MACHINE LEARNING-ENHANCED SUPPLY CHAIN RISK MANAGEMENT FOR DEMAND FORECASTING ACCURACY: EVIDENCE FROM THE EGYPTIAN FMCG SECTOR

Abstract

The name of the author who will actually present it at the Conference should be underlined. Traditional demand forecasting methods in fast-moving consumer goods (FMCG) supply chains increasingly prove inadequate for mitigating demand variability driven by supply disruptions, market volatility, and complex risk factors. While machine learning (ML) and supply chain risk management (SCRM) have emerged as distinct optimization strategies, empirical evidence integrating the two remains limited. This study examines whether ML-enhanced SCRM improves demand forecasting accuracy in FMCG operations. Using a sequential mixed- methods design, we conducted semi-structured interviews (n = 7 companies) and administered a structured survey (n = 55 professionals; 58.5% response rate) across Egyptian FMCG firms. Multiple regression analysis revealed that both ML integration (β = 0.20, p < .05) and SCRM maturity (β = 0.35, p < .01) significantly predict forecast accuracy (R² = .21, F(3,116) = 10.23, p < .001). Mediation analysis (Sobel z = 2.15, p < .05) confirmed that SCRM partially mediates the relationship between ML-forecast accuracy and the outcome. Cronbach’s alpha coefficients demonstrated strong internal consistency across constructs (α = .83-.89). Findings underscore that ML efficacy requires complementary risk governance structures; standalone algorithmic implementation yields suboptimal forecasting improvements. This study contributes a validated empirical framework linking predictive analytics to risk governance for FMCG managers and supply chain scholars navigating digital transformation.