{ "background": "Process-control systems in manufacturing are critical for operational efficiency and yield optimisation. In many developing industrial contexts, such as those in East Africa, diagnostic tools for yield improvement are often limited by sparse, heterogeneous data and the need to quantify uncertainty in system performance.", "purpose and objectives": "This Data Descriptor presents a methodological evaluation dataset constructed to assess a novel Bayesian hierarchical model for diagnosing yield improvements in industrial process-control systems. The objective is to provide a reusable, structured dataset that enables the validation of the proposed model's capacity to isolate process-level effects from plant-level variability.", "methodology": "The dataset was synthetically generated to reflect realistic operational parameters from multiple manufacturing plants. It incorporates structured noise, planned interventions, and covariates. The core statistical model is defined as $y{ij} \\sim \\text{Normal}(\\alphaj + \\betaj x{ij}, \\sigmay)$, with $\\alphaj \\sim \\text{Normal}(\\mu\\alpha, \\sigma\\alpha)$ and $\\betaj \\sim \\text{Normal}(\\mu\\beta, \\sigma\\beta)$, where $i$ indexes observations and $j$ indexes plants. Posterior distributions were estimated using Hamiltonian Monte Carlo.", "findings": "Methodological evaluation using the dataset indicates that the hierarchical model successfully partitions variance, attributing approximately 65% of yield variation to plant-level random effects. Crucially, the 90% highest posterior density interval for the global intervention coefficient $\\mu\\beta$ was [0.15, 0.42], excluding zero, which demonstrates the model's inferential capability to detect a positive systematic effect despite heterogeneous baseline performance.", "conclusion": "The constructed dataset provides a robust foundation for evaluating the proposed Bayesian hierarchical diagnostic model. It confirms the method's utility in delivering nuanced, probabilistic insights into yield drivers within complex, multi-plant industrial settings.", "recommendations": "Future applied research should utilise this dataset as a benchmark for comparing diagnostic models