Demand Forecasting Models and Inventory Optimization Strategies in the Context of E-Commerce Promotional Scenarios

Abstract

E-commerce promotions can lead to significant fluctuations in traffic and sales in a short period of time, which poses higher demands for demand forecasting and inventory management. This paper focuses on demand forecasting models and inventory optimization strategies in the context of e-commerce promotions, proposing a feasible systematic solution. The solution firstly separates the trend and seasonality of baseline demand using time series methods, and then quantifies elements such as promotional rules, price trajectories, coupon distribution, traffic allocation, and external events into features introduced into machine learning models, in order to capture short-term nonlinear disturbances and adjust the baseline accordingly. Based on the forecast outputs and uncertainties, it designs layered stock replenishment, regional multi-warehouse allocation, and real-time allocation rules to balance the risks of stockouts and inventory holding costs. The paper further discusses key points regarding feature engineering, data update cadence, and system deployment, including automated data pipelines, short-cycle retraining trigger mechanisms, and organizational support for operational and supply chain collaboration. This solution emphasizes methodology and implementation pathways, aiming to provide actionable references for platforms to enhance forecasting reliability and inventory decision-making efficiency during large-scale promotional periods.