Abstract
{ "background": "The assessment of efficiency gains in industrial machinery fleets within developing economies remains methodologically challenging, often relying on deterministic models that inadequately capture operational heterogeneity and uncertainty. In the context of industrialisation, there is a pressing need for robust statistical frameworks that can inform maintenance and capital investment strategies.", "purpose and objectives": "This study aims to develop and validate a comparative Bayesian hierarchical model for quantifying and predicting efficiency gains within industrial machinery fleets. The objective is to provide a probabilistic framework that accounts for sectoral and temporal variations, enabling more reliable long-term planning.", "methodology": "A comparative study was conducted using operational data from multiple industrial sectors. The core methodology employs a Bayesian hierarchical model specified as $y{it} \\sim \\text{Normal}(\\alphai + \\betat, \\sigma^2)$, with $\\alphai \\sim \\text{Normal}(\\mu{\\alpha}, \\tau{\\alpha}^2)$ and $\\betat \\sim \\text{Normal}(\\mu{\\beta}, \\tau_{\\beta}^2)$, where $i$ indexes fleets and $t$ indexes time. Hamiltonian Monte Carlo sampling was used for inference, with posterior predictive checks for model validation.", "findings": "The model indicates a positive trajectory for aggregate fleet efficiency, with a mean posterior annual gain of 2.3% (95% credible interval: 1.7% to 2.9%). The analysis reveals significant sectoral divergence, with the processing sector showing markedly slower efficiency convergence compared to manufacturing. The hierarchical structure effectively captured unobserved heterogeneity, with robust standard errors confirming model stability.", "conclusion": "The Bayesian hierarchical framework provides a superior, probabilistically rigorous tool for analysing machinery fleet efficiency, directly quantifying uncertainty in a way traditional methods cannot. It offers a validated approach for strategic asset management in industrialising contexts.", "recommendations": "Industrial policymakers and fleet managers should adopt probabilistic forecasting models for capital planning. Future research should integrate real-time sensor data