Declining response rates have made traditional, probability-based sampling methods more resource-intensive and thus more expensive. Studies of population subgroups are particularly vulnerable to this trend, as smaller group sizes as well as other factors often make these groups "hard-to-reach" or "hard-to-survey". In response, researchers have increasingly relied on network sampling methods such as snowball sampling and respondent-driven sampling (RDS) to identify research participants. RDS has become particularly popular as it improves upon snowball sampling to allow researchers to make probability-based estimates. However, the assumptions on which RDS is based increase its difficulty of implementation, a fact often glossed over in the literature. This presentation will provide an overview of respondent-driven sampling with comparison to other network sampling methods along with several use cases and discussions of lessons learned in implementing RDS.