Getting Croct experiment data into BigQuery, via GA4 export or directly?
We already pipe GA4 into BigQuery. Now we want Croct variant assignments in the warehouse to join against revenue tables (some of which come from Snowflake via transfer).
Two paths as I see it: rely on the GA4 BigQuery export since Croct already streams into GA4, or pull from Croct directly if there is an API for it.
Which is the intended route? mainly care about completeness, GA4 processing has burned us before
1 answer
Both paths work, and they serve different needs.
The GA4 route is the zero-effort one: the Croct dimensions exported to GA4 (audiences, experiences, experiments, variants) arrive as event data, so they flow into your existing GA4 BigQuery export automatically. If your joins can live with GA4's processing, you are done already.
For raw completeness, pull from Croct directly. The data-export client offers three methods: exportEvents, exportSessions, and exportUserById. The export service is part of the HTTP API, so you can also hit it without the client: base URL is https://api.croct.io and authentication is an API key in the X-Api-Key header.
Given that GA4 processing has burned you before, I would use the direct export as the source of truth for the variant-to-revenue joins and keep the GA4 copy for the analysts who live in that dataset anyway.
Direct export it is. Is the export paginated or streamed? Planning a daily incremental load.
The client handles iteration for you, so a daily incremental job that pulls events and sessions since the last run is the straightforward pattern. Check the data-export client reference in the docs for the exact parameters.