When Retail Buyers Panic, A.I. Keeps Calm
During the pandemic, algorithms overrode the human tendency to panic buy.
When Retail Buyers Panic, A.I. Keeps CalmCristina Spanò
A shopper fills her virtual cart with the items she wants, goes to check out, confirms the details, and clicks a button to complete the purchase.
Then what happens? The question is fundamentally important for retailers. Efficiently fulfilling customer orders—sending them from the right place at the right time—can generate huge savings.
One tactic that has emerged to capture these savings is a brief delay between when a customer completes a purchase and when the merchandise is actually picked from the shelf and prepared for shipment. Missing, though, has been a theoretical foundation to explain how and why this practice makes sense. Using data from one of China’s largest online retailers, JD.com, Chicago Booth PhD student Yaqi Xie, Columbia’s Will Ma, and Booth’s Linwei Xin built a model and conducted a series of experiments to estimate the benefits of delay.
They find that the potential savings can be attributed to more information about future demand: the longer a company waits to fulfill one customer’s order, the more it learns about other orders as they arrive.
Xin offers a hypothetical example of three Chicago-based consumers placing separate orders with Amazon. The first customer orders shoes and a jacket. Ten minutes later, one customer buys toothpaste and the same shoes, and the other purchases the jacket and candy.
Amazon has the choice of filling the first order from a smaller local warehouse or a bigger distant one. However, the local warehouse stocks only one unit of each item. Therefore, if it fills the first order locally, it has no choice but to route the second and third orders to the faraway warehouse.
“If you instead wait for a short period, then you can route the first order to a more distant warehouse and fill more orders locally,” he says, noting that this would reduce shipping costs. “The benefits of delay are about making better decisions with more information.”
The researchers emphasize that retail isn’t the only domain where delay is useful. Chess.com, for instance, matches chess players with similar skills, and intentionally lets players wait for a few seconds to increase the likelihood of pairing players with close ratings. Somewhat similarly, Uber uses a “batching algorithm” to match riders with drivers in groups as opposed to immediately pairing a ride request with the nearest driver. The researchers note that while Uber’s delay gives the company additional time to learn about incoming requests, it also allows Uber to increase the number of feasible matches it makes—because unlike Amazon, which has a pretty fixed inventory of products, Uber has an inventory of drivers that might grow while a rider’s dispatch order is delayed.
Analyzing one month of sales data from JD.com, the researchers find that a one-hour delay between customer purchase and fulfillment reduced costs by 1.4 percentage points—a substantial amount given the volume of online purchases. A four-hour delay led to a 3.8 percentage point reduction in costs.
The researchers point out that the value of delay erodes exponentially. Though longer delays provide greater savings, the marginal benefit decreases. In their calculations, the 1.4 percentage-point savings for a one-hour delay became only a 0.95 percentage point average hourly benefit when the delay extended to four hours. Also, companies can’t delay forever; overly long wait times clog up warehouses and annoy consumers. Companies must therefore look for the “sweet spot.”
Finding this sweet spot requires balancing savings and costs. While their model focused exclusively on how to fulfill orders efficiently, the researchers acknowledge that managers have to consider the context and other factors when determining how much delay makes sense, including the downsides of an idle warehouse and consumer dissatisfaction. But the general conclusion holds: a little bit of delay goes a long way.
Yaqi Xie, Will Ma, and Linwei Xin, “The Benefits of Delay to Online Decision-Making,” Working paper, November 2022.
During the pandemic, algorithms overrode the human tendency to panic buy.
When Retail Buyers Panic, A.I. Keeps CalmRegional transportation networks can blunt the impact of supply-chain disruptions.
Why Some Regions See Gas Price Surges After a Storm—And Why Others Don’tResearchers create a computationally efficient model for assemble-to-order operations.
Ever Closer to an Optimally Cost-Efficient Assembly-Line OperationYour Privacy
We want to demonstrate our commitment to your privacy. Please review Chicago Booth's privacy notice, which provides information explaining how and why we collect particular information when you visit our website.