Whatnot Acquires Shaped AI To Cut Live Shopping Recommendation Lag
Serge Bulaev
Whatnot, a live-shopping platform, has agreed to acquire Shaped, a startup specializing in real-time recommendations, to make product suggestions much faster - possibly in milliseconds instead of minutes. Shaped's technology may help Whatnot adjust quickly as auctions and viewer interest change, which is important since items can sell just seconds after appearing. Experts suggest this deal shows that being able to personalize suggestions very quickly may be key for live commerce to grow. The companies have not shared financial details, but early tests suggest Whatnot might see bigger audiences and more engagement if Shaped's technology works as expected.
Live-shopping marketplace Whatnot has acquired Shaped AI, a move set to solve one of live commerce's toughest challenges: recommendation lag. The acquisition, detailed in a recent TechCrunch report, will integrate Shaped's real-time AI to slash product suggestion delays from minutes to milliseconds on the multi-billion-valued platform.
Why Whatnot Acquired Shaped AI
Whatnot acquired Shaped AI to integrate its real-time recommendation engine, which was specifically designed for the fast-paced, volatile inventory of live commerce. This technology provides personalized product suggestions in milliseconds, solving a critical lag issue that traditional e-commerce recommendation systems cannot handle at scale.
Shaped's infrastructure was purpose-built for live marketplace dynamics, where inventory and viewer intent change by the second. Standard recommendation engines designed for traditional e-commerce fail in this environment. The acquisition brings Shaped's 12-person team, led by founder and former Meta engineer Tullie Murrell, to form Whatnot's new Applied AI Research group.
How Fast Are the New Recommendations?
The technology promises a dramatic leap in performance. Whatnot previously spent six years reducing its recommendation latency from one day to several minutes. Shaped's engine compresses that delay to near real-time performance. This speed is critical in a live auction environment where items sell seconds after appearing. As Emmanuel Fuentes, VP of Data and AI at Whatnot, explained, "Inventory changes by the second, and buyer intent shifts throughout a show, making traditional batch recommender systems unusable at scale."
What Technical Capabilities Does Shaped AI Bring?
According to industry reports, Shaped contributes several specialized technical strengths that show significant improvements over incumbent solutions in key areas including hit rate performance, ranking relevance, latency, and model deployment speed. The company's system demonstrates meaningful advantages compared to existing platforms like AWS Personalize across these metrics.
Beyond these performance improvements, Shaped's system is tuned for sparse product data, inferring relevance from live chat, bidding patterns, and on-screen text. Its context engine updates user profiles with every tap, bid, or scroll, ensuring recommendations adapt instantly as a show progresses.
How This Acquisition Positions Whatnot Against Competitors
The acquisition strengthens Whatnot's position in a rapidly growing market. The U.S. live shopping market continues to expand significantly, with intense competition from major platforms:
- TikTok Shop: Leverages its massive user base and native video integration.
- Amazon Live: Benefits from a seamless checkout process within the Amazon ecosystem.
- Meta: Currently hosts a significant portion of U.S. live shopping events on Facebook.
Poor recommendations cause users to abandon streams. With Whatnot expanding into many new product categories, as noted by AlleyWatch, real-time ranking becomes essential for discovery at scale and defending its market share.
What Are the Long-Term Strategic Implications?
Beyond immediate user experience gains, Whatnot views this as a foundational investment in its AI infrastructure. The new research group will focus on teaching models to understand live video content, enabling search across products that appeared just seconds ago. While not an immediate priority, Whatnot may eventually offer its recommendation infrastructure as a service to other platforms.
The acquisition highlights a major trend in the e-commerce arms race: platforms are now competing on the speed of their personalization AI. This in-house technical differentiation creates a significant competitive moat, suggesting the future of live shopping may hinge less on studio polish and more on milliseconds of relevance.