dbt-ga4: the community standard for modelling the GA4 BigQuery export

The GA4 BigQuery export lands as one semi-structured events_* table per day, an event per row, parameters nested and repeated. Useful for a warehouse, useless for a straight SELECT. dbt-ga4 is a dbt package that unnests it into staging and mart-level models: sessions, users, pageviews, ecommerce, and multi-channel attribution, ready to query or plug into a BI tool.

Why it made the cut

It is the closest thing GA4 modelling has to a default answer. It sits on dbt's official Package Hub, has been maintained since GA4 export first shipped, and is the package most agencies and in-house data teams reach for before writing their own unnesting logic from scratch. When someone asks "how do I get sessions out of the GA4 export properly", this package is usually the reply.

How to run it

1. Add the package to packages.yml in your dbt project and run dbt deps.
2. Point the package's source config at your GA4 export dataset and property ID.
3. Run dbt build to materialise the staging and mart models, then query the sessions or ecommerce marts directly.
4. Read the package's own docs for the session and channel-grouping logic before you rely on it for reporting, so you know how it defines a session and where that might differ from the GA4 UI.

Worth knowing

This models the export faithfully, it does not reconcile GA4's own session counts, which use a different, partly undocumented methodology. Expect some drift between what this package reports and what the GA4 interface shows. It is also a serious dbt undertaking, not a five-minute query: budget time for the initial build and for understanding the session/channel logic before you trust it in front of a client.