Alibaba held grocery orders briefly so two baskets could become one parcel

Online grocery fulfillment is expensive because speed pushes each order into the system immediately. But many customers place multiple orders within a short interval, creating duplicate parcels that could have shipped together.

Alibaba's answer was selective patience. Its system predicted real-time multiorder probability, then used a dynamic program to decide release timing so likely pairs could be consolidated without broadly slowing service.

IDEAS records the article's estimates: the scheme captured 92.8% of multiorders while holding orders for about 20.3 minutes on average, translating to more than USD 10 million in annual fulfillment-cost savings.

The clever part was not just batching. It was making waiting conditional, so the operation held only the orders where delay had a high chance of turning into a cheaper parcel.

Delay only the orders likely to merge

The model separated useful waiting from blanket delay.

Consolidation reduced parcels and shipping cost while preserving the normal fulfillment process.

The dynamic program converted a prediction into an operational release rule.

A little delay can improve speed economics when it is targeted at orders with high consolidation odds.

The work appeared in INFORMS Journal on Applied Analytics in 2024. It shows that the fastest supply chain is not always the smartest one: sometimes the valuable move is to pause just long enough for demand to reveal itself.