Intertemporal Demand Allocation for Inventory Control in Online Marketplaces

Researchers Rene Caldentey and Tong Xie have published a significant paper, 'Intertemporal Demand Allocation for Inventory Control in Online Marketplaces,' that reveals a hidden mechanism of control in digital platforms. The study investigates how marketplaces like Amazon or Alibaba, which increasingly offer fulfillment services (Fulfill-by-Platform, or FBP), can influence sellers' inventory decisions without directly controlling their stock. The core finding is informational: by strategically routing customer orders over time, a platform can change the predictability of a seller's sales stream. This directly impacts the seller's perceived need for safety stock, as unpredictable demand forces sellers to hold more inventory as a buffer.

The key mechanism is that platforms can implement simple, low-memory allocation rules to create varying levels of forecast uncertainty for sellers. For example, a uniform splitting of orders minimizes uncertainty, while other rules can increase it. Crucially, to increase uncertainty, the platform must use routing rules that prevent sellers from inferring total marketplace demand from their own sales history. This reduces the platform's strategic problem to choosing an optimal level of uncertainty that trades off two goals: encouraging sellers to adopt the platform's more profitable FBP service, versus managing how much inventory those adopting sellers decide to hold. The paper frames demand allocation as a powerful dual-purpose tool for operational and informational design in multi-agent marketplace systems.