Simulation-Based Forecasting and Inventory Control Models For Consumer Goods Networks: A Quantitative Study Using Monte Carlo Simulation and Time-Series Methods

Abstract

This quantitative study examined the integrated effects of time-series demand forecasting and inventory control policies within consumer goods networks using Monte Carlo simulation. The study was designed to evaluate how demand uncertainty, forecast accuracy, forecast error dispersion, lead-time variability, and inventory policy structure jointly influenced service performance, stockout behavior, cost variability, and overall inventory stability. Historical demand data were modeled at the SKU level using time-series methods, and empirically estimated forecast error distributions were embedded into a simulation framework to generate repeated stochastic demand and lead-time scenarios. The analytical sample consisted of 312 SKUs observed over a median horizon of 104 weekly periods, yielding 32,448 SKU–period observations. Demand segmentation indicated that 51.9% of SKUs exhibited stable high-volume patterns, 30.8% displayed moderate variability, and 17.3% were characterized by intermittent demand with frequent zero observations. Descriptive results showed pronounced heterogeneity in demand uncertainty, with coefficients of variation increasing from 0.42 for stable SKUs to 1.76 for intermittent SKUs. Forecast accuracy deteriorated with horizon length and demand irregularity, with mean absolute error increasing from 9.6 units for stable SKUs at short horizons to 31.6 units for intermittent SKUs at longer horizons. Simulation-based inventory evaluation demonstrated that continuous review policies achieved higher median fill rates (0.963) and lower stockout frequency (0.42 stockouts per cycle) than periodic review policies, which exhibited lower median fill rates (0.941) and higher stockout frequency (0.88). Total cost variability, measured by the coefficient of variation, was lower under continuous review policies (0.31) than under periodic review policies (0.47). Regression analyses confirmed that forecast error magnitude, forecast error dispersion, demand uncertainty, and lead-time variability were statistically significant predictors of total cost, service performance, and inventory dispersion. Interaction effects indicated that forecast error impacts were amplified under higher lead-time variability and attenuated under continuous review policies. Overall, the findings demonstrated that distribution-sensitive forecasting evaluation and integrated simulation-based inventory analysis provided stronger evidence on service stability, cost dispersion, and policy robustness than approaches relying on point forecasts or mean performance metrics alone.