Early in the COVID-19 pandemic, Columbia Business School Visiting Professor Liat Hadar found herself doing what millions of people stuck at home were doing: shopping online. Hadar, then home with her young daughters, was browsing a website where she liked to buy them dresses. She carefully picked out her favorites and added them to her shopping cart. But when it came time to check out, she paused.
“Wait a minute, why do I even need that?” she remembers thinking. Her daughters weren’t going anywhere. When would they possibly wear fancy new dresses? Hadar abandoned the cart, but the experience prompted her to wonder what it is, exactly, about the contents of a cart that influence whether a shopper ultimately completes a purchase.
That question became the basis for a recently published study by Hadar and co-authors Yael Steinhart and Yaniv Shani, both of Tel Aviv University, and Gil Appel of George Washington University. Their study finds that shoppers are more likely to abandon online carts containing a higher proportion of items designed primarily for pleasure than products designed for more practical purposes.
The reason, the researchers determined, comes down largely to guilt. And the findings have potentially far-reaching implications for how retailers design the online shopping experience.
Looking inside the cart
Online cart abandonment is remarkably common: more than 70% of online shopping carts are abandoned, according to industry research cited in the paper. Previous studies have examined a range of reasons why that happens, from unexpected costs and slow delivery to the different ways consumers use online carts. Researchers have shown that shoppers may treat a virtual cart as a kind of digital consideration set, collecting products they are still evaluating rather than those they have firmly decided to buy. But the impact of cart composition on the decision to buy has received far less attention.
Hadar and her team focused on the distinction between what they call hedonic products, which provide pleasure or enjoyment, and utilitarian products, which primarily fulfill a practical need. A perfume or box of chocolates would fall on the hedonic end of the spectrum, while trash bags or a vacuum cleaner would fall on the utilitarian end.
The researchers combined two real-world datasets with four controlled experiments. The field data encompassed nearly 15 million items purchased or abandoned across different e-commerce websites and product categories. In the experiments, the researchers manipulated cart composition and tested whether it caused changes in shopper behavior.
Across the field data, the pattern was consistent: as the proportion of hedonic products in a cart increased, so did the likelihood that the shopper would abandon it. The experiments provided an explanation. Shoppers appeared to mentally combine the items in their carts into an overall assessment of how indulgent the purchase felt. A cart weighted toward hedonic goods felt more hedonic overall. That produced more guilt, which in turn made abandonment more likely.
Rethinking the recommendation engine
Perhaps the most surprising finding came when the researchers explored whether retailers could counteract the cart composition effect. Recommendation engines commonly suggest products like ones shoppers have already viewed or purchased. But if someone has put an indulgent item in their cart, recommending another indulgence may be precisely the wrong approach.
In experiments, recommending a utilitarian product reduced guilt and made shoppers less likely to abandon hedonic items — even when they didn’t add the recommended product to their cart. In other words, merely considering something practical appeared to alter how shoppers perceived the purchase as a whole.
That suggests retailers may want to think about recommendations not only in terms of what a shopper is likely to buy, but also in terms of how a suggested product changes the psychological context surrounding everything already in the cart.
“If someone puts a candle in their online cart, you might not want to offer another candle in a different scent,” Hadar says. Instead, she suggests that a retailer might recommend something more utilitarian, like a pet hair remover.
According to Hadar, AI could make such an approach relatively straightforward. The researchers themselves used an AI model to help classify product categories as hedonic or utilitarian, and Hadar says retailers could potentially incorporate similar classifications into their recommendation algorithms.
According to Hadar, the lesson isn’t that retailers should steer shoppers away from indulgence altogether. After all, she acknowledges, “People want to feel good.” Rather, the research suggests that retailers need to better understand the psychology surrounding those indulgences and recognize that what appears alongside them can change whether consumers ultimately feel comfortable clicking “buy.”
