Full Answer

GA4 maintains two distinct data layers: the raw event stream collected from users whose tracking was fully operational, and the modelled layer that estimates activity for users who rejected consent, had cookies blocked, or browsed across devices without being stitched. The BigQuery export draws from the raw event stream only.

This has important analytical implications. When a WooCommerce store's GA4 interface shows 10,000 sessions and the BigQuery export contains 7,200 session records, the 2,800-session difference represents GA4's modelled estimate of activity that was not directly measured. The BigQuery data is not missing records due to an export bug — it accurately reflects the scope of directly observed measurement.

For stores using BigQuery as their primary analytics platform, this creates a baseline accuracy advantage. Every row in the BigQuery export corresponds to a real event from a real session with real parameters. Revenue figures in the BigQuery data match actual tracked transactions. There is no modelled component inflating the numbers.

The tradeoff is coverage. BigQuery's export shows a store's measured reality, which — in markets with high consent rejection or ad blocker usage — may represent only 50-70% of actual site activity. Stores that supplement the GA4 BigQuery export with server-side event pipeline data to the same warehouse close this gap with directly measured events rather than modelled estimates. The pipeline captures events at the WooCommerce hook level regardless of consent status or browser restrictions, producing a BigQuery dataset that is both complete and accurate — the combination that GA4's architecture cannot deliver through either its interface or its export.