GA4's reports are fine for many daily questions. They become limiting when you want multi-year comparisons, want to join website data with CRM revenue, need to query without thresholds, or want to define your own attribution logic. The BigQuery export solves those problems by sending raw event data to a data warehouse you control. It sounds like a data engineering project, and some of it is, but the first step is simple and worth doing even if your team cannot use the data yet. This guide explains the export from a marketer's point of view, using [Google's BigQuery Export documentation](https://support.google.com/analytics/answer/9358801) for the limits. Those limits can change, so recheck them before you plan around them.
Why should a marketing team use the GA4 BigQuery export?
The export gives you things GA4 reports do not:
- Your own history. GA4's event data retention for explorations is limited, and the export keeps a copy for as long as you choose to store it.
- Event-level data. You can see each event and its parameters rather than pre-aggregated numbers.
- Freedom from report limits. You avoid the sampling, thresholding and "(other)" row problems that appear in some interface reports.
- Joins with other data. Connect web behaviour to CRM stages, revenue, ad cost or offline data, which is how you answer questions like "which campaigns generated closed revenue?"
- Custom logic. Build your own session definitions, channel rules or attribution models.
- Fallback protection. Your data sits outside GA4, in case of changes to the interface or your account.
The main reason to set it up now is simple: it does not backfill. Data collection starts from when you create the link. If you might want to compare 2027 with 2026, you need the export running in 2026, even if nobody queries it for a while.
Read the transcript
You can see what you spent. Not what it earned. If tracking stops at the form, you are setting budgets without the revenue data.
Ad platforms and your CRM disagree. Signal is lost at every step of the journey. So budget follows whichever report is loudest.
Connect the ad click to the closed deal.
Tracked from the ad click to closed revenue. Each stage keeps its source and click ID. Budget decisions come from one trusted report.
Marketing attribution, with Rahul D Sarker. Book a call at rahuldsarker.co.
Do you need it?
Not every company needs to use this data straight away. This helps you decide what to do now.
| Your situation | What to do |
|---|---|
| Small site, simple questions, nobody can write SQL | Enable the daily export if the cost is negligible, and set a budget alert. Keep using GA4 reports. |
| Long B2B sales cycle, need to link web sources to CRM revenue | Use it. This is one of the most valuable applications. |
| High-traffic site hitting GA4 reporting limits | Use it, and plan event filtering and cost controls first. |
| Team wants custom attribution or a data warehouse | Use it as a data source alongside CRM and ad data. |
| Your GA4 setup is broken or inconsistent | Fix tracking first, but link the export anyway so history starts. A bad export only copies bad data. |
If tracking quality is in doubt, start with a GA4 audit. Raw data does not fix a poor event plan.
How do you set up the GA4 BigQuery export?
You need a Google Cloud project, Editor access or above on the GA4 property, and appropriate rights on the Google Cloud project. The usual route:
- Create or choose a Google Cloud project. Use one owned by your company's account rather than an agency's or a freelancer's personal login. This matters more than it sounds, because ownership disputes are painful.
- Enable the BigQuery API, if it is not already enabled, and set up billing if you plan to use streaming export or go beyond free limits.
- Link from GA4. Go to Admin, then Product links, then BigQuery links, and choose Link. Select your project and the data location. Choose carefully, as you generally cannot change the location of an existing dataset later.
- Choose data streams and events. You can include or exclude certain streams and events. Excluding noisy events helps to control volume.
- Choose the export frequency. Daily, Streaming or both, where available. See the next section.
- Check for the dataset the next day. It should be named analytics_ followed by your property ID. Daily tables are named events_ followed by the date.
Add budget alerts in Google Cloud at this point. They do not hard-stop spending by default, but they do warn you. Document the owner, location and frequency.
What are the GA4 BigQuery export limits and costs?
Check the current limits in Google's documentation before you commit, but as of this writing the important points are these.
Daily export limit. Standard GA4 properties have a daily export limit of 1 million events. If you consistently exceed it, Google warns you and may pause the daily export. You can reduce volume with event or stream exclusions.
Streaming export. This sends events through the day into intraday tables. It has no stated event cap, but it costs money and Google describes it as best effort, which means gaps are possible, and some user and session traffic source data is missing. Use the daily tables as your source of truth for finished days.
Costs. The link itself is free. You pay Google Cloud for storage and queries, and for streaming where you use it. BigQuery has a free monthly allowance for storage and queries. For small properties, costs can stay low. For larger ones, they rise fast if analysts run careless queries.
Keep query costs down:
- Query only the date ranges you need by filtering on the table suffix.
