Whatnot Acquires Shaped, Cuts Live Shopping Recommendation Lag
Serge Bulaev
Whatnot has bought the machine-learning company Shaped to help make live shopping recommendations much faster, possibly cutting delays from minutes to milliseconds. Shaped's technology may help Whatnot suggest products in real time during live auctions, which could be important as inventory and bids change quickly. Analysts suggest that better, faster recommendations are a key way for online marketplaces to grow, and Whatnot appears to be betting on Shaped's tools for this reason. Some experts note there are still open questions about privacy and costs for these new systems.
Livestream marketplace Whatnot has acquired New York-based machine-learning startup Shaped in a strategic move to slash live shopping recommendation lag from minutes to milliseconds. TechCrunch reports the deal strengthens Whatnot's discovery capabilities as resale giants like eBay and Poshmark race to integrate AI, but does not explicitly state it 'intensifies the race for real-time personalization'[1].
Why Whatnot bought Shaped to speed up live shopping
Whatnot acquired Shaped to integrate its real-time machine learning engine, which reduces recommendation latency from minutes to under 50 milliseconds. This speed is critical for suggesting relevant products during fast-paced live auctions, where inventory and bids change in seconds, aiming to improve user engagement and sales.
Real-time discovery remains live commerce's most significant technical challenge. With inventory and bids fluctuating by the second, traditional recommendation models are too slow. Whatnot sellers stream over 175,000 hours of footage on the platform weekly, facing auctions that can conclude in less than two minutes. Shaped's sub-50-millisecond ranking engine will become the core of a new Applied AI Research group, led by former Shaped CEO Tullie Murrell.
Key details of the acquisition include:
| Attribute | Detail |
|---|---|
| Announcement date | 15 Jul 2026 |
| Purchase price | Undisclosed |
| Incoming staff | ~12 engineers and researchers |
| Core tech | Machine-learning service that ranks text, user and session data in <50 ms |
| Goal | Serve dynamic product suggestions during live auctions |
Market context for Whatnot Buys AI Firm Shaped to Speed Up Live Shopping
The acquisition comes as the live commerce market continues to grow rapidly. With personalization engines driving a significant portion of revenue for major marketplaces, Whatnot's decision to acquire Shaped's proven infrastructure over building an in-house solution is a calculated strategic bet.
The move aligns with key industry trends in real-time recommendations:
- Contextual Personalization: Reacting to micro-behaviors like pausing or rewinding a stream.
- Generative AI: Creating conversational product suggestions from viewer questions.
- Automated Overlays: Displaying related items seconds after a product appears on screen.
In a company blog post, Shaped's founders noted that live marketplaces need new metrics, as a single recommendation can impact an auction's outcome. Emmanuel Fuentes, Whatnot's VP of Data and AI, emphasized that extreme speed is essential in an environment of constant, volatile inventory.
Implications for Whatnot's growth strategy
Following its recent funding round, Whatnot has aggressively expanded into over 45 categories and seen significant growth in cross-category buying. Integrating Shaped's low-latency models is expected to boost conversions in these new verticals and help retain buyers in the face of fierce competition from Amazon and TikTok.
The acquisition supports Whatnot's strategic priorities:
- Embed Shaped's models across all platforms (mobile, web, and replays).
- Pursue international expansion in Europe and Asia.
- Introduce advanced advertising slots based on AI-driven placement.
- Consider licensing its recommendation technology after meeting internal goals.
While challenges like real-time content moderation, data privacy, and infrastructure costs persist across the industry, Whatnot's acquisition of Shaped signals that the future of livestream retail may be won in milliseconds.