Web and product analytics
GA4 BigQuery export: set-up, limits and costs
The short answer
Every GA4 property, including free ones, can export raw event data to BigQuery from Admin, Product links, BigQuery links. A standard property's daily export is limited to 1 million events a day (Analytics 360 allows up to 20 billion); streaming export has no volume limit but costs $0.05 per GB in BigQuery. Linking is free, but BigQuery storage and queries are billed by Google Cloud beyond its free tier or sandbox.
On this page 6 sections
Why export at all
- History you control. GA4 keeps event-level data for explorations for at most 14 months on standard properties. BigQuery keeps it as long as you pay for storage.
- No sampling or thresholds. You query every event.
- Joins. Combine web events with CRM, call tracking or order data on your own keys.
- Custom attribution and cohort analysis that the GA4 interface does not offer.
If nobody on the team writes SQL, the export is still worth switching on early because it is not retroactive: it only collects from the day you link it.
Set it up
- Create a Google Cloud project and enable BigQuery (the free BigQuery sandbox works to start; exports beyond sandbox limits incur charges).
- In GA4: Admin, Product links, BigQuery links, Link. Choose the project and a data location (pick the region nearest your other data; it cannot be changed later).
- Choose the streams and events to include. Excluding noisy events helps standard properties stay under the daily limit.
- Choose Daily, Streaming, or both. Submit.
- Data starts arriving within about a day as a dataset named analytics_<property ID>.
Daily or streaming
| Daily export | Streaming export | |
|---|---|---|
| Table | events_YYYYMMDD | events_intraday_YYYYMMDD (deleted once the daily table is complete) |
| Limit, standard property | 1 million events per day | No volume limit |
| Completeness | Complete for the day | Best effort, can have gaps |
| Extra cost | BigQuery storage and queries | Plus $0.05 per GB streamed (about 600,000 events per GB) |
| New-user attribution | Included | Campaign, source and medium for new users not included |
Query the daily table for anything you will report on. If a standard property goes over 1 million events a day, Google can pause the daily export, so filter out events you do not need.
What it costs in practice
Storage is cheap; queries are where bills appear, because BigQuery charges by data scanned. Always filter on the date suffix (_TABLE_SUFFIX BETWEEN '20261001' AND '20261031'), select only the columns you need instead of SELECT *, and set a project budget alert. A small business site will often stay inside BigQuery's monthly free query allowance.
For dashboards on top of the export, Looker Studio has a BigQuery connector that avoids the GA4 API quota: see building a GA4 dashboard in Looker Studio. For retention analysis, see cohort analysis explained.
Questions
- Is GA4 BigQuery export free?
- Linking GA4 to BigQuery is free for standard and 360 properties. You pay Google Cloud for BigQuery storage and queries beyond the free tier, and streaming export adds $0.05 per GB. The BigQuery sandbox lets you start without billing, within its limits.
- What is the GA4 BigQuery export limit?
- Standard GA4 properties have a daily BigQuery export limit of 1 million events per day. Analytics 360 properties allow up to 20 billion events per day. Streaming export has no volume limit.
- Can I export past GA4 data to BigQuery?
- No. The export starts from the day you create the link and does not backfill earlier data, so link it as early as possible.
Sources
Next step
Explore web analytics
Related guides
How to build a GA4 dashboard in Looker Studio
Connect GA4 to Looker Studio, build a one-page marketing dashboard, avoid GA4 API quota errors, and know when a reporting tool beats a home-made dashboard.
Cohort analysis explained, with GA4 steps
What a cohort retention table shows, how to read the triangle, standard vs rolling vs cumulative retention, and how to build a cohort exploration in GA4.
Marketing attribution models explained
First-click, last-click, linear, time-decay, position-based and data-driven attribution, explained with one worked example and how to pick one.
Build or buy a predictive model?
When a no-code predictive analytics platform makes sense, when to build your own model, and what training your own model costs in people and compute.