Wayfair Improves Catalog Accuracy and Support Speed by Adopting OpenAI

Wayfair integrated OpenAI models into its vendor and catalog systems to improve data accuracy for millions of products and automate workflows. Starting as a small-scale test in 2024, the project has now grown into a large-scale production system managing tens of millions of product attributes.

In particular, Wayfair focused on consistently managing tens of thousands of attribute tags—such as color, material, and size—across about 30 million product listings. Previously, this relied on manual work or individually customized AI models, but managing all 47,000 tags this way had scalability limits.

To solve this, Wayfair built a tag-agnostic system based on a single OpenAI model. A 'definition agent' learns from web and internal definitions to generate the contextual meaning of each tag, which has increased the speed of applying new attributes by 70x compared to a year ago.

As a result of adopting the system, 2.5 million product tags were corrected, and 41,000 vendor support tickets per month were automated. In addition, improved data quality led to better SEO and PLA performance, confirming positive effects across the entire customer journey.

When correcting data, a human review process runs in parallel to ensure reliability. High-confidence data is updated automatically, while high-risk tags are changed only after vendor confirmation, ensuring data accuracy.