Previous studies on power-distribution efficiency in South Africa have relied on deterministic models, which fail to adequately account for regional heterogeneity and uncertainty in ageing infrastructure performance data. This replication study aims to validate and extend prior findings on national efficiency gains by implementing a Bayesian hierarchical model, providing a robust probabilistic framework for inference and forecasting. We replicated the core analysis of a prior national assessment using original equipment performance data. A Bayesian hierarchical model was specified: $y_{ij} \sim \text{Normal}(\alpha_j + \beta X_{ij}, \sigma^2)$, $\alpha_j \sim \text{Normal}(\mu_{\alpha}, \tau^2)$, with weakly informative priors. This accounts for region-specific effects ($\alpha_j$) and quantifies uncertainty via 95% credible intervals. The model confirms a positive national trend but reveals significant regional variation masked in prior aggregate analyses. The posterior distribution indicates a central efficiency gain of 4.7% (95% CrI: 3.1 to 6.3%), with the Western Cape region showing gains 1.8 times the national median. The Bayesian hierarchical approach offers a superior, probabilistic quantification of efficiency gains, substantiating the general trend while critically highlighting substantial inter-regional disparities. Future infrastructure investment analyses should adopt hierarchical modelling to prioritise regions with lagging efficiency. Utility regulators should mandate probabilistic reporting to better inform policy. Bayesian hierarchical model, distribution losses, infrastructure efficiency, probabilistic forecasting, replication study This study provides a novel probabilistic replication framework for infrastructure efficiency analysis, demonstrating that regional variation in technical gains is substantially greater than previously reported.