Efficiency gains in South African smallholder farming systems are crucial for sustainable agriculture development. A Bayesian hierarchical model was developed to analyse data from multiple farms, accounting for variability between different scales of production units within the same farm. The model revealed significant gains in efficiency across various farm sizes, with a mean improvement rate of 15% in productivity metrics. The Bayesian hierarchical approach effectively captures intra-farm and inter-farm variations in smallholder farming systems, providing robust insights into efficiency dynamics. Further research should explore the model's scalability to other regions and its impact on policy interventions aiming at increasing agricultural productivity. The empirical specification follows $Y=\beta_0+\beta^\top X+\varepsilon$, and inference is reported with uncertainty-aware statistical criteria.