A mean-variance optimisation model for produce export distribution with postponed shipment and market allocation options
针对生鲜农产品出口商面临的价格波动问题，提出一个基于两阶段随机优化的决策工具，通过分散市场、分批发货和延迟分配来降低风险，并用智利葡萄出口数据验证了模型优于传统方法。
Global exporters of fresh agricultural produce must determine where and when to ship their goods while facing significant price volatility in their sales markets. Unlike non-perishable commodities, agricultural produce cannot be relocated once shipped, eliminating any opportunity for arbitrage across markets. To address this problem, we present a decision-support tool for fresh produce exporters based on a two-stage stochastic optimisation model that extends Markowitz's Mean–Variance model. Our proposed model implements three risk-mitigation strategies: distribute shipments across multiple markets, use two shipping periods, and delay a fraction of market allocation decisions. Diversifying exports across markets and shipment times enables the exporter to capitalise on negative price correlations, while delaying allocations reduces uncertainty with updated price information. We empirically tested our approach using real Chilean grape export price data to simulate price scenarios and compare our model's results with two benchmark models: one that assumes one shipment period, and another that allows two shipment periods but decides all allocations in advance, without adapting to newly revealed information. Our model significantly outperforms both benchmarks, indicating the value of the three risk-mitigation strategies. Furthermore, as inventory costs increase, the value of postponement decreases; also, having a wider range of markets, with negative correlation, adds value.