Events

Overview

The sessions table is a flattened date-partitioned session-level aggregation of the events table. The attibution model defaults to last_click, but can also be configured as first_click, first_click_non_direct or last_click_non_direct.

One row represents one single discrete session.

Schema

Aggregation Columns

The following columns are used as aggregation dimensions, meaning that one row will correspond to a unique combination of these column values.

COLUMN NAME	DATA TYPE	DESCRIPTION
session_id	STRING	Unique ID per session
project_id	STRING	project_id containing the ga4_dataset_id
dataset_name	STRING	dataset_name of the GA4 dataset
analytics_property_id	STRING	analytics_property_id of the GA4 dataset
event_id	STRING	Unique ID per event
stream_id	STRING	ID of GA4 stream
platform	STRING	Web or App
user_pseudo_id	STRING	Unique ID corresponding to identified user

Aggregated Metadata Columns

The following columns are aggregated in order to make sense at a session-level aggregation.

COLUMN NAME	DATA TYPE	DEFINITION	DESCRIPTION
event_date	DATE	MIN(event_date)	The event_date of the first occurring event in a session
session_start_timestamp	DATE	MIN(event_timestamp)	The event_timestamp of the first occurring event in a session
session_end_timestamp	TIMESTAMP	MAX(event_timestamp)	The event_timestamp of the last occurring event in a session
event_names	ARRAY	ARRAY_AGG(DISTINCT event_name IGNORE NULLS)	An array of unique event_name values in a session

Channel Grouping Column

The channel_grouping STRING column is added, in alignment with the Google Default channel group logic. This enables the last_click_non_direct and last_click_non_direct attribution models.

Aggregated Metric Columns

All columns in the count metric STRUCT in the events table are summed across all events in a session, with the naming convention unchanged (e.g. SUM(count.page_view) AS page_view). Additionally, the following fields are summed across all event metrics:

COLUMN NAME	DATA TYPE	DEFINITION	DESCRIPTION
event_value_in_usd	FLOAT	SUM(event_value_in_usd)	Sum of event_value_in_usd across all events in a session
count.total_sessions	INTEGER	1 AS total_sessions	An integer flag to enable total_sessions to be used as an output metric
count.total_events	INTEGER	SUM(count.total_events)	The count of events in a session
count.total_conversion_events	INTEGER	SUM(count.total_conversions)	Total conversion events in a session
count.total_conversion_sessions	INTEGER	MAX(count.total_conversions)	Flag (1) if a session contains at least one conversion event
parameter.total_engagement_time_msec	FLOAT	SUM(parameter.engagement_time_msec)	The sum of engagement time across all session events

Attributed Columns

The following column values are selected based on the attribution model configured upon deployment(first_click, last_click, first_click_non_direct and last_click_non_direct), with direct-attributed events defined by the channel_grouping. The attribution model for the deployment is then recorded in the attribution_model STRING column.

COLUMN NAME	DATA TYPE
privacy_info	STRUCT
user_first_touch_timestamp	TIMESTAMP
user_ltv	STRUCT
device	STRUCT
geo	STRUCT
app_info	STRUCT
traffic_source	STRUCT
event_dimensions	STRUCT
collected_traffic_source	STRUCT
channel_grouping	STRING
local	STRUCT

Aggregated Array Columns

Additionally, for parameter and property values observed at the event level, ARRAY fields of unique, non-null values are included for downstream anaytical purposes. The naming convention is unchanged from the event table.

If specific parameter or property sub-columns require different aggregation treatment (e.g. SUM), this is configurable upon deployment. By default the only parameter sub-column which is summed is parameter.total_engagement_time_msec, which is defined as SUM(parameter.engagement_time_msec) across all session events.