Dual-stage procurement with forecast updating for seasonal products at Amazon.com
2023
We present a procurement model for highly seasonal products, such as toys or fashion products, which are usually purchased through a combination of import and domestic buys, months ahead of typical vendor lead times, with potential riskier buys just-in-time. This procurement process differs significantly from the more common just-in-time buying, or repeated dual sourcing, since the decision to split orders between import, domestic and just-in-time channels reflects a trade-off between cost and information gain. We propose a dynamic programming approach to the problem that captures the value of information gain by modeling the information gain through the Martingale Method of Forecast Evolution. We establish structural results of the optimal policies and present illustrative results. Experiments were conducted within a commercial system during the years 2018 and 2019, and directed to the toys category in anticipation of the Q4 peak season. It evidenced performance that would have enabled cost of goods purchased to grow by 24% without incremental retail headcount, and handled hundreds of millions of items during the course of the experiments.
Research areas