Identity Resolution – The Brain of Customer Data Platform (CDP)
summary
Explains a production identity-resolution pipeline that combines rule-based real-time merging with an out‑of‑band ML re‑consolidation, ingesting billions of events into about million profiles. Uses attribute embeddings, cross-polytope LSH for blocking to avoid quadratic compares, and Siamese neural networks for pair scoring; outputs merge recommendations back to the CDP. Operates near‑real‑time but still requires daily consolidation and leaves shared‑device merging as an unresolved failure mode.
AI-written summary · about
chapters (9)
- 0:00Introduction to CDP
- 3:41Fundamentals of CDP
- 8:16Role of identity resolution
- 12:57How identity resolution works
- 16:11Problems with rule-based matching
- 19:55Profile blocking and embeddings
- 23:56Out-of-band consolidation
- 27:31Locality sensitive hashing
- 28:35Automated corrections and challenges
related talks
29mUnified Customer Data Management
31mInsights & Engagements with Customer Data Platforms
23mRemoving Data Blindspots For A Connected Experiences Beyond Marketing
06mUnifying Online and Offline Worlds
25mRevolutionizing Customer Experience Through AI & Data
11mMaximizing personalization and AI models with quality customer data
16mBuilding an End-to-End De-Identification Pipeline for Advertising Activity Data at...
16mExploring Customer Data Platforms and Testing