Aisle Wanderer

The Challenge

In e-commerce, showing a user ten similar items is standard practice. In groceries, it's a disaster. With 20+ items per cart, this approach creates massive cognitive load, causing users to abandon purchases and shrinking the transaction value.

The challenge was to get the first product suggestion right, every time.

Our Solution

We solved this by creating a dynamic, LLM-based taxonomy from the ground up. This system learned the deep attributes of several million grocery SKUs, allowing it to make highly accurate, single best-match recommendations that traditional ML models couldn't.

Results

• Single-item recommendation accuracy that outperformed traditional similarity engines
• Reduced cognitive load for users during the shopping experience
• Increased transaction values through more relevant product suggestions
• Scalable system handling millions of grocery SKUs