Objectives: Previous studies of parenting programs suggest that facilitator fidelity, participant attendance and engagement often influence treatment outcomes. While the number of parenting program evaluations has been growing in low- and middle-income countries, little is known about the implementation processes and their impact on participant outcomes in these settings. Methods: This study was nested within a cluster-randomised trial of a parenting program in South Africa. The paper aims to, first, describe the implementation of the intervention over 14 weeks. Second, using longitudinal multilevel analyses, the paper examines the impact of variation in observer-rated fidelity, attendance, and engagement on participant outcomes – parenting and maltreatment reported by caregivers and adolescents aged 10-18 (N=270 pairs), 14 outcome constructs. Results: Fidelity, attendance and participant engagement rates were similar to those reported in high-income country studies. However, the participation and implementation characteristics did not predict participant outcomes. This may be due to limited variation in dosage as home visits were comprehensively provided when participants could not attend group sessions, and fidelity was monitored by the implementers and researchers. One statistically significant predictor after the multiple testing correction was higher fidelity predicting an increase in adolescent-reported maltreatment at follow-up, possibly due to an increase in reporting (incidence rate ratio 1.33, 95% CI [1.19, 1.49], p<0.01). Conclusions: Our study confirms that a high quality of implementation can be achieved in a low-resource context. Suggestions for future research on parenting programs include examining therapeutic alliance alongside program fidelity and facilitator skill as well as systematically recording program adaptations.