The GA4 BigQuery Export Is Raw Material, Not a Data Model

The free GA4 BigQuery export gave analysts access to raw event data, and many teams now treat it as the end point of their behavioural data strategy. It is better understood as the starting point.

The export does solve real problems with the GA4 interface. Reports in the UI are subject to sampling, data retention limits and platform-specific processing rules, so they are an interpretation of the data rather than the data itself. The export provides event-level records without sampling or aggregation, which makes it the right foundation for serious analysis.

It is still a vendor's view of the business. Google defines the schema and the event model and can change them, and the data reflects only what the GA4 tags managed to capture.

On its own, the export also lacks the context needed to describe a full customer journey: CRM records, product database attributes and offline touchpoints. Projects run into trouble when teams expect the raw stream to answer causal questions or provide a single customer view. Survival analysis of customer lifecycles or accurate revenue attribution requires a coherent, cross-platform user profile to exist first.

That is where the architecture work begins. Treat the export as high-quality raw material and build a transformation layer on top of it, typically dbt models in BigQuery, that cleans the event stream, models it and joins it to every other first-party source.

Consent-based, privacy-first collection can feed the same warehouse. Combining those signals with the modelled GA4 data widens coverage and reduces dependence on increasingly unreliable third-party cookies.

The analyst's role therefore extends beyond building dashboards into data architecture: modelling, integrating sources and turning raw events into a consistent, organisation-wide telemetry dataset that can support causal analysis.