GOV.UK GA4 improvements roadmap
This page details upcoming and recent changes to the GOV.UK GA4 data collection and processing.
See the changelog for previous releases.
What we’re working on now
Creating useful tables joining GOV.UK GA4 and Knowledge Graph data
Creating summarised tables or views combining GOV.UK GA4 and Knowledge Graph data.
Evaluating the utility of implementing a GOV.UK GA4 GTM spam filter
Evaluating the utility of implementing a filter which ensures that we are only collecting GOV.UK GA4 data via our Google Tag Manager