Messages about the environmental impact of returns reduced return shipments by 2.6% in a randomized experiment involving 117,304 online fashion shoppers, researchers found.
The study suggests that information presented during shopping can influence returns without necessarily requiring a retailer to restrict its return policy. The finding comes from one European fashion retailer, however, and does not establish that similar messages will work equally well elsewhere.
Published online in 2024 in Marketing Science, the research examined both a tested message and a separate model that estimated whether more selective messaging might improve the result.
What customers saw
The researchers worked with the retailer to show information about the environmental consequences of returning products during the shopping journey and again near its conclusion. Participants in the field experiment were randomly assigned, allowing the researchers to compare behavior with a control group.
The reported 2.6% reduction was relative to the comparison group, not a 2.6 percentage-point fall in the return rate. The researchers reported no adverse effect on sales in the experiment.
The authors said further analysis suggested that part of the reduction came from less “bracketing”, ordering multiple versions with the intention of sending some back. The study measured short-term behavior and cannot establish whether shoppers would respond in the same way over a longer period.
Why returns matter to retailers
Returned goods can generate transport, handling and inspection costs. Some products need to be repackaged, discounted or otherwise processed before resale. The environmental effect depends on factors including transport, packaging and what happens to the returned item.
Retailers also need to consider the role that return policies play in helping customers buy with confidence. A person who cannot try on clothing before ordering may reasonably rely on the ability to return an item that does not fit.
The experiment tested an informational prompt, not the removal of return rights. Better sizing guidance, accurate photographs and useful product descriptions may address different causes of returns and deserve separate testing.
Personalization was a model estimate
After the field experiment, the researchers used browsing and shopping-cart information to examine whether the message appeared to work differently for different customers. They developed a statistical approach known as causal machine learning to estimate variation in the effect of the prompt.
Their subsequent evaluation suggested that selectively showing the message could reduce returns further. That projection was an analysis of existing experimental data, not a second live experiment demonstrating the same benefit in a new campaign.
For businesses considering personalization, the distinction matters. A prediction can help identify an idea worth testing, but it does not guarantee that a future system will produce the predicted savings. Using browsing data also calls for attention to privacy, consent and applicable law.
How a retailer could assess the idea
A business can compare a clear, accurate message with an otherwise similar shopping experience without the message. Useful measures include return shipments, completed purchases, revenue, processing costs and customer complaints. The test should be large enough and run long enough to produce meaningful evidence.
Messages should not imply that every return is avoidable or that customers should keep unsuitable products. A helpful approach gives shoppers information without obscuring their rights or making the returns process unnecessarily difficult.
The results complement research on the sales effects of retail return policies. Together they illustrate the trade-off: a policy can make purchasing easier, while better information may help customers make choices that lead to fewer unnecessary returns.