Global Retailer: AI Personalization on Snowflake

The challenge

A global retailer with 18 M SKUs was running personalization on a legacy Hadoop stack. Recommendations were batch, 24 hours stale, and the merchandising team had no way to ask natural-language questions of the catalog.

What we delivered

• 01
  Customer 360 on Snowflake with Cortex Search for hybrid retrieval.
• 02
  RAG merchandising copilot on Cortex — merchandisers ask ‘which products should I cross-sell with this SKU?’ and get grounded answers.
• 03
  Real-time propensity scoring with Snowpark Python UDFs.
• 04
  Personalized recommendation API serving the web and mobile app at <80 ms p95.
• 05
  Full PII / consent governance via Horizon Catalog masking policies.

Outcomes

• +14% Average basket size.
• +28% CTR on personalized recommendations.
• −81% Time from data load to live personalization.
• 0 PII findings in the post-launch privacy review.

“The merchandising copilot is the first time our team has been able to ask the catalog a question in English and get a real answer back.”

— VP Merchandising, Global Retailer