Skip to content

LLM Based Ranking and Relevance

Marqo's LLM Based Ranking and Relevance uses large language models (LLMs) trained specifically on your customer interaction data to optimize product rankings. The Marqo pixel automatically collects user interaction data (clicks, purchases, browsing behavior) from your site, so no manual data submission is required. This modern approach learns patterns from your customer behavior and applies them intelligently across your entire catalog—even for new products with no sales history.

Overview

Traditional learning-to-rank systems are outdated: they can only rank products they've seen before with historical sales data. If you launch a new product or customers search for something new, these systems have nothing to work with.

LLM Based Ranking and Relevance is different. The Marqo pixel automatically collects your customer behavior data—what they click, what they buy, how they browse—and trains LLMs on this real interaction data. The system learns not just what sells, but why it sells. This understanding applies to new products and new queries immediately, without waiting weeks or months for data to accumulate.

Why LLM Based Ranking Beats Traditional Learning-to-Rank

Immediate Performance for New Products

Traditional learning-to-rank requires extensive historical data for each product before it can rank well. New product launches, seasonal items, and niche products struggle because there's no sales history to learn from.

LLM Based Ranking and Relevance understands patterns: what colors your customers prefer, what price points convert, what styles resonate. When a new product launches, the AI applies these learned patterns immediately—no waiting period, no poor initial rankings, no lost sales.

Smarter Pattern Recognition

Traditional systems work like a spreadsheet: they need exact matches. If customers haven't searched for a specific product variation before, the system can't rank it well—even if similar products convert highly.

LLM Based Ranking and Relevance recognizes patterns across your entire catalog. It understands that certain product combinations, styles, and attributes work well together. So when customers search for variations you haven't seen before, the AI still delivers relevant results because it understands the underlying patterns.

Built Specifically for Your Business

Every ecommerce business is unique. Your customers have specific preferences, your products have unique characteristics, and your brand has its own voice. LLM Based Ranking and Relevance learns your business specifically:

  • What product attributes matter most to your customers
  • How your customers browse and discover products
  • What price points and styles resonate with your audience
  • How your brand positioning affects what converts

This isn't a generic algorithm—it's an AI system trained specifically on your customer data, understanding your business like a seasoned merchandiser.

Business Benefits

LLM Based Ranking and Relevance delivers measurable results:

  • Higher conversion rates: Products that match customer preferences automatically rank higher
  • Better new product launches: New items perform well from day one, not after weeks of data collection
  • Improved revenue: Higher-margin products that customers actually want rank better
  • Less manual work: The AI handles ranking optimization automatically, reducing need for constant boost/bury adjustments
  • Scalability: Works seamlessly across millions of products and thousands of search queries

Monitoring Performance

Track how well your AI ranking is performing:

  1. Go to Analytics → Ranking Performance
  2. View key metrics:
  3. Conversion rate improvement: How much has conversion improved vs. baseline?
  4. Revenue per search: Is revenue per search increasing?
  5. New product performance: How well are new products ranking?
  6. Query coverage: Are more queries getting relevant results?

Best Practices

  1. Give it time to learn: The Marqo pixel automatically collects customer interaction data as users browse your site. The more interaction data collected, the better the model performs. Allow a few weeks for the system to learn your patterns.
  2. Monitor performance: Track conversion and revenue metrics to ensure the model is working well for your business.
  3. Combine with manual controls: Use boost/bury and pinning for strategic products or campaigns, let AI handle the rest automatically.
  4. Trust the patterns: The AI learns real patterns from your customer data—if it's ranking something highly, there's usually a good reason.
  5. Review periodically: Check analytics monthly to ensure the model continues performing well as your business evolves.