Case Study: Houzz improves product search accuracy and conversions with Provectus generative AI

Redefining Home Improvement Experiences with Generative AI

Houzz, an online platform for home renovation and design, sought to improve its search engine to better connect homeowners with products and professionals. Their existing system, powered by a classic NLP model, struggled to interpret detailed user queries, leading to poor result accuracy. To enhance user experience and boost conversion rates, they partnered with vendor Provectus to leverage generative AI for semantic query understanding.

Provectus developed a solution using Amazon Titan Text Embeddings and synthetic data generated by Anthropic's Sonnet. The new system significantly improved the identification of product categories and attributes within queries. This resulted in a search accuracy of 78%, a 79% precision rate, and 85% recall, far surpassing the old model. For Houzz, this led to a more efficient search experience, increased user engagement, and a significant rise in conversions on product pages.

Houzz