Ingest behavioral events from web and mobile channels into a unified customer record

Behavioral event ingestion is the mechanism by which anonymous intent signals become part of a customer's history. The transition from anonymous to known — and the retroactive stitching of pre-registration behavior to the authenticated profile — is the most commercially important capability in a CDP's event pipeline.

The anonymous-to-known transition. Every visitor starts as an anonymous device identified by a session-scoped or persistent anonymous identifier (ECID in AEP; anonymous_id in Segment; custom UUID in CDW-native stacks). When a visitor authenticates (registers, logs in), the platform must merge the anonymous identifier's event history to the authenticated profile. How quickly this merge happens — and whether it is deterministic or probabilistic — directly affects attribution accuracy and audience freshness.

Cross-device continuity. A customer who browses on desktop and converts on mobile is a single customer. The CDP must link both device sessions to the same profile via a shared persistent identifier (email, loyalty ID, phone). The event history from both channels should be queryable as a single timeline.

Event schema compliance. Events must arrive in the schema the profile store expects. Incomplete or malformed events are silently dropped in most CDPs, creating invisible gaps in the behavioral history. Schema validation at the edge (before events reach the data store) is the architectural safeguard.

Parallel viability (high). Event ingestion from web and mobile has direct parallels in composable stacks: Snowpipe Streaming or Kafka Connect for the web channel, mobile SDKs feeding the same stream, and dbt incremental models computing the identity stitching logic in the CDW. Phase 3 will document the Snowflake/dbt parallel path; Phase 4 will cover Hightouch Identity Resolution.