How Bazaarvoice scaled AI-powered product insights with Qdrant - Qdrant

What this post added

Bazaarvoice details their successful migration of billions of vectors from PostgreSQL with pgvector to Qdrant. Key technical challenges and solutions included: scaling to billions of embeddings with cost-efficiency, implementing tenant isolation via payload-based partitioning for scoped searches, and leveraging Qdrant's quantization for significant storage and RAM reduction (~100x). The migration was performed under real-world constraints with streaming data ingestion and disk-backed collections, still achieving sub-100ms query latency and ~98% recall. This migration enabled new product development by removing operational overhead of manual partitioning and enabling flexible, cross-product queries at query time.