Informant accuracy is a pervasive issue in the social sciences and persists through the ongoing use of self-report measures of behaviors that are subject to recall errors. The current study reports findings from methods we used to measure substance use among adolescents in South India. We used a repeated-measures, mixed-methods design that began with semistructured interviews (N = 60) to determine the substances used by adolescents. Next, adolescents were recruited to participate in a four-week study that assessed self-reported substance use at three time points and urinary cotinine, a nicotine metabolite, at two time points. Fifty-six participants completed the study protocol. Findings revealed that during the last phase of the study, participants were more willing to admit using substances. Our results add to existing literature that demonstrates the need to move beyond self-reported data of consumption, and to consider using repeated measures, direct observation of behavior, and biometric markers of behavior to ensure more accurate assessments of sensitive topics.