Iron ore, a major raw material for steel production, plays a critical role in infrastructure development, with Australia contributing 38 per cent of global supply (Holmes, Lu and Lu, 2022; Lundaev et al, 2023). The beneficiation of magnetite, one of the most prevalent iron oxide minerals, relies heavily on wet drum separators (WDS) to separate magnetic from non-magnetic fractions. Despite decades of industrial application, WDS operations face persistent performance challenges. Recovery rates fluctuate between 75 per cent and 95 per cent, even within the same operation (Svoboda and Fujita, 2003) yet the mechanisms driving this variability remain poorly understood. Efforts to improve WDS performance have followed two approaches. The first focuses on advancing separator designs through innovations in magnetic configurations. Whilst these developments show promise in controlled settings, new equipment cannot resolve process issues if the interactions between slurry properties and separation mechanisms remain poorly understood. The second approach emphasises process optimisation through adjustment of operational parameters (Chen et al, 2024; Dworzanowski, 2010, 2012). However, these studies examine individual parameters in isolation – particle size effects (Arol and Aydogan, 2004) or feed rate impacts (Wang et al, 2022) – without considering how multiple slurry properties collectively influence performance. Recent works have applied modelling techniques to investigate the influence of slurry properties on separation efficiency (Tian et al, 2025; Zhao et al, 2026). While these studies represent progress in understanding WDS behaviour, further investigations remain necessary to elucidate how interactions between slurry properties influence WDS performance. Furthermore, models developed from industrial data sets face challenges in validation due to temporal inconsistencies, changing standards and instruments, changes in operators, raw material variations, and seasonal effects. These factors raise questions about whether models developed from historical plant data can reliably predict WDS performance under controlled or varying operational conditions, limiting confidence in their industrial application. This study addresses these challenges by developing and validating a predictive model for WDS performance using complementary data sources. The study aims to: (1) develop a predictive model for the effects of four feed slurry variables (particle size (F80), SiO2, Al2O3, and total Fe grades) on WDS separation efficiency using historical plant data; (2) validate the developed model using independent survey data collected under controlled conditions. This approach ensures that the resulting models are both statistically rigorous, providing confidence that performance predictions derived from historical data can be reliably applied to guide operational decisions and process optimisation.