Data-driven optimal dynamic pricing strategy for reducing perishable food waste at retailers

Approximately forty per cent of fresh products are wasted in low and middle-income countries before reaching consumers. Perishable foods have only a certain shelf-life and they need to be sold for consumption before a specific date. When a product is priced incorrectly, it is often disposed of directly or redistributed. Redistribution of surplus food also has an economic impact on food prices. Therefore, setting an optimal pricing strategy is crucial to reduce inventory and surplus food in an environment with volatile demands. In this context, big data analytics can help managers forecast customer behaviour and determine pricing strategies throughout the retail industry. This study focuses on food waste at the retailer stage of food supply chain (FSC). We present a dynamic pricing model that uses real-time Internet of Things (IoT) sensor data as a novel contribution to decide pricing at different stages of a sales season for retailers. The food waste problem at the retail stage of a FSC is investigated in a pilot project for bulk apple sales to address the research question. This study proposes a four-stage data-driven optimal dynamic pricing strategy for bulk produce to reduce food waste for retailers in low and middle-income countries. A multi-stage dynamic programming method is used to decide on a pricing strategy for bulk produce, with real-time IoT sensor data being retrieved to analyse and determine the length of freshness scores. The effect of the sale price, replenishment amount, discount rate, and freshness score on profit and food waste are evaluated. All these analyses assist managers in taking the best possible actions and remedies. Appropriate interventions boost sales, increase profits by reducing waste and determining competitive sales price, while improving customer loyalty and satisfaction by striking the right balance between food quality and price. Our results show the huge potential of using hyperspectral imaging sensors in the FSC of a retailer. The model is demonstrated empirically to test its practicability.