Target Enhances Product Discovery with Spanner Graph and Cuts DB Maintenance by 50%

Target integrated search, vector, graph, and transaction data into Spanner Graph to enhance personalized product discovery and conversational shopping experiences. This enables richer semantic search and real-time responses based on connections between products, categories, and customer intent.

Previously, Target operated an Elasticsearch cluster alongside a separate NoSQL database. This architecture made it difficult to synchronize search, vector, and transaction data, imposed a high operational burden for managing search indexes and custom aggregation logic, and created bottlenecks when scaling data domains.

Instead of adding multiple specialized databases, Target built an enterprise ontology based on Spanner Graph. By using Spanner as a single source of truth for managing both transactional state and semantic intelligence, the following capabilities are handled within a single database:

• Integration of enterprise product catalogs and metadata
• Product data enrichment using generative AI
• Storage of entity nodes and relationship edges
• Vector embeddings and semantic similarity search
• Graph traversal and full-text search on relational tables

This architecture supports AI-driven features, such as Target's Gift Finder conversational shopping agent, which require understanding both product connectivity and customer intent. Consolidating the database also reduced the burden of database maintenance by 50%.