GA4 BigQuery export guide: setup, costs and first queries

The GA4 interface answers most day-to-day questions. But questions like "which campaign's leads turned into revenue?", "how does this period compare with last year, event by event?" or "why do GA4 and the CRM count differently?" need raw data. GA4's BigQuery export moves that raw data into BigQuery, Google's data warehouse. This guide covers the setup, the limits, costs as of October 2026, the table schema, first queries and joining CRM data.

Why export to BigQuery?

According to Google's help page, the export sends all the raw, unsampled event data GA4 receives to BigQuery, and the exported data is yours, with access managed through BigQuery permissions. The practical gains:

• Retention: on standard GA4 properties, event data retention can be set to at most 14 months, and that setting governs explorations and funnel reports. Data in BigQuery stays until you delete it.
• Joins: you can combine GA4 data with CRM, ad spend and revenue data in one place.
• Reporting: Looker Studio dashboards can read from BigQuery instead of running into GA4 API quotas. See what Looker Studio is for the details.

One caveat: Google says BigQuery data excludes the modelling and similar additions GA4 applies in its reports, so BigQuery and the GA4 interface will not always match exactly.

Export types and limits

Details worth knowing:

• The 1 million event limit. If a standard property keeps exceeding it, the daily export is paused and earlier days are not reprocessed. Properties that exceed it by a wide margin may be paused right away. The fix is to exclude data streams and unneeded events in the link settings.
• Streaming leaves out traffic source data for new users. Google recommends the daily export for user attribution data.
• Late events. Daily tables can be updated for a few days with events that arrive late. Treat the last three days as provisional.
• No backfill, no re-export. The export runs from the moment you link; Google says exported data cannot be re-exported.

Costs: sandbox, free tier and prices

The link itself is free on the GA4 side; the cost is in BigQuery. The figures below are those shown on Google Cloud's BigQuery pricing page as of October 2026, in US dollars, for the page's default region, Iowa (us-central1). They can differ by region.

The BigQuery sandbox works without a credit card or billing account. Google's sandbox docs list its limits: a lifetime limit of 10 GiB of storage (not refunded when you delete data), 1 TiB of queries a month, all tables and partitions expiring after 60 days, and no streaming. For the GA4 export that makes the sandbox fine for a trial but not for lasting history: daily tables older than 60 days disappear. Google's setup page also warns that moving from the sandbox to paid can cause export failures if the project is already over 10 GB.

The most effective way to keep costs down is to write queries carefully. BigQuery charges by the columns you read, so selecting only the columns you need instead of SELECT *, and narrowing the date range with _TABLE_SUFFIX, cuts the data scanned sharply.

Setup, step by step

Google's setup docs describe three stages:

1. Create a Google Cloud project and enable the BigQuery API.
2. Prepare the project: start with the sandbox or attach a billing account. Google notes that the export needs a valid payment method to continue, and days interrupted by a payment problem cannot be re-exported.
3. Link GA4: in GA4 go to Admin > Product links > BigQuery links > Link. Choose the project and the data location, pick the data streams to export and the events to exclude, tick Daily, Streaming or both, and submit.

Permissions: at least the Editor role on the GA4 property, and an email address with Owner access to the BigQuery project. Data should start arriving within 24 hours of linking.

Choose the location carefully. The data location is set during linking if the project has no dataset for the property yet. Changing it later means deleting the link, backing up the data and recreating the dataset, and can leave a gap in the data; Google suggests cross-region dataset replication as an alternative. As of October 2026 BigQuery's location list has no region in Türkiye; more on that in the data protection section below.

Schema: tables and fields

Each property gets a dataset named analytics_<property_id> in the project. Inside it:

• events_YYYYMMDD: the full daily table from the daily export.
• events_intraday_YYYYMMDD: the streaming export's intraday table, deleted once the daily table is complete.

Each row is one event. Common fields: event_date, event_timestamp (microseconds, UTC), event_name, user_pseudo_id (a pseudonymous user ID), user_id (if you send one), event_params (event parameters, a repeated RECORD), collected_traffic_source (UTM values collected with the event) and session_traffic_source_last_click (the session's last-click source). Because parameters are nested, queries use UNNEST.

First queries

In the examples below, replace project.analytics_123456789 with your own project and dataset.

Daily events and users:

SELECT
  event_date,
  COUNT(*) AS events,
  COUNT(DISTINCT user_pseudo_id) AS users
FROM `project.analytics_123456789.events_*`
WHERE _TABLE_SUFFIX BETWEEN '20261001' AND '20261007'
GROUP BY event_date
ORDER BY event_date;

Leads by source and medium (change generate_lead if your lead event has a different name):

SELECT
  session_traffic_source_last_click.manual_campaign.source AS source,
  session_traffic_source_last_click.manual_campaign.medium AS medium,
  COUNT(*) AS leads
FROM `project.analytics_123456789.events_*`
WHERE _TABLE_SUFFIX BETWEEN '20261001' AND '20261031'
  AND event_name = 'generate_lead'
GROUP BY source, medium
ORDER BY leads DESC;

Reading a parameter (the page the lead came from):

SELECT
  (SELECT value.string_value FROM UNNEST(event_params)
   WHERE key = 'page_location') AS page,
  COUNT(*) AS leads
FROM `project.analytics_123456789.events_*`
WHERE _TABLE_SUFFIX BETWEEN '20261001' AND '20261031'
  AND event_name = 'generate_lead'
GROUP BY page
ORDER BY leads DESC;

Before running a query, check the "this query will process X MB" estimate in the top right of the BigQuery editor; you see the cost before you pay it.

