I've been tasked with comparing the identity stitching capabilities of Segment and RudderStack for a new customer data pipeline. The goal is a single, unified user profile.
We ran both platforms in parallel for a month, ingesting identical web and mobile events via their respective SDKs. Primary key was `anonymous_id` with `user_id` on identify calls.
The core difference was in the merge model configuration. RudderStack's logic is defined in code, which was more transparent for our data team. Segment's merging felt more like a black box, though their docs claim deterministic rules.
We used the reverse ETL features to push computed identities to our warehouse (Snowflake) from both. The RudderStack output was easier to map to our internal models because we controlled the transformation SQL. Segment's merged profiles required more post-processing.
Main observations:
* RudderStack gave us more control over the identity graph logic (e.g., custom merge windows, attribute precedence).
* Segment's merged user tables were simpler to query for basic marketing use cases out of the box.
* Cost models diverged significantly at scale, with RudderStack's warehouse-native approach being cheaper for our high-volume event streams.
Key question for others: did you find Segment's proprietary merging algorithm to be a blocker for data governance requirements, or was it sufficient?