On this page
BigQuery Export (GA4)
BigQuery Export (GA4)
GA4's built-in integration that streams every raw, unsampled event row collected by a Google Analytics 4 property into a Google BigQuery dataset, queryable with SQL. Available to all standard GA4 properties at no platform charge (the feature required GA360 until 2023) and to Analytics 360 properties with higher quotas and additional export modes. The export is the foundation of custom attribution modeling, unbounded historical analysis, and CRM joining in ecommerce analytics stacks.
How it works
GA4 creates a BigQuery dataset named analytics_<property_id> in the linked Google Cloud project. Within it, three table types can be populated depending on which export modes are enabled (as-of 2026-09-26):
| Table | Export mode | Notes |
|---|---|---|
events_YYYYMMDD | Daily / Fresh Daily | One row per event; batch finalized once per day |
events_intraday_YYYYMMDD | Streaming | Continuously populated; deleted at day-end once daily table is complete |
Google Analytics Help Center confirms the export is forward-only — there is no historical backfill. Events collected before the BigQuery link is activated are not exported.
(Google Analytics Help Center — BigQuery Export overview)
Export modes
Three modes exist and can be enabled independently (as-of 2026-09-26):
- Daily — available to Standard and 360 properties. Exports all events for the previous day once per day; timing is not guaranteed but typically runs mid-afternoon in the property's timezone; can be delayed until later or the following day. Tables are back-filled for up to 2 calendar days plus today to capture late-arriving events from Measurement Protocol (GA4) or Firebase SDKs.
- Fresh Daily (360 only, Normal and Large properties; not XL) — delivers data typically by 5am in the property timezone with batched updates throughout the day, typically within 60 minutes. A completeness signal is emitted to Cloud Logging when all previous-day data has been exported.
- Streaming — available to Standard and 360. Populates
events_intraday_YYYYMMDDwithin minutes; best-effort service with no completeness SLO and may contain data gaps. The intraday table is deleted at day-end once the daily table is finalized. Streaming explicitly excludes new-user traffic source attribution fields (traffic_source.name,traffic_source.source,traffic_source.medium); user attribution for existing users requires ~24 hours to fully process.
(Google Analytics Help Center — BigQuery Export overview)
Limits and quotas
| Property type | Daily export limit (Daily mode) | Streaming limit |
|---|---|---|
| Standard | 1 million events/day (as-of 2026-09-26) | None |
| Analytics 360 | Up to 20 billion events/day (as-of 2026-09-26) | None |
If a standard property consistently exceeds the 1M-event daily limit, the daily export is paused and previous days' exports are not reprocessed. If it significantly exceeds the limit, Google may pause daily exports immediately. Property editors and administrators receive an email notification each time a property exceeds the daily export limit, indicating when the export will be paused if action is not taken.
Operators can exclude specific data streams and individual events from the export during or after the linking process to manage volume.
(Google Analytics Help Center — Set up BigQuery Export)
A practical ecommerce estimate: a busy WooCommerce store generating 5–10 events per visitor (page_view, scroll, add_to_cart, purchase and enhanced-ecommerce events) can reach 1 million daily events with approximately 100,000–200,000 daily sessions during peak traffic periods such as Black Friday. (as-of 2026-01-07 — practitioner estimate)
(Seresa — GA4 BigQuery event limit 2026)
Schema and data structure
The export schema uses deeply nested, repeated records. Key structural points confirmed by Google official docs and practitioner sources (as-of 2026-05-29):
event_paramsis a REPEATED RECORD (array of key/value structs); extracting any custom parameter requires anUNNESTsubquery.event_timestampis stored in microseconds (not milliseconds); conversion withTIMESTAMP_MICROS(event_timestamp)is required before applyingDATE().- No native session table — a unique session key is built by concatenating
user_pseudo_idwithga_session_id(extracted fromevent_paramsvia UNNEST). - Custom parameters are capped at 25 per event, 40 characters per parameter name, 100 characters per parameter value; truncated values are not recoverable after export.
events_YYYYMMDDdaily tables are updated by Analytics for up to 2 calendar days plus today beyond the table date.- wBRAID and gBRAID identifiers are not included in the export. (as-of 2026-09-26)
- When Google Consent Mode v2 is implemented, cookieless pings collected by Analytics appear in the BigQuery export. (as-of 2026-09-26)
- Modeled conversions, modelling applied by GA4, and value additions made in standard reports are not included — the export contains only raw event data as received from the client. This is a primary source of UI-vs-BigQuery discrepancy.
