Retail

+25% search conversion with semantic catalog search

Retrieval-augmented search and recommendations across a large retail product catalog.

The problem

Keyword search returned poor matches on a large catalog, and shoppers abandoned search results without finding relevant products.

The solution

We indexed the catalog into a vector store with a re-ranking layer, serving grounded, relevance-ranked results and recommendations.

+25%

Search conversion

−35%

Query latency

100k+ SKU

Catalog coverage

Stack

pgvectorOpenAI embeddingsCohere rerankFastAPI