United StatesRetail and e-commerce2024 – ongoingWalmart
Generative AI for product catalog data
Walmart used large language models to create or improve more than 850 million catalog data points
Problem
Walmart's online catalog was growing with more marketplace sellers and items. Missing or poor product attributes hurt how customers found products and how stores stored and delivered them.
Approach
The company applied several large language models, some from outside providers and some built in-house, to populate product attributes and improve product pages. The cleaner data also feeds store associates' mobile tools, which show packaging images and item locations when they pick orders or stock shelves.
Results
Takeaway
Walmart used the improved catalog data in both product pages and staff tools for finding and stocking items. The "100 times" staffing comparison was an estimate, and the company disclosed no error rates or costs.
Categories
- Size
- Large enterprise
- Industry
- Retail and e-commerce
- Business function
- Internal work and knowledge managementMarketing and sales
- Innovation type
- Process innovation
- AI technology
- Generative AI
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
Walmart chief executive Doug McMillon said in the fiscal second-quarter earnings call that multiple language models created or improved more than 850 million pieces of catalog data. He estimated that the same work would have needed nearly 100 times the current headcount in the same period. Walmart U.S. chief John Furner adds that AI filled in attributes for hundreds of millions of items. Two outside analysts comment that catalog work is a narrow, well-defined task suited to AI.
Retail Dive reported that large language models created or improved more than 850 million product data points. Walmart executives estimated that the work would otherwise have needed about 100 times the headcount. Catalog quality affects search, inventory storage and delivery, and staff used the cleaner data in mobile tools to find products and view packaging images.
Full details are in the original sources.
