A randomized, placebo-controlled, efficacy trial of Carraguard was unable to demonstrate a reduction in women’s risk of HIV infection, which may have been due, in part, to low adherence (gel used in 42 % of vaginal sex acts, on average). A secondary analysis was undertaken to understand baseline factors associated with high adherence (gel used in ≥85 % of sex acts). Women who reported ≥1 vaginal sex act, returned ≥1 opened applicator, and had ≥1 conclusive post-enrollment HIV test (N = 5990) were included. Adherence was estimated as the ratio of average weekly applicator insertions (based on a dye stain assay indicating vaginal insertion)/average weekly sex acts (by self-report). Multivariate logistic regression modeling indicated that coital frequency, site, contraception, and partner age difference had a significant impact on adherence. Women reporting >1 and ≤2 vaginal sex acts per week, on average, were half as likely to be adherent as those reporting 1 vaginal sex act per week or less [adjusted odds ratio (AOR): 0.48; 95 % CI 0.38–0.61]; women from the Western Cape had one-third the odds of being adherent compared to women from KZN (AOR: 0.31; 95 % CI 0.23–0.41); compared to women using injectable contraception, women using any other or no method were more likely to be adherent (AOR: 1.30; 95 % CI 1.04–1.63); and women who had a larger age gap from their partners were more likely to be adherent (AOR: 1.03; 95 % CI 1.01–1.05; p = 0.001). Despite low adherence, overall, 13 % of participants achieved nearly perfect adherence, indicating a potential niche for a coitally dependent microbicide. More research is needed on the impact of sexual patterns and HIV risk perception on product acceptability and adherence to improve counseling in ongoing trials and when products are eventually introduced.