User Behavior Insights
summary
An open source pipeline that actually captures granular search events is available: a client-side JS library sends impressions, mouseovers, clicks and add-to-cart events into OpenSearch (optionally via Data Prepper or S3), using a shared schema and unique query and result IDs so causality is traceable. Data stays under customer control (PII policy left to implementers), is near real-time, and already powers demos that extract user preferences for personalization and A/B tuning. The project has an RFC and a PR to upstream OpenSearch.
AI-written summary · about
chapters (8)
- 0:00User behavior insights project
- 2:35Behavioral data for search relevance
- 12:09Virtuous circle of search relevance
- 17:30Tuning search with user behavior
- 26:43Demo of insights in action
- 31:18Integrating and extending the schema
- 32:24Standard for search analytics
- 33:32Q&A
related talks
29mMastering Search Performance: Advanced Query Insights for Production OpenSearch
23mBeyond Slow Logs: Implementing Advanced Query Insights in OpenSearch
39mOpenSearch Helps You to Adopt a Friend
42mLeveraging Digital Twins and OpenSearch for Personalized Customer Experiences
15mFrom Top N Lists to Visual Analytics: Query Insights Dashboard in OpenSearch
41mScaling Search and Powering Vector-Based Retrieval With OpenSea...
39mDemystifying Anomaly Detection in OpenSearch
18mFrom Logs To Signals: Rethinking System Monitoring With OpenSearch