Shipping costs e-commerce companies up to 20% of their revenue—not to mention its contribution to air pollution, global warming, and packaging waste.
But making shipping more efficient has been challenging, says Harvard Business School Assistant Professor Aliya Korganbekova. “It’s a very difficult optimization problem,” she says. “These products come from different warehouses, from different sellers, and the various departments inside organizations are focused on different goals.”
Online shopping has been growing rapidly for several years, a shift accelerated by the COVID-19 pandemic, as millions of consumers have become accustomed to the convenience of having goods delivered to their homes. That growth has made shipping a bigger operational and environmental challenge for retailers, who often bear the cost of getting products to customers.
Korganbekova and her coauthors together with the data science team of a large US company, tested a new approach to reducing shipping costs and emissions without sacrificing sales. They designed an algorithm that ranks products partly by how much they cost to ship. In a study involving millions of shoppers, the approach reduced expenses by 4.8% and delivery-related emissions by 3.8% per order, while slightly boosting profits.
Aliya Korganbekova wrote the May working paper “Ranking Algorithms for Shipping Cost Optimization” with her sister Malika Korganbekova, an assistant professor at Stanford University’s Graduate School of Business, and her brother Alibek Korganbekov, an assistant professor at Indiana University’s Kelley School of Business.
Finding a solution to expensive shipping
Aliya Korganbekova was spending time with her siblings when she recalls, “We were getting all these orders from different retailers, and sometimes we would order different things, and they would come in different packages within one day, so we started to think: ‘Wouldn't it be more efficient if they would bundle these together?’”
Getting that item overnight comes at a price, one that retailers typically absorb. In 2025, for instance, Amazon reported nearly $103 billion in shipping costs, about 14% of its revenue—twice the share it represented a decade earlier. Some 53% of the expense comes from the last leg of the journey, as a product makes its way from warehouses to sorting facilities and finally to doorsteps.
A lot of companies have set goals to reduce shipping costs and the emissions that go along with them but have mostly focused on the supply side of the equation, often by building more warehouses. The researchers were curious whether they could instead tackle the problem by adjusting the websites that determine which products customers see first when they shop online.
The team partnered with an e-commerce retailer to design an algorithm that shows items to shoppers based in part on shipping costs. The company has 5 million products, 50,000 sellers, and 150,000 seller warehouses, and the researchers tested their algorithm over eight days on 2.8 million consumers.
To build the algorithm, researchers incorporated a cost penalty based on the product’s size, the consumer’s location, the warehouse’s location, and the carrier’s rate. Products that were cheap and easy to ship got a smaller penalty, while products that were expensive to ship got a larger one. However, products that customers strongly preferred earned some protection from that penalty.
Costs shrank and profits increased slightly
The researchers randomly assigned shoppers to use either an established algorithm or the one they built, finding their platform:
Significantly shortened shipping length. Delivery distances dropped by 34 miles per order, creating annual savings comparable to the electricity consumption of 50,000 US households or emissions from 58,000 gas-powered cars.
Reduced shipping expenses without impacting the customer experience. While shipping costs dropped 4.8%, conversions and customer satisfaction remained steady and profit grew 0.2%.
Shifted more orders closer to consumers. Centralized fulfillment—which is less expensive—increased by nearly 13%.
The researchers found that the effect depended on the type of product category. In non-branded categories, as a result of the implemented algorithm, orders rose by 0.1% and profit by 0.4%. In branded categories, where shoppers care more about the brand name and are, therefore, less willing to switch to a substitute, orders fell by 0.3%.
What’s stopping e-retailers from optimizing shipping?
Guiding consumers to products that are less expensive to ship at the search stage makes sense, says Korganbekova. But she points to three barriers that may realistically make the idea difficult to scale:
Different teams often make decisions that inadvertently work against one another. For example, shipping cost information may live in a system that’s not available to marketing teams that build ranking algorithms, so the two departments may end up working toward different goals. “Decentralized decision-making in retail creates real economic losses,” the paper explains. “When different teams optimize independently, the lack of coordination produces inefficiencies that joint optimization could potentially eliminate.”
The technology needed to coordinate information is relatively new. Until recently, it hasn’t been possible to coordinate this level of data without slowing down page loading times, which is critical to survival in the industry. Some organizations may still be catching up by implementing the correct technology.
Companies are concerned about disrupting sales. If consumers don’t want the algorithms they rely on to change, retail companies are reluctant to dedicate the resources to a project that could potentially hurt sales.
What business leaders can do
Korganbekova argues that business leaders should look beyond traditional supply-side solutions and consider how making demand-side changes, using data judiciously, and altering organizational structures can help improve results. She provides three pieces of advice:
Consider that reducing environmental impact could be less expensive
Companies often equate sustainability with extra costs, says Korganbekova, but “here, the financial and sustainability components work together. Finding cases when that happens is important because when companies can understand that they can save money but also improve their environmental performance, there will be more options to scale such approaches.”
Use data wisely to make system improvements
So much data is available now that companies run the risk of information overload. “Can we use the data that companies and managers have, which they built for a long period of time, and [choose the relevant information] to embed it into some of the existing processes?” asks Korganbekova.
Avoid optimizing different departments in isolation
Instead of delegating tasks to specialized teams, each with their own objectives and timelines, Korganbekova suggests companies look across units to find common objectives and make sure teams work together on goals. “The data can connect different departments,” she says.
Image by Ariana Cohen-Halberstam, created with photo from AdobeStock.
