Its catalog includes approximately 30 million items across approximately 1,000 products. Approximately 47,000 attributes cover details such as color, material, and dimensions. These details determine how your products appear in search results and recommendations. Before adopting artificial intelligence (AI), correcting bad listings relied primarily on suppliers and shoppers reporting errors. Manual review was not possible for a catalog of this size.
Through its partnership with OpenAI, Wayfair has fixed 2.5 million product attribute tags and now automatically processes 41,000 supplier support requests each month, the AI startup shared in a March post. This work includes two jobs that previously relied on people. One is the merchandising staff who checks product listings for accuracy, and the other is the supplier support team who sorts through requests one at a time.
Wayfair has built a system that reviews product listings against explanations of what each detail means and flags and corrects mistakes. If your coffee table is listed as walnut, but the photo and description show pine, the system will detect it.
OpenAI added that the system is running on more than 1 million products to date. OpenAI says that when they tested the revised listings, they received more clicks and were ranked higher in search than before.
Wayfair did not allow the system to make changes without checking. Staff will manually inspect a sample of the revised list. If the system is confident of the fix, it will update the list and notify the supplier. If not, ask your supplier to confirm the change first. Wayfair said it plans to expand the program to four times as many products within six months.
Advertisement: SCROLL TO CONTINUE
PYMNTS reports that by using Google’s AI models to categorize products, Wayfair has already reduced the time it takes to curate listings by 67% and improved some conversion rates by 2%.
Regular requests from suppliers can be handled without staff
Wayfair works with tens of thousands of suppliers, and its support team reads every request it receives, understands what the supplier needs, and sends it to the appropriate internal group. According to Wayfair’s supplier support team, requests span hundreds of different issue types, more than any one staff member can track.
Currently, the system reads each request, retrieves the missing details from Wayfair’s internal records, and sends it independently to the appropriate team.
For some teams, this system is even more advanced. One example covers requests for replacement parts, reviewing the case history and creating a suggested response for a staff member to review before submission. Wayfair checks how often these suggestions match what our staff would do. For tasks with consistently high match rates, the system will work without a staff member reviewing them first. Through these efforts, Wayfair said it now automatically processes up to 70% of its ticket volume in some workflows.
Catalog cleanup becomes a bigger target for AI
Wayfair’s approach reflects a broader shift in how retailers apply AI, including adding chatbots for shoppers as well as cleaning up product data and back-office workflows. In January, Microsoft introduced a tool for retailers that captures product details from photos and automatically fixes catalog errors, with apparel brand Guess as an early user.
Wayfair also made its product data available to third-party AI tools that shoppers use to search and compare products. These tools, including ChatGPT, can pull directly from retailers’ product feeds to answer shopping questions and display products in response to searches. This means that products modified through Wayfair’s internal systems will be displayed more accurately when shoppers ask the AI assistant for product recommendations.
For all of our coverage of PYMNTS AI, subscribe to our daily subscription AI Newsletter.
