What should a team validate first when ga4 bigquery schema cheat sheet: every field, what it does appears?

Reproduce the problem in the live implementation, isolate whether it is scoped to one report or flow, and compare it against at least one secondary source before changing the setup.

How do I know whether the fix actually worked?

You need before-and-after evidence in the browser and in the downstream report. A clean-looking dashboard without validation is not enough.

When should this become a full GA4 audit instead of a quick fix?

If the issue touches attribution, consent, revenue, campaign quality, or data trust for more than one workflow, it is usually safer to audit the surrounding implementation than patch only the visible symptom.

GA4 BigQuery Schema Cheat Sheet: Every Field, What It Does (2026)

What's in the GA4 BigQuery export schema?

The GA4 BigQuery export contains 30+ top-level fields per event row, organised into seven groups: identity (user_pseudo_id, user_id), timing (event_timestamp, event_date), event metadata (event_name, event_params), traffic source (traffic_source.*, session_traffic_source_last_click.* added 2024), device and geo (device.*, geo.*), e-commerce (ecommerce.*, items array), and derived fields (privacy_info.*, is_active_user, session_traffic_source_last_click.*).

Three of those fields are nested arrays — event_params, user_properties, and items — which need UNNEST() to query. The schema is consistent across all GA4 properties; the only variation is whether session_traffic_source_last_click is present (depends on export configuration date).

This post is the reference. Bookmark it.

The complete schema (30+ top-level fields)

Identity fields

Field	Type	What it contains
`user_pseudo_id`	STRING	Anonymous user ID (the GA cookie value). Persistent across sessions for the same browser.
`user_id`	STRING	Custom user ID set via the User-ID feature. NULL if not implemented.
`pseudo_user_properties`	RECORD (repeated)	User-scoped properties set with set_user_properties calls.
`user_first_touch_timestamp`	INT64	Timestamp of the user's first touch (first session start).

Timing fields

Field	Type	What it contains
`event_timestamp`	INT64	Microseconds since Unix epoch when the event was logged client-side.
`event_date`	STRING	Event date in YYYYMMDD format (the table suffix).
`event_previous_timestamp`	INT64	Microseconds since Unix epoch of the previous event from the same user.
`event_server_timestamp_offset`	INT64	Difference between client and server timestamp in microseconds.

Event metadata

Field	Type	What it contains
`event_name`	STRING	The event name (e.g., page_view, purchase, session_start).
`event_params`	RECORD (repeated)	Array of key-value parameters attached to the event. Critical field — most analysis lives here.
`event_value_in_usd`	FLOAT64	Currency-converted value if event has a value parameter.
`event_bundle_sequence_id`	INT64	The sequential ID of the bundle this event was uploaded in.
`stream_id`	STRING	Identifies the data stream the event came from.
`platform`	STRING	The platform that generated the event (WEB, IOS, ANDROID).

Traffic source

Field	Type	What it contains
`traffic_source.source`	STRING	First-touch source (lifetime user attribution).
`traffic_source.medium`	STRING	First-touch medium.
`traffic_source.name`	STRING	First-touch campaign name.
`session_traffic_source_last_click.manual_campaign`	RECORD	Last-click source for the current session (added 2024).
`collected_traffic_source`	RECORD	Source data from the actual collected event (newer field).

The session_traffic_source_last_click field, added in 2024, gives session-level last-click attribution directly. Before this field, you had to extract source/medium from the session_start event's params.

Device and geo

Field	Type	What it contains
`device.category`	STRING	desktop, mobile, or tablet.
`device.mobile_brand_name`	STRING	Apple, Samsung, etc.
`device.mobile_model_name`	STRING	iPhone 15, Galaxy S24, etc.
`device.operating_system`	STRING	iOS, Android, Windows, macOS, Linux.
`device.operating_system_version`	STRING	OS version string.
`device.web_info.browser`	STRING	Chrome, Safari, Firefox, Edge.
`device.web_info.browser_version`	STRING	Browser version.
`device.language`	STRING	Browser language preference (en-us, fr-fr).
`geo.continent`	STRING	The continent.
`geo.country`	STRING	The country (e.g., "United Kingdom").
`geo.region`	STRING	Subnational region (state, province).
`geo.city`	STRING	City.
`geo.metro`	STRING	Metro area code (US-specific).