Joining CRM data

This is where the real value is. GA4 knows which campaign a lead came from; the CRM knows whether that lead was qualified and whether it closed. Joining them needs a shared key.

The method we recommend:

1. Generate a random lead ID (a UUID, for example) when the form is submitted.
2. Write it both to the CRM record and, as a parameter (lead_id), to the GA4 generate_lead event. For the tagging side, see what Google Tag Manager is.
3. Load the CRM's lead table (ID, stage, deal value) into BigQuery on a schedule.
4. Join the two tables on lead_id.

WITH ga4_leads AS (
  SELECT
    (SELECT value.string_value FROM UNNEST(event_params)
     WHERE key = 'lead_id') AS lead_id,
    session_traffic_source_last_click.manual_campaign.campaign_name AS campaign
  FROM `project.analytics_123456789.events_*`
  WHERE _TABLE_SUFFIX BETWEEN '20260701' AND '20260930'
    AND event_name = 'generate_lead'
)
SELECT
  g.campaign,
  COUNT(*) AS leads,
  COUNTIF(c.stage = 'qualified') AS qualified_leads,
  SUM(c.deal_value) AS won_value
FROM ga4_leads g
LEFT JOIN `project.crm.leads` c USING (lead_id)
GROUP BY g.campaign
ORDER BY won_value DESC;

Do not use an email address or phone number as the join key, and do not send them to GA4: Google Analytics policies do not allow personally identifiable information to be sent. A technical ID is both safer and enough. We build this whole set-up as our marketing data warehouse service.

A note on KVKK: data location and transfers abroad

BigQuery has no region in Türkiye, so for a Turkish company any location chosen for the GA4 export is abroad. If the exported data includes user_pseudo_id, user_id or IDs from the CRM, review with your counsel whether these count as personal data and on what basis they are transferred.

The rule on transfers abroad is Article 9 of Law No. 6698 on the Protection of Personal Data (KVKK), as rewritten by Article 34 of Law No. 7499 (Official Gazette, 12 March 2024, issue 32487). In outline:

• KVKK Art. 9(1): personal data may be transferred abroad if one of the conditions in Articles 5 and 6 applies and there is an adequacy decision for the destination country, sector or international organisation.
• KVKK Art. 9(4): without an adequacy decision, data may be transferred with one of the listed appropriate safeguards, provided a condition in Articles 5 and 6 applies and the data subject can exercise their rights in the destination country; one safeguard is the standard contract announced by the Board (Art. 9(4)(c)).
• KVKK Art. 9(5): a standard contract must be notified to the Authority by the controller or processor within five working days of signing.
• KVKK Art. 9(6): with neither an adequacy decision nor a safeguard, transfers are allowed only occasionally, in one of the listed cases (for example, explicit consent after being informed of the risks).

On the technical side, the most effective step is data minimisation: send no personal data to GA4, exclude unneeded events from the export, limit BigQuery access by role, and set a retention period. Google notes that when Consent Mode is in place, cookieless pings also appear in the export.

How Roicool works on this

As part of GA4 consulting we switch on the export, tune event volume to the 1 million limit and set up the first queries; the CRM join and the reporting layer are part of a marketing data warehouse project. For measurement as a whole, see our analytics consulting page. How GA4 differs from product analytics tools is covered in our GA4 vs Mixpanel comparison.

Sources

• Google Analytics Help, BigQuery Export: support.google.com/analytics/answer/9358801 (accessed October 2026)
• Google Analytics Help, Set up BigQuery Export: support.google.com/analytics/answer/9823238 (accessed October 2026)
• Google Analytics Help, BigQuery Export schema: support.google.com/analytics/answer/7029846 (accessed October 2026)
• Google Analytics Help, Data retention: support.google.com/analytics/answer/7667196 (accessed October 2026)
• Google Analytics Help, Best practices to avoid sending PII: support.google.com/analytics/answer/6366371 (accessed October 2026)
• Google Cloud, BigQuery pricing: cloud.google.com/bigquery/pricing (accessed October 2026)
• Google Cloud, Try BigQuery using the sandbox: cloud.google.com/bigquery/docs/sandbox (accessed October 2026)
• Google Cloud, BigQuery locations: cloud.google.com/bigquery/docs/locations (accessed October 2026)
• Law No. 7499, Art. 34 (KVKK Art. 9), Official Gazette 12 March 2024, issue 32487 (Turkish): resmigazete.gov.tr (accessed October 2026)