Why Visual Discovery Wins in Retail

This one was a treat to give. Recorded live at Suzhou Industrial Park, the talk lays out the case for why retail discovery is shifting from keyword search to visual, intuitive, profitable — and what that means for conversion, AOV, and returns.

The reference point

"You've probably heard of Google Lens, Amazon Lens, and in China, AliExpress and Temu's lens. To give you one example: Amazon spent millions and over 100 machine learning engineers and over a year to develop their version. We built ours with four engineers in a few months."

The point isn't the brag — it's that the technology has matured. What used to be a moonshot research project is now something a focused team can ship in production. And once it's in production, the rest of e-commerce stops being a list of grids and a search box.

What e-commerce actually has wrong

"Conversion rates are very low. The user checks out a few products and leaves — especially in furniture, where someone shops once every couple of months and then disappears. With AI, that's solvable now. The users want better tools."

The four products in the talk:

• Youzu Lens — visual search, but with object-level segmentation, so it doesn't just match the whole image, it understands every item inside it
• Room Visualizer — AI room generation with shoppable hotspots, not just a pretty picture
• Magic Background — replace plain white catalog photos with lifestyle imagery without booking a photographer
• 3D Viewer — turn one flat image into an interactive 3D model

The recommendation engine in disguise

The reframe in the middle of the talk is the most important one:

"Don't think about Lens just as product discovery. Think about it as a recommendation engine. If a product is sold out, we recommend visually similar items from the same catalog. It can even be a cross-category recommendation — and it enriches the data for text search too."

And on velocity

"We onboard customers very fast. Even for large customers with billions of SKUs. China's bias to action is something we resonate with — turn decisions into pilots you can actually measure."

The customer story

The talk closes with Vivre — the largest furniture marketplace in Romania, deployed across 10 Central and Eastern European countries. Higher conversion, higher average order value, more time spent on the platform. And our Room Visualizer is sitting on their live website right now.

"We focus on furniture today. But our technology is zero-shot — it works on fashion, food, and beyond."