Modernizing Search and Relevance for Large‑Scale B2B Commerce

Overview

I worked with Sysco , a global leader in foodservice distribution, to modernize and strengthen the relevance, performance, and scalability of their enterprise search platform. The engagement focused on evolving search from a primarily keyword‑driven utility into a strategic commerce capability—one that better supports customer intent, product discovery, and measurable business outcomes.

Rather than treating search as an isolated technical component, the work positioned it as a core driver of digital commerce performance, influencing conversion, basket composition, and customer lifetime value. The engagement combined strategic advisory, hands‑on relevance engineering, and enablement of Sysco’s internal teams to operate and extend the platform over time.

The Challenge

Sysco operates a massive and dynamic product catalog serving diverse customer segments, ordering patterns, and fulfillment contexts. While Elasticsearch was already in place, the platform needed to evolve to support more advanced use cases and growing expectations around relevance, personalization, and performance.

Key challenges included:

• improving relevance across high‑volume, long‑tail product searches,
• reducing “no results” and low‑quality result scenarios,
• supporting intent‑driven discovery in a commerce context,
• introducing modern vector‑based and machine‑learning techniques without disrupting existing workflows,
• and establishing clear KPIs and governance for ongoing relevance optimizat

Representative Search Intents

Exact and High‑Confidence Product Lookup

Typical query form: Known‑item or SKU‑driven searches with little tolerance for error. Expected behavior: Fast, precise retrieval with strong typo tolerance and ranking confidence.

Exploratory and Category‑Driven Discovery

Typical query form: Broad or semi‑structured searches such as product categories or use‑case driven terms. Expected behavior: Relevant category‑level results with strong ranking signals and guided refinement.

Intent‑Driven and Behavioral Queries

Typical query form: Searches that imply purchase intent, substitution, or replenishment behavior. Expected behavior: Results informed by behavioral signals, relevance models, and emerging personalization strategies.

Solution

The engagement delivered a phased search modernization and relevance enablement program, aligned with Sysco’s commerce roadmap and operational constraints.

1 - Relevance Strategy and Maturity Assessment

Established a shared relevance roadmap by mapping business goals, user journeys, and search behaviors. This created a common framework for prioritizing relevance improvements and measuring progress over time

2 - Platform Modernization and Architecture Guidance

Provided architectural recommendations to support expanded use cases, including improved ingestion pipelines, indexing strategies, and document modeling. The work aligned Sysco’s Elasticsearch implementation with newer platform capabilities and long‑term scalability goals

3 - Relevance Engineering and Signal Optimization

Conducted hands‑on relevance tuning across analyzers, queries, and ranking logic. This included recommendations for synonym handling, query restructuring, and signal collection to improve result quality across core commerce scenarios

4 - Introduction of Vector and ML‑Based Techniques

Advised on the adoption of native vector search and machine‑learning‑driven relevance techniques to support similarity, intent detection, and future personalization use cases—without forcing a disruptive architectural reset

5 - Enablement, Workshops, and Knowledge Transfer

Led relevance enablement sessions and working workshops designed to transfer both strategic context and practical tuning techniques to Sysco’s internal teams. This ensured the organization could sustain and evolve relevance improvements independently.

Technologies Used

• Elasticsearch (including vector search capabilities)
• Custom analyzers and query strategies
• Relevance testing frameworks and KPIs
• Analytics and reporting for search performance

Results

• Improved relevance and consistency across high‑volume commerce searches
• Reduced “no results” and low‑quality result scenarios
• Established a clear relevance roadmap tied to business outcomes
• Enabled internal teams to operationalize and continuously improve search
• Positioned search as a strategic lever for conversion and basket optimization