1.9; 95%CI: 0.8–4.6). Positive cultures for coagulase negative staphylococcus (CNS) at dry-off accounted for 95% of the non-cures (n = 19/20). Risk ratio for new infection was 1.4 for TS:DCT (95%CI: 1.0–2.0). CNS accounted for 50% of new infections (n = 86/135). There were no statistical effects of treatment group on culling (DCT n = 14; TS n = 18), clinical mastitis (DCT n = 5; TS n = 3), milk (kg) (DCT = 38.9; TS = 39.8), or LS (DCT = 2.4; TS = 2.5). The impact of CNS to increased new infection risk and decreased bacteriological cure needs to be further investigated. These results suggest that the employed algorithm decreased dry cow antibiotic use by 64% without adversely impacting production outcomes.