"Visual search has been a revelation. Our customers can now find the products they love instantly, and the 2.5x increase in conversion rate for those users speaks for itself."
The Challenge
This home goods brand has a highly visual product catalog, but their search functionality was purely text-based. This created a frustrating experience for users inspired by images they found on social media or in magazines, leading to high bounce rates and lost sales when they couldn’t find the right keywords to describe what they were looking for.
Our Solution
We integrated our AI-Powered Visual Search service directly into their eCommerce platform. By adding a simple camera icon to their search bar, we empowered users to “shop with their camera.” Customers can now upload a photo of a room, a piece of furniture, or a decor item, and our AI instantly returns a curated list of the most visually similar products available for purchase.
Key Results
- 2.5x Higher Conversion Rate: Users who engaged with visual search converted at a rate 2.5 times higher than the site average.
- 40% Increase in User Engagement: Visual search users spent more time on the site and viewed more products per session.
- Reduced Bounce Rate: The intuitive search experience significantly lowered the bounce rate for users arriving with visual inspiration.
- Enhanced Product Discovery: The feature surfaced relevant items from the long-tail of their catalog that were previously difficult to find via text search.
Technology Spotlight
The visual search experience was powered by a state-of-the-art computer vision pipeline:
- Computer Vision Model: We used a CLIP-based model to generate rich vector embeddings for every product image in their catalog.
- Vector Database: These embeddings were stored and indexed in a Weaviate vector database, optimized for ultra-fast similarity search.
- Frontend Integration: A lightweight JavaScript component was added to their site to handle image uploads and display the visual search results seamlessly.
Our Engagement Model
- Catalog Analysis & Indexing: We began by processing the client’s entire product image catalog, generating and storing embeddings for each item.
- Infrastructure Deployment: Our team deployed the vector database and the API endpoint required to handle visual search queries.
- Frontend Integration & Testing: We worked with the client’s development team to integrate the visual search component into their UI, followed by a thorough testing phase.
- Launch & Performance Monitoring: After a successful launch, we provided ongoing monitoring of search performance and user engagement metrics to ensure a flawless experience.