(Digital Applied — GA4 BigQuery Export 2026 Reference); (Google Analytics Help Center — BigQuery Export overview); (Google Analytics Help Center — BigQuery Export schema)
Traffic source structures
Four distinct traffic-source structures exist in the schema with different scopes and attribution (as-of 2026-05-29):
| Structure | Scope | Export mode | Notes |
|---|---|---|---|
traffic_source | User / first-acquisition | Daily only | Original structure |
collected_traffic_source | Event-scoped (raw) | Daily + Streaming | Added May 2023 |
session_traffic_source_last_click (STSLC) | Session-scoped, last-non-direct-click | Daily only | Added July 17, 2024; includes cross_channel_campaign sub-record (added Oct 9, 2024) with default and primary Custom Channel Grouping (GA4) |
event_params source/medium keys | Legacy event-scoped | Both | Original method |
(Digital Applied — GA4 BigQuery Export 2026 Reference)
Costs
Enabling the BigQuery Export is free. Costs are incurred on the Google Cloud side:
- Streaming export: $0.05 per gigabyte of data written; approximately 600,000 Google Analytics events equate to 1 GB (varies by event size). (as-of 2026-09-26, official)
- BigQuery compute (on-demand): $6.25 per TiB scanned beyond the free tier of 1 TiB per billing account per month. (as-of 2026-05-29, practitioner-cited — verify against current Google Cloud pricing page)
- BigQuery storage: charged per GB per month (active and long-term rates apply; long-term rate applies after 90 days unmodified). (as-of 2026-09-26)
- New from February 1, 2026: BigQuery users are charged for Cloud Storage multi-region data transfer fees when a BigQuery job reads data stored in a multi-region Cloud Storage bucket from a different location.
- BigQuery Sandbox: GA4 data can be exported to the sandbox at no charge, but the 10 GB sandbox storage limit applies; data can expire after ~60 days without an active billing account.
(Google Analytics Help Center — BigQuery Export overview); (Google Cloud — BigQuery pricing); (Digital Applied — GA4 BigQuery Export 2026 Reference)
Cost optimization
- Always include a
_TABLE_SUFFIXdate filter on wildcardevents_*queries; omitting it causes BigQuery to scan every matching table in the dataset — described by practitioners as "the most common way teams accidentally blow through the free query tier in an afternoon." - Consolidating date-sharded tables into a single native partitioned table clustered by
event_nameanduser_pseudo_idreduces query cost; vendor benchmarks cite 80–95% reduction in bytes scanned for typical analytics workloads (as-of 2026-05-29, vendor upper-bound estimate; actual savings depend on query patterns). - Practitioners recommend waiting 48–72 hours before treating a daily table as final, as late-arriving events back-fill for up to 2 days.
(Digital Applied — GA4 BigQuery Export 2026 Reference)
Setup
- Permissions required: Editor or Administrator on the GA4 property + Owner-level access to the BigQuery project.
- Navigate to Google Analytics Admin > Product Links > BigQuery Links; data should start flowing within 24 hours.
- The data location chosen at link creation cannot be changed — switching it later requires deleting and recreating the link, opening a gap in exported data.
- If a GA4 property and a Firebase project are integrated, they cannot be linked to separate BigQuery projects — they must share the same Cloud project.
- Individual data streams and events can be excluded from the export during or after link setup.
- If the property's timezone is changed, the export window shortens or lengthens on that day, producing an unusual event count.
