GA4 to BigQuery: How to Move Your Data
Connect GA4 to BigQuery with Airbyte. Sync report-shaped analytics data, set lookback windows to avoid drift, and consolidate multiple properties.

Moving GA4 to BigQuery is worth starting with an honest question: do you need a pipeline at all? Google Analytics 4 has a native BigQuery export built into the product, it is free on the standard tier, and it delivers raw event-level data. If that covers your case, use it.
This guide covers the managed path with Airbyte, which solves a different problem: report-shaped data rather than raw events, history predating the export, and several properties through one configuration. Two settings decide whether the numbers you land can be trusted.
GA4 to BigQuery at a glance:
Why move data from GA4 to BigQuery?
Because the native export, good as it is, does not cover everything.
The good case is one of four things. You need history from before the export was switched on, which the export cannot give you retrospectively. You want report-shaped data matching what the GA4 interface shows, rather than raw events you must aggregate yourself. You run several properties and want them in one place. Or you want web analytics landing on the same schedule as every other source your team syncs.
The poor case is raw event data from today onwards into BigQuery and nothing more, which the native export does for free and without a pipeline to operate. The two are not mutually exclusive: plenty of teams run the export for events and a connector for reports, joining them in the warehouse.
What do you need before you start?
Four things, and the first one disqualifies some properties entirely:
A GA4 property, not a Universal Analytics one. This connector uses the Google Analytics Data API, which reports on GA4 only. If your property settings show a tracking ID in the UA format, it cannot be synced here. The GA4 source documentation sets out the full configuration.
Your numeric property ID. Found in property settings, and not the measurement ID beginning with G. The field accepts a list, so collect every property you intend to sync now rather than adding them piecemeal.
A view on how long your conversions take to settle. This becomes your lookback window, and the default of two days is almost certainly wrong for anything with a considered purchase cycle.
The questions your custom reports need to answer. Each report takes between 1 and 9 dimensions and between 1 and 10 metrics, so knowing the questions up front tells you how many reports to define.
If your environment restricts outbound traffic by IP, add the Airbyte Cloud IP addresses to the relevant allow lists before you begin.
How do you build a GA4 to BigQuery pipeline in Airbyte?
Step 1: Grant access to the property
On Airbyte Cloud you authenticate with OAuth as a Google account that can already see the property. On Open Source, create a service account with a JSON key, then add that service account to the GA4 property itself. That second half is the step people skip, and its symptom is an authenticated connection that returns nothing at all.
Step 2: Configure the GA4 source
Click Sources in the left navigation, then New Source, and select Google Analytics 4, following adding a source. Enter your property IDs, set the lookback window to match your attribution cycle rather than leaving it at two, and define any custom reports as a JSON array. Enable the property metadata stream while you are here, so dashboards can show property names instead of numeric IDs.
Step 3: Configure the BigQuery destination
Click Destinations, then New Destination, and select BigQuery, following adding a destination. Supply a service account with BigQuery User and BigQuery Data Editor. Batched Standard Inserts suits report volumes. If you also run the native export, put this dataset in the same location so the two can be joined, since BigQuery cannot query across physical locations and the location is fixed at creation.
Step 4: Create the connection and set sync modes
Click Connections, then New connection, select your reports and a sync mode. Use deduplication on the record key, because the lookback window means the same report date arrives repeatedly and you want the latest version to replace earlier ones rather than sit beside them.
Then model a table partitioned on the report date, because the default partitioning works against you here.
Why don't your numbers match the GA4 interface?
Two causes, and they have different fixes. The first is restatement. GA4 figures for a given day keep changing after that day ends, because attribution resolves late and the platform carries its own processing latency, so a report pulled this morning will not match the same report pulled tomorrow.
The lookback window handles that by refreshing a number of past days on every run. It accepts 2 to 60 and defaults to 2, which covers processing latency and not attribution. Raise it to match how long your conversions actually take, and deduplicate in BigQuery so restated days replace earlier versions.
The second cause is sampling, and it is the more dangerous one. The Data API enforces compute thresholds on queries, and when a request exceeds them the API analyses a subset of sessions rather than all of them. The result still arrives and still looks like an ordinary number. Keeping the data request interval small keeps individual queries under the threshold, which is the practical defence against sampled figures reaching your warehouse unnoticed.
How should you handle several GA4 properties?
By consolidating, if your connector version offers it. Property IDs accepts a list, and by default each report then produces one stream per property, named with the property ID appended. For an organisation with a dozen properties that is a great many near-identical tables, and every cross-property question becomes a union.
The connector added an opt-in One Stream per Report mode for exactly this. It combines all configured property IDs into a single stream per report, named with a Consolidated suffix, with schemas merged across properties and the property ID in the primary key. One table per report, one query across the estate.
Enable it before anyone builds on the per-property streams if you can. Turning it on later means running Refresh source schema, after which the old streams are reported as removed and the consolidated ones appear in their place, so every downstream model referencing the old names breaks at once. And if you add a property ID at any point, clear the affected streams to backfill it.
Frequently asked questions
Should I use this or GA4's native BigQuery export?
Use the native export if raw events from today onwards is what you need, since it is free and built in. Use the connector for report-shaped data, history predating the export, several properties in one configuration, or orchestration alongside your other sources. Many teams run both.
Can I sync a Universal Analytics property?
No. The connector uses the Google Analytics Data API, which reports on GA4 properties only. A tracking ID in the UA format indicates a property it cannot read.
What lookback window should I set?
Long enough to cover your attribution cycle, anywhere from 2 to 60 days. The default of 2 covers processing latency only, so a considered purchase cycle needs considerably more.
How do I know whether my data was sampled?
Sampled results look like ordinary numbers, which is why the defence is preventative. Keep the data request interval small so individual queries stay under the API's compute thresholds.
Why are my queries scanning more than expected?
Tables are partitioned on the extraction timestamp, and the lookback window writes one report date across many extraction dates, so filtering on report date prunes nothing. Model a partitioned table downstream.
Get your GA4 data into BigQuery
Decide first whether the native export already serves you, because for raw events alone it does and costs nothing to run. If you need reports, history or several properties, raise the lookback window to match your attribution cycle, keep request intervals small so sampling never reaches you, consolidate properties before anyone builds on per-property streams, and model a table partitioned on the report date.
Airbyte's connector catalog includes 600+ pre-built connectors, so web analytics can land beside the ad spend and revenue data that give it meaning. For other marketing sources in the same warehouse, see Google Ads to BigQuery and YouTube Analytics to BigQuery.
Integrate with 700+ apps using Airbyte
Move data from 700+ sources into warehouses, lakes, and beyond. Set up pipelines in minutes with pre-built connectors and the Connector Builder.
