Target revamps search with Google Cloud tech

Target has rebuilt its retail product discovery platform using Google Cloud Spanner Graph, replacing a patchwork of search and transactional systems with a single database.

The update consolidates technology behind features like product ratings, reviews, and the retailer’s Gift Finder chat agent. This tool helps shoppers find items through conversational prompts on Target’s website and app.

From fragmented systems to a unified database

Previously, Target depended on Elasticsearch clusters for search and separate NoSQL databases for transactional data. The setup caused synchronization problems and forced engineers to manage data consistency across multiple systems.

Expanding the platform proved challenging. Adding new retail data domains required extra database collections, complex joins, and weaker transactional links. The old design couldn’t handle graph relationships, vector similarity, and keyword search indexes in a single transaction.

After evaluating standalone vector and graph databases, Target selected Spanner Graph to create an “enterprise ontology.” The system combines semantic data, graph relationships, vector embeddings, and transactional records in one environment.

The change addressed more than efficiency. The fragmented approach had created bottlenecks for AI-driven tools, which often failed to retrieve accurate product context. Traditional retrieval-augmented generation methods missed structured relationships between items, such as compatibility or age suitability, because they relied too much on vector similarity alone.

How the migration worked

Target transitioned the platform in four stages to avoid disrupting shoppers. First, it mapped retail entities—products, categories, brands, and guest preferences—into a Spanner Graph schema.

Next, engineers built a parallel data pipeline to replay live transactional updates, apply schema changes, generate embeddings, and write results into Spanner Graph in real time. They gradually shifted read traffic through a canary deployment before completing the cutover and retiring the legacy Elasticsearch stack.

The final step removed the maintenance burden of those clusters. It also consolidated workloads that had been spread across separate technologies. Target reported a 50% reduction in infrastructure upkeep.

Developers now spend less time on database administration. The new setup lets teams query relational catalogue data and graph relationships together using SQL and Graph Query Language. This eliminated some duplication and extract-transform-load work required under the old system.

Spanner’s autoscaling handled traffic spikes during major retail events like Black Friday and Cyber Monday without issues. The retailer noted improved recommendation relevance and customer satisfaction but didn’t share specific figures.

The project reflects a trend among large retailers. Many are moving away from stitching together separate platforms and toward unified systems that combine transactional databases, search, and AI retrieval. For Target, the result is a single source of truth for both operational retail data and semantic product intelligence.

Shoppers don’t just search for products—they ask questions. A unified system allows AI tools to understand context, such as whether a gift suits a toddler or if two products work together. The previous approach couldn’t reliably answer those kinds of queries.

Target’s Gift Finder chat agent demonstrates the new system’s capabilities. Instead of forcing shoppers to browse categories, it uses conversational prompts to narrow down options based on intent. The underlying data model supports these relationships, making the experience smoother for customers and easier to maintain for engineers.

The architecture consists of three layers. One captures catalogue data and metadata from internal sources, enriching product information with generative AI. A second stores entity nodes, relationship edges, and vector embeddings in Spanner Graph. The third supplies structured context to the large language models powering conversational shopping tools.

The company didn’t disclose the migration cost or exact timeline. The effort was driven by the need for better scalability and AI integration. The move also aligns with a broader push in retail to reduce complexity in backend systems, where each additional database or synchronization layer adds risk.

For now, the focus remains on maintaining the new system and expanding its capabilities. The retailer hasn’t ruled out further integrations but prioritizes stability during peak shopping periods.

The shift simplifies operations while improving how AI interprets product relationships.