- Select only the columns you need. Avoid selecting all columns.
- Build smaller summary tables for dashboards, and do not point Looker Studio directly at raw events.
- Set maximum bytes billed, and consider project-level quotas.
- Review expensive queries each month.
What the data looks like
The raw table often surprises people who are used to the GA4 interface. Each row is one event, not one session. Several things are worth knowing up front:
- Parameters are nested. Details like page_location and ga_session_id live inside an array called event_params. To get them out, analysts use UNNEST in SQL. This is the first skill anyone needs.
- There is no sessions table. You build sessions yourself by combining user_pseudo_id and ga_session_id.
- Revenue and items have their own fields. Ecommerce data has dedicated fields and an items array.
- Traffic source has several meanings. The traffic_source field reflects the user's first acquisition, not the current session. Newer fields, such as collected_traffic_source and session_traffic_source_last_click, are more useful for session analysis. Check that they exist in your export version before you rely on them.
- Time is in microseconds. event_timestamp needs converting, and the time zone needs handling.
Numbers will not exactly match GA4 reports. The interface can include modelled data for users who refused consent and can apply thresholds, while the export has neither. Do not spend weeks trying to reconcile every row. Choose one source of truth per use case, understand the gap and write it down. A clean event taxonomy makes all of this much easier.
Five useful first analyses
Do not begin with a data warehouse project. Start with questions the marketing team already has:
- Multi-year trend by channel. Build a monthly table of sessions and key events by channel, so that you can compare beyond GA4's retention.
- Landing page to lead quality. Join form submissions to CRM outcomes. Which pages bring leads that actually become opportunities? This is often the first eye-opening result.
- Paths before conversion. Look at the sequence of pages for users who converted and those who did not. It can reveal whether pricing, case study or comparison pages matter.
- Time to convert. Measure the days from first visit to lead by source. Very useful in long-cycle B2B, where short attribution windows hide channels' contribution.
- Data quality checks. Look for missing parameters, sudden event count changes, unknown sources and duplicated events. Set simple alerts on these.
Use AI assistants to help draft SQL if you like, but have someone check the logic and test it on a small date range. A wrong join can look entirely plausible and misdirect a budget.
How do you connect GA4 BigQuery data to your CRM?
For B2B companies, this is usually why the export is worth having. The goal is to see which marketing touches led to pipeline and revenue.
The basic method:
- Capture the GA4 client ID on every form. Use a hidden field with Google Tag Manager, then store it on the CRM contact. Store UTM values as well.
- Export CRM data to BigQuery. Start with contacts, deals, stages, amounts and dates. A scheduled export or connector will do.
- Join on the client ID. The user_pseudo_id in GA4 usually holds this value. Test on known examples, because its formatting and consent status can vary.
- Build a simple model. First touch, last touch and any reasoned variant. Then report pipeline and revenue by channel and campaign.
- Show it in a dashboard. Use a summary table with Looker Studio.
Check privacy here. Joining pseudonymous analytics data with named CRM contacts changes what the data can reveal, so review your consent language and access rules. Our first-party data article and DPDP guide cover important considerations. The article on the full stack from GTM to Looker shows how the parts fit together.
Who should own it
This often falls between marketing, IT and an agency. Decide early:
- Google Cloud ownership: a company account, and not an individual. Review who has admin rights.
- Budget: who gets the alerts and who has authority to act on them.
- Data model: a person who owns definitions like "lead", "session" and "channel".
- Access: limit raw data access to those who need it. Use views for others.
- Retention: how long you keep the data, and why.
- Change log: tracking changes can break queries. Tell the analyst before you rename an event.
A small team can use one analyst or part-time specialist, plus a few well-maintained summary tables. You probably do not need a complex data stack at first. If you need to build this quickly, our attribution modelling service and marketing attribution consultant pages cover the options. Whatever route you pick, turn on the export early, because that is the one part you cannot do retrospectively.
FAQ
Linking is free for standard GA4 properties, but BigQuery charges apply for storage and queries beyond the free allowance, and streaming has its own cost. Small properties may pay little or nothing. Set a budget alert and check it.
Not through the standard export, which starts when you link it. You can export some aggregated data through the GA4 Data API for reporting, but that is not the same as the raw event history.
Reports may include modelled data, thresholds and different processing or attribution, while the export is raw event data. Time zones and query logic can also cause differences. Understand the gap rather than forcing an exact match.
Not for the link or for basic queries. An analyst with SQL skills can get a lot done. If you are joining many sources or building production dashboards, some engineering support will help with reliability and cost.