Whatnot has acquired Shaped, an artificial intelligence company specializing in recommendation and ranking infrastructure, to strengthen product discovery across its live shopping platform. Financial terms of the transaction were not disclosed.
Shaped’s team of engineers and researchers will join Whatnot and focus on developing AI systems for live commerce, where inventory, buyer behavior and marketplace activity can change continuously throughout a broadcast.
Whatnot is a live shopping marketplace that enables sellers to host interactive video shows, demonstrate products and sell items to buyers in real time. The company describes itself as the largest live shopping platform in North America and Europe.
Shaped develops technology that helps marketplaces and content companies determine which products, listings or experiences should be shown to each user. Its infrastructure has been used by hundreds of companies seeking advanced personalization without building large internal machine learning teams.
The company was founded around the idea that any organization with a product catalog and a stream of customer activity should be able to access sophisticated recommendation technology.
Traditional recommendation systems often analyze historical behavior and relatively stable inventories to determine which products should be ranked most prominently. Live commerce presents a more complicated challenge because listings, prices, availability, and audience engagement can shift from minute to minute.
A Whatnot seller may introduce new products throughout a show, while viewers continuously place bids, make purchases, send messages or move between broadcasts. Each interaction can change what content should be recommended next.
Shaped said these conditions require a different approach to retrieval, ranking, evaluation, and exploration than the systems generally used by conventional e-commerce platforms.
Recommendation technology in live commerce must consider both immediate buyer intent and the longer-term health of the marketplace. A system may need to identify the product or show the most likely to generate a purchase while also helping users discover new sellers, categories, and communities.
The acquisition gives Shaped’s team the opportunity to develop these systems directly within a large live marketplace rather than serving the industry solely as an outside technology provider.
Whatnot processes more than 500,000 hours of live video and millions of real-time interactions each week. This scale provides Shaped’s researchers with a substantial volume of behavioral and marketplace data to test and improve discovery systems.
Cross-category purchasing on Whatnot has also increased 170% year over year, indicating that users are increasingly buying products outside the categories that initially brought them to the platform.
Improved AI-powered discovery could help accelerate this trend by identifying relationships between buyer interests that may not be obvious from individual searches or previous purchases.
For example, a customer who primarily watches sports card auctions may also be interested in memorabilia, vintage apparel or other collectible categories. A more advanced recommendation system could recognize these connections and introduce relevant sellers or shows at the appropriate time.
The acquisition could also help Whatnot personalize the sequence of live broadcasts, product listings, and notifications presented to each user.
Because live inventory is temporary, the system must make recommendations quickly. A show or product may be available only for a limited period, reducing the usefulness of slower recommendation processes that rely primarily on historical data.
Shaped’s team will work on AI systems that respond to live signals while balancing short-term conversion opportunities with sustained engagement and marketplace diversity.
These systems could help buyers find relevant products more efficiently while giving sellers greater exposure to audiences likely to engage with their shows.
The transaction also reflects the growing importance of artificial intelligence in live shopping. As platforms add more sellers, categories and broadcasts, users can face an overwhelming number of options.
Effective ranking and recommendation systems can reduce that complexity by identifying the most relevant experiences from a continuously changing marketplace.
Whatnot can also use AI to evaluate whether recommendations are improving broader marketplace outcomes, including customer retention, seller growth, repeat purchases and activity across multiple categories.
Shaped’s team believes many of these live-commerce discovery challenges remain open areas of AI research. Joining Whatnot gives the company access to a real-time environment where new approaches can be deployed and evaluated at significant scale.
Shaped’s customers will continue receiving support during the transition. The company said there are no immediate changes to its service and that it is contacting customers directly regarding future plans.
The acquisition represents an exit for Shaped’s investors, which include Madrona and Y Combinator.
By bringing Shaped’s recommendation expertise in-house, Whatnot aims to create more personalized shopping experiences while improving how buyers discover sellers, products and communities across its platform.
KEY QUOTE:
“When we started Shaped, the thesis was simple: every company with a catalog and an event stream deserves state-of-the-art ranking, not just the ones that have huge ML teams and AI labs. Over the past few years, we built that platform, and it’s been used by hundreds of marketplaces and content companies. Whatnot is the most interesting place in the world to work on the live marketplace discovery problem. Rather than serving that problem from the outside, we decided to go work on it directly, at the company defining the future of live commerce.”
Tullie Murrell, CEO of Shaped
