Secondary data analysis was employed to scrutinize factors affecting sample retention in a randomized evaluation of an early childhood intervention. Retention was measured by whether data were collected at 3 points over 2 years. The participants were diverse, immigrant, and U.S.-born families of color from urban, low-income communities. We examined how the initial recruitment and enrollment process, and sample demographics related to retention. Effects that adversely related to retention included recruitment from a public area (e.g., bus stop) versus personally salient locations (e.g., child's school); assignment to the control group versus the treatment group; longer time lapses in communication between researchers and participants; and living in a less-resourced, low-income neighborhood relative to higher resourced, low-income neighborhoods. Being born outside the United States was positively associated with retention relative to participants born in the U.S. Implications for evaluators and recommendations for evaluation methodology are discussed.