How Booking.com selected Weaviate as its vector database standard

Summary

When Booking.com's internal embedding service requirements outgrew OpenSearch, its Machine Learning & Data Science team moved to Weaviate to handle the wide variety of AI use cases across the company, with faster performance and roughly 40% lower cost.

The Challenge

Booking.com's Machine Learning & Data Science team runs a centralized embedding service that powers vector search and data management for machine learning, agentic, and GenAI projects across the organization. The service originally launched on OpenSearch because the team already had familiarity with it.

As more teams adopted the service, datasets grew to hundreds of millions of embeddings, queries increased in complexity, and concurrency and low-latency expectations rose, especially for user-facing applications. Holding OpenSearch performance steady meant constant tuning and ever-increasing costs as cluster sizes and operational overhead grew.

“As more teams started using our vector store, we began seeing highly diverse requirements across use cases. Some teams needed advanced capabilities like hybrid search or multi-vector support, while others demanded larger vector capacities and higher RPS metrics. These requirements brought us to a point where we needed to reassess whether our current setup could support this next phase of growth.” - Başak Tuğçe Eskili, Machine Learning Engineer, Booking.com

Why Weaviate?

Booking.com built a performance benchmark that closely resembled its production workloads: 100 million embeddings, increasing concurrent threads, and nearest-neighbor, filtered KNN, and mixed read/write queries. Weaviate delivered the most consistent performance of the evaluated databases.

• 20x faster performance at scale: Weaviate outperformed OpenSearch by 20x in the benchmark, with a 40x reduction in usage cost.
• 40% lower usage cost: Weaviate ran on a substantially smaller compute and memory footprint than the heavily tuned OpenSearch database.
• Fit for purpose: Weaviate scored better in operational maturity, deployment flexibility, cost predictability, and integration with the Booking.com ML ecosystem.

What's Next?

With Weaviate, Booking.com now has an AI data platform suited to the wide variety of AI systems across the enterprise. Because database access was abstracted behind its internal embedding service, migration from OpenSearch to Weaviate meant little more than a configuration change. As the team brings on new AI applications, it relies on Weaviate's continuous innovation and support.