Exercise referral schemes (ERS) are internationally recognised, yet little attention has been paid to discrete referral groups or the influence of wider social determinants of health. The primary quantitative element of this mixed methods study used a mixed effects linear model to examine associations of sociodemographic predictors, obesity class and profession of referrer on weight and physical activity (PA) variables for weight-related referrals (n = 3624) to an established 24-week ERS. Chained equations modelling imputed missing data. The embedded qualitative element (n = 7) used individual semi-structured interviews to explore participant weight-related expectations and experiences. Age, gender and profession of referrer influenced weight loss. PA increased and was influenced by age and gender. The weight gap between the most and least obese narrowed over time but the PA gap between most and least widened. Age, employment and obesity class were most predictive of missing data but would unlikely alter overall conclusions. Qualitative themes were weight-loss support, personal circumstances and strategies, and weight expectations versus wellbeing rewards. This ERS worked, did not widen existing obesity inequalities, but demonstrated evidence of PA inequalities for those living with deprivation. To improve equity of experience, we recommend further stakeholder dialogue around referral experience and ongoing support needs.