Blurbs Logo
Blurbs Logo Create Profile | Sign In

Marqo

Claim page

"AI-driven search enhancing customer interaction accuracy."

Case Study
Frequently Asked Questions
This solution tailors search results by analyzing customer interactions, using AI to predict their needs more accurately. It reduces irrelevant results, automates tagging, and enhances engagement by personalizing based on user behavior and search intent, ultimately boosting conversions and satisfaction.
SHOPTALK SPRING '25
BlurbSTAR Case Study
Marqo & Redbubble
Marqo greatly enhanced Redbubble's search and metrics.
30%
Improvement in add-to-cart rate
3.85%
Increase in average search revenue
1.
Situation
Challenge with User-Generated Content
→
Difficulty with user-generated content accuracy
→
Challenges in representing unique artworks
→
Existing search system limited for long-tail queries
Redbubble, a global online marketplace known for print-on-demand products, faced challenges with their existing search capabilities. Their platform was mainly powered by user-generated content, making it difficult to accurately represent artwork. The user-generated metadata, such as tags and titles, was prone to inaccuracies and manipulation, which created difficulties in capturing the true essence of the content. With a sophisticated search stack already in place, improving the relevance of search results was challenging, especially for long-tail queries which had sparse interaction signals. Their OpenSearch system worked well for popular queries, but there was room for improvement with more descriptive, three-plus-word queries.
2.
Task
Introduce Vector Search Capabilities
→
Enhance search results relevance
→
Integrate vector search with current system
→
Improve user experience for complex queries
Redbubble aimed to enhance its search experience to better capture user intent and improve the relevance of search results, particularly for long-tail queries. The existing OpenSearch system needed to be complemented by a more sophisticated vector search engine capable of understanding complex, descriptive queries. The task was to find and implement a solution that could process the subjective and varied nature of artwork while optimizing the shopping experience by improving metrics like add-to-cart rates and order values for descriptive and multi-word queries.
3.
Action
Implement Marqo Vector Search
→
Evaluated and selected Marqo's advanced functionalities
→
Integrated CLIP models for image and text search
→
Aligned models with customer intent using Marqtune
After extensive evaluation, Redbubble chose Marqo's vector search engine due to its advanced functionality and ability to integrate seamlessly with their system. Marqo's end-to-end capabilities, including model fine-tuning with Marqtune, allowed for a transformation in capturing the relevancy of artwork. By leveraging state-of-the-art CLIP models, search results were enhanced to include text and image features, producing an understanding that was independent of user-generated tags. This approach enabled Redbubble to generalize across unseen works and align search relevance with customer intent and historical data, significantly improving the online search experience.
4.
Result
Significant Metric Improvements Achieved
→
30% increase in add-to-cart rates
→
19% improvement for long-tail query add-to-cart rates
→
2.3% increase in average order value
The implementation of Marqo and its continuous fine-tuning of models resulted in notable improvements across various performance metrics for Redbubble. Tailored models outperformed the existing search algorithms, significantly increasing the Add-to-Cart rates for searches with longer queries. Specifically, a 30% improvement in add-to-cart rates for descriptive, five-plus-word searches, and a general improvement of 19% for four-plus-word queries were recorded. Additionally, the average order value increased by 2.3% across all queries, signifying an enhancement in user engagement and conversion rates. This transformation in search efficiency and accuracy showcased a significant return on investment for Redbubble.
Keywords
MARQO
REDBUBBLE
VECTOR SEARCH
MULTIMODAL SEARCH
MARQTUNE
ADD-TO-CART IMPROVEMENT
SEMANTIC SEARCH
ECOMMERCE SEARCH
OPENSEARCH
CLIP MODELS
Marqo
Website
Back to Top
© 2025 Blurbs
Privacy Policy