Getting product returns right in retail
We value making things right. Yet when something goes wrong, the next step can still be difficult to work out.
Think of something you bought that did not work out. Maybe it was damaged, missing a part, or simply not right for your home. You want to know what to do next. Instead, you may find yourself searching through a policy, trying to work out which conditions apply, and wondering whether to call customer service or make the trip to a store. It is easy for us to miss how difficult that moment can feel. We know our policies, processes and exceptions. Customers encounter them when they are already trying to solve a problem—and may not know which options are available or what conditions apply. Even a generous policy can feel complicated if the next step is hard to find. That matters because a return is not just a transaction. It is a moment when people need clarity and confidence. Someone with a damaged chair, for example, may not know whether to ask for a spare part, return the whole product, or contact us for help. The right answer can depend on the product range and the situation, but people should not have to become policy experts to understand their options.
That is why we took on Return & Resolution. Instead of asking how we could automate a return, we asked how we could help people find and understand a suitable resolution for their case.
In practice, the AI-powered guide let customers identify a product and describe the issue, including by sharing photos. It assessed the case and suggested relevant next steps, such as a return, a spare part, or further support. The intention was to help people get personalised guidance without having to start by calling customer service or travelling to a store
We explored the customer experience and business opportunity, then partnered with Pixelabs to develop and test the guide. After internal testing and customer interviews, we ran a summer pilot on IKEA’s Spanish returns website and invited customers to share feedback.
Customers completed 1,100 flows during the pilot. That showed the tool was used, while feedback helped us see where the experience could be clearer and more reassuring.
Not every interaction met every customer’s expectations. The comments also showed where a solution like this could improve. “The tool does not allow me to proceed with the return,” wrote one customer. Another felt the assessment had missed important context: “I expected you to take into account that I did not use it; the photos show that it has not been used.” And one customer wanted more room to explain: “There is no way to write and explain the case in more detail. It is a very general robotic system, and it is not possible to describe the customer’s exact problem with the product.”
The guide performed with very high accuracy in validation, but technical accuracy alone was not enough. Return rules can vary by product range, and customers may not know the alternatives or the conditions that apply. If a recommendation is hard to understand, people may still doubt it—even when the tool has reached the right answer. The guide needs to do more than tell people what they can or cannot do: it should explain why, make the options easier to understand, and help customers feel confident about what happens next.
The core learning is that a resolution feels right not only when the recommendation is accurate, but when people understand and trust it. A helpful tool should guide, explain and reassure—not just decide.
Next time something you bought does not work out, pause before assuming a return is the only answer. Ask what would make the situation right for you.
Outcome / next step: The project delivered a tested tool and recommendations for its further development. Customers completed flows during the pilot, but a decision about whether or how to deploy the tool has not yet been made.
insight
news
More from the workbench. Explore the projects, ideas and stories we’re excited to share.
