BigQuery streaming export

In this article:

About streaming export

BigQuery streaming export makes fresher data for the current day available within a few minutes via BigQuery Export.

When you use this export option, BigQuery will have more recent information you can analyze about your users and their traffic on your property.

For each day, streaming export creates 1 new table and 1 (BigQuery) view of that table:

  • Table: ga_realtime_sessions_YYYYMMDD is an internal staging table that includes all records of sessions for all activity that took place during the day. Data is exported continuously approximately every 15 minutes. Within this table are multiple records of a session when the session spans multiple export operations.
     
    The ga_realtime_sessions_YYYYMMDD tables should not be used (and are not supported by Google Analytics technical support) for queries. Queries on these tables may yield unexpected results as they may contain duplicate records of some sessions. Query the ga_realtime_sessions_view_YYYYMMDD view instead.
  • View: ga_realtime_sessions_view_YYYYMMDD sits on top of the exported tables and is there to deduplicate multiple records of repeated sessions that exist across export boundaries. Query this table for deduplicated streaming data. Learn more about BigQuery Views

Query performance

While ga_realtime_sessions_view serves to deduplicate users and sessions, the deduplication adds an additional computational step for each query, which increases query time. The increase in query time varies depending on data volume, and so varies from client to client.

The increase in query time is, however, mitigated by the overall increase in data freshness and your opportunity to respond to more current data.

Billing

You will incur additional costs for using streaming export at the rate of $0.05 per gigabyte of data. 1 gigabyte equates to approximately 600,000 Google Analytics hits, though that number will vary depending on hit size. These additional costs are covered by the existing BigQuery coupon that every Google Analytics 360 customer receives. Learn more bout BigQuery pricing

Field coverage

Available fields

Most first-class Analytics dimensions (native, non-widened dimensions available in standard reports) are available, except as noted below.

Unavailable fields

Data that is widened through other Ads sources like Google Ads, Campaign Manager, Google Ad Manager etc. is not available.

Channel Grouping data is not available.

The following fields are not available:

  • userId
  • trafficSource.*
  • hits.latencyTracking.*
  • hits.publisher.*

Prerequisites

Google Analytics 360 views from which you export data to BigQuery must be eligible for enhanced data freshness.

Implement

If you haven’t already done so, set up BigQuery Export and link BigQuery to your Analytics property.

After you’ve set up BigQuery Export:

  1. Sign in to Google Analytics. Use an email address that has OWNER access to the BigQuery project, and also has Edit permission for the Analytics property that includes the view you want to link.
  2. Click Admin, and navigate to the property that contains the view whose data you want to export.
  3. In the PROPERTY column, click All Products > BigQuery > Adjust link.
  4. Select Data exported continuously.
  5. Click Continue.
  6. Click Done.

Impact of changing export frequency

Changing from Data exported in batch multiple times a day to Data exported continually: Changes for a property do not take effect until 12 am the following day, based on the view with the earliest time-zone setting in the property.

Changing from Data exported continually to Data exported in batch multiple times a day: Streaming export is deactivated immediately and we will stop streaming data within a few hours. Data will start exporting during the next regular export window (exports occur in batch multiple times per day). That next intraday export will contain the complete dataset for the current day, as is expected with this export frequency selection.

Was this article helpful?
How can we improve it?