AI-Powered Demand Forecasting in FMCG Supply Chains: Integrating LSTM and External Market Indicators

Abstract

Demand volatility remains a critical challenge in Fast-Moving Consumer Goods (FMCG) supply chains, often leading to stockouts, excess inventory, and reduced profitability. Traditional statistical methods like ARIMA and exponential smoothing struggle to capture nonlinear patterns and the influence of external factors on consumer demand. This secondary data-based study proposes an AI-powered demand forecasting framework that integrates Long Short-Term Memory (LSTM) networks with external market indicators to enhance prediction accuracy in FMCG contexts. The research utilizes publicly available secondary datasets spanning 2019–2024, including SKU-level sales from NielsenIQ syndicated reports, weather data from the India Meteorological Department, commodity price indices from the Reserve Bank of India, and search interest data from Google Trends. An LSTM model with an attention mechanism is developed to process both historical sales and external variables. Performance is benchmarked against ARIMA and Facebook Prophet using Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and simulated inventory holding costs. Results demonstrate that the LSTM-external indicator model reduces MAPE by 24.6% compared to ARIMA and 17.3% compared to Prophet for 4-week ahead forecasts. Weather fluctuations significantly impacted beverage demand, while inflation indices showed a strong correlation with personal care and packaged foods. The model also exhibited superior stability during high-promotion periods and regional supply disruptions. This study contributes to commerce and management literature by validating the deep learning models using secondary data can deliver enterprise-grade forecasting without access to proprietary firm data. For FMCG managers, the findings offer a cost-effective approach to improve service levels and reduce working capital tied to inventory. The framework is scalable across geographies and adaptable to other retail verticals facing similar external demand shocks.