(Google Analytics Help Center — Set up BigQuery Export)
Ecommerce use cases
The export is the foundation of serious GA4-based analytics stacks for ecommerce. Confirmed practitioner use cases (as-of 2026-05-29):
- Custom attribution modeling — raw event rows allow custom attribution windows, multi-touch attribution, and CRM joining, bypassing GA4's sampled UI reports; this is the primary practitioner workaround cited in GA4 Attribution pages.
- Unbounded historical analysis — GA4's UI caps data retention at 14 months; BigQuery stores data indefinitely.
- Bypassing UI sampling — GA4 Explorations apply sampling above ~10 million events or high-cardinality dimensions; BigQuery contains 100% of collected events.
- Joining external data — combining GA4 event-level data with CRM, OMS, or ad platform records for revenue attribution and LTV modeling.
- Feeding BI and data warehouse tools — BigQuery export is the standard first step for teams using Looker Studio, dbt, Tableau, or other downstream tools.
- Google publishes an obfuscated sample ecommerce dataset at
bigquery-public-data.ga4_obfuscated_sample_ecommerce(3 months of real GA4 ecommerce events from the Google Merchandise Store, Nov 1 2020 – Jan 31 2021) accessible in BigQuery Sandbox with no billing required.
(Digital Applied — GA4 BigQuery Export 2026 Reference); (Find Anomaly — GA4 BigQuery Export complete guide 2026)
Analytics 360 SLA
For Analytics 360 properties using Fresh Daily export (the mode required to be covered by the SLA):
| Property size | SLA coverage |
|---|---|
| Normal (< 25B events collected) | Covered |
| Large (≥ 25B events) | Covered |
| XL (≥ 250B events) | No SLA |
The SLA on Fresh Daily export is not currently available (the SLA applies to a different coverage tier). To be covered, users must select the Fresh Daily export option — not Daily or Streaming.
(Google Analytics Help Center — BigQuery Export SLA)
UI vs BigQuery discrepancies
The GA4 interface and BigQuery export consistently differ in ways that are documented (as-of 2026-09-26):
- Modeled conversions not in BQ — GA4's UI can show processed, estimated (modeled) numbers; BigQuery contains only raw event data.
- Privacy thresholding not applied in BQ — GA4 UI hides small user cohorts; BigQuery shows raw counts.
- Sampling not applied in BQ — GA4 Explorations sample above an unofficial ~10 million event threshold; BigQuery is always unsampled.
- Streaming tables missing new-user attribution —
traffic_source.*fields are not populated for new users in streaming intraday tables. - wBRAID/gBRAID not in BQ — certain click identifiers are excluded.
- June 2026 incomplete export incident — a Google Analytics Community thread (2026-06-01) reported daily export delivering only a fraction of events for a specific date while the GA4 UI showed full data; no retroactive reprocessing. (single community report; not confirmed as a resolved systemic Google incident)
(Find Anomaly — GA4 BigQuery Export complete guide 2026); (Google Analytics Help Center — BigQuery Export overview)
Contradictions
GA4 Explorations sampling threshold: Digital Applied (2026-05-29) states Google "does not publish an official sampling threshold and it varies by report type and date range." Multiple practitioner sources (citing GrowthNirvana 2025) name 10 million events as the threshold. The 10M figure is widely repeated but not confirmed in official Google documentation. Sources: Digital Applied 2026 | Seresa 2026
Key terms
| Term | Meaning |
|---|---|
events_YYYYMMDD | Date-sharded daily export table, one per day |
events_intraday_YYYYMMDD | Streaming export table, deleted at day-end |
analytics_<property_id> | BigQuery dataset created by the export link |
event_params | REPEATED RECORD containing custom event parameters, requires UNNEST |
session_traffic_source_last_click | Session-scoped last-non-direct-click traffic attribution structure, added July 2024 |
| Date-sharding | GA4's default pattern of one table per day (vs a single partitioned table) |
| Fresh Daily | GA360-only export mode with ~5am delivery SLA |
| Standard property limit | 1 million events/day for Daily export (as-of 2026-09-26) |