The Architecture
Warehouse as source of truth. CDP or Data 360 federates. CRM surfaces the activations. Analytics happen in the warehouse. Reverse ETL pushes updates back. One data layer, many consumers.
Vendor Choices
Snowflake for ease-of-use, ecosystem, SQL performance. Databricks for data + ML unification, Spark workloads. BigQuery for GCP-native, serverless economics. Each has merits; consolidation around one is the modern pattern.
Modeling
Medallion architecture (bronze/silver/gold). Customer dimensional models. Activation-ready golden records. Invest in modeling before activation — bad models make every downstream job painful.
CDP Integration
Warehouse feeds Data 360 (or other CDP). CDP resolves identity, activates segments, personalizes in real-time. Avoid duplicating warehouse logic in CDP; keep CDP focused on activation.