The encyclopedia · Software & IT · Operational decision · 2020–2024
JD.com chose daily assortment and stock levels to lift order fulfilment
JD.com built a system that jointly sets each front depot's assortment and daily allocation, lifting fulfilment and cutting costs.
JD.com
the move
JD.com fulfils orders from front distribution centres fed by regional warehouses, so the assortment each front depot carries and the inventory it receives daily determine how locally an order can be served.
The team built a decision-support system that integrates optimization and machine learning to decide front-depot assortments and the daily allocation from regional depots, replacing decisions that had been made locally and separately.
The system substantially improved order-fulfilment efficiency while reducing inventory and transfer costs, and won the 2024 INFORMS Daniel H. Wagner Prize.
why it works
- A local assortment ignores what the rest of the network can supply
- Assortment and allocation interact, so deciding them separately loses value
- Optimization plus machine learning matches stock to forecast demand
- Stocking close to demand cuts fulfilment distance and transfer cost
what transfers
Where you stock and how much you send each day are one problem; solving them together moves stock toward demand instead of shipping from farther away.
what came after
The decision support system is implemented across JD.com's inventory network in China. The work won the 2024 INFORMS Daniel H. Wagner Prize and was honored at the 2024 INFORMS Annual Meeting.
references
- Berkeley IEOR Shines at the 2024 INFORMS Annual Meeting
- Introduction: 2024 Daniel H. Wagner Prize for Excellence in the Practice of Advanced Analytics and Operations Research
spotted an error? The archive wants to know.