E-commerce

Field	Type	What it contains
`ecommerce.transaction_id`	STRING	Transaction ID for purchase events.
`ecommerce.purchase_revenue`	FLOAT64	Revenue from the transaction (excluding tax/shipping).
`ecommerce.purchase_revenue_in_usd`	FLOAT64	Revenue in USD.
`ecommerce.refund_value`	FLOAT64	Refund amount.
`ecommerce.shipping_value`	FLOAT64	Shipping cost.
`ecommerce.tax_value`	FLOAT64	Tax amount.
`ecommerce.unique_items`	INT64	Number of unique items in the transaction.
`ecommerce.total_item_quantity`	INT64	Total quantity across all items.
`items`	RECORD (repeated)	Array of items in the event (product details).

Derived and metadata

Field	Type	What it contains
`is_active_user`	BOOLEAN	Whether the user is considered active per GA4's definition.
`privacy_info.analytics_storage`	STRING	Consent state for analytics_storage.
`privacy_info.ads_storage`	STRING	Consent state for ad_storage.
`privacy_info.uses_transient_token`	STRING	Whether the user uses a transient token (consent-denied modelled events).
`app_info.id`	STRING	App package name (mobile only).
`app_info.version`	STRING	App version (mobile only).

Want to see which hidden implementation gaps are affecting your GA4 data quality?

Start free audit

How to unpack nested arrays

Three fields are repeated records that need UNNEST() to query:

event_params

The value.string_value, value.int_value, value.float_value, and value.double_value fields hold the actual parameter values depending on type. Use the COALESCE pattern when the type is mixed:

items

user_properties

Standard event_params worth knowing

Every GA4 event includes a default set of parameters. The most common:

Param key	Value type	What it contains
`ga_session_id`	int_value	The session identifier (used for sessionisation)
`ga_session_number`	int_value	Sequential session number for the user (1, 2, 3...)
`page_location`	string_value	Full URL of the page
`page_title`	string_value	The page title
`page_referrer`	string_value	The referring URL
`engagement_time_msec`	int_value	Engagement time in milliseconds for this event
`session_engaged`	int_value	Whether the session is engaged (1 or 0)
`entrances`	int_value	Whether this is the entry event of the session (1 or 0)
`source`, `medium`, `campaign`	string_value	Traffic source from the URL on the event
`gclid`, `dclid`	string_value	Click ID parameters when present

For purchase events specifically, additional params:

Param key	Value type	What it contains
`currency`	string_value	Currency code (USD, GBP, EUR)
`value`	float_value or int_value	Total transaction value
`transaction_id`	string_value	Transaction ID (mirrors ecommerce.transaction_id)
`coupon`	string_value	Coupon code applied

What changed in 2024–25

Three meaningful schema additions in the past 18 months:

`session_traffic_source_last_click` (added 2024). Top-level field giving session-level last-click attribution without needing to extract from session_start event params. Older properties may not have it for historical data — only forward.

`is_active_user` (added 2024). Boolean indicating whether GA4 considers the user active per its internal definition. Useful for filtering bot or spam events that GA4 has flagged as inactive.

`privacy_info.uses_transient_token` (added 2025). Indicates whether the user uses a transient token — the marker for consent-denied modelled events. Useful for separating modelled from observed data in BigQuery analyses.

Older properties may have legacy fields no longer documented (e.g., event_dimensions.hostname was deprecated in 2023 in favour of extracting hostname from page_location). When in doubt, query the schema directly: SELECT column_name, data_type FROM project.analytics_PROPERTY_ID.INFORMATION_SCHEMA.COLUMNS WHERE table_name LIKE 'events_%' LIMIT 100;