How e-commerce platforms should deal with multiple orders

Aminian, Ma and Xin proposed three algorithms and analyzed them in the model. The first is the order’s immediate “reward rate,” which is the shipping cost savings earned in proportion to the customer’s multi-order rate.

Let’s assume that Customer A orders socks on Amazon. Let’s assume that the maximum order processing time is three hours, there is one fixed shipping cost for each shipment, and Amazon can only fulfill one order. There is a 50% chance that Customer A will order something else in the next few hours. Now let’s assume that Customer B orders a coffee machine two hours later. If past behavior suggests that Customer B has only a 20 percent chance of placing another order in the next few hours, Amazon may pay to ship the coffee machine and keep the socks on for another hour, especially if the seller’s primary concern is maximizing the immediate reward rate.

This is the model’s “most conservative” algorithm for deciding whether to further hold an order, Aminian says, and works best when the platform is very busy. Scientists have found that when demand is high, it is optimal to seek immediate benefits and take fewer risks.

However, the disadvantage of the first algorithm is that it ignores how long each order has been in the system. The second algorithm aims to maximize the total remaining reward for each customer. Consider the same scenario. Since the maximum wait time is three hours, Customer A’s socks order must be completed within an hour before Customer B purchases the coffee machine. However, Customer B’s order could have been held for a full three hours longer.

Based on the potential return and the remaining time, the second algorithm would first send the socks and stop the coffee machine. This is because the three-hour window is perceived as an advantage over the remaining one-hour window, even though Customer B is less likely to place a second order.

This is the “least conservative” algorithm, Aminian says. “He is the most changeable – he tries to change as often as possible whenever he sees a chance.” This would be the best solution for a very small number of orders, he says.

The third algorithm aims to fall somewhere in the middle, managing a wide range of ordering situations. For example, when order demand is moderate, it tries to strike a balance between immediate and total remaining rewards and hold or ship orders accordingly. The researchers found that this strategy would also be effective – although not optimal – when the volume is high or low. A retailer trying to cover all situations without changing strategy may prefer this option.

The researchers note that their model could further reduce costs by recording order modifications or cancellations before delayed shipments are made. Beyond e-commerce, the researchers suggest it could be useful for cloud computing platforms managing capacity, hospitals deciding on discharges based on patient urgency, and businesses deciding whether to outsource new or ongoing projects when they run out of capacity.

In addition to this research shedding light on key trade-offs, these algorithms could “affect the decision-making efficiency of such companies under future uncertainty by increasing prosperity in healthcare, gains in cloud computing, and savings in fulfillment costs for e-commerce retailers” – says Aminian.