ML Recommendation Engine That Lifted AOV by 34%
Real-time product recommendations built on TensorFlow Serving that grew average order value 34%.
Client: Velora Home
The Challenge
Velora Home's 40,000-SKU furniture catalog overwhelmed shoppers, and generic bestseller carousels converted poorly. Cross-sell attach rates sat below 8% despite heavy email promotion.
Our Solution
We trained a two-tower neural recommendation model on three years of browsing and purchase data, deployed via TensorFlow Serving behind a low-latency API. A Next.js storefront integration rendered personalized rails in under 80ms, with an A/B framework to iterate on ranking signals weekly.
The Results
Average order value rose 34% and cross-sell attach rate more than doubled to 19%. Personalized sessions converted 2.4x better than the control experience.
Ready to be our next success story?
Get a free consultation and a tailored roadmap for your project.
Talk to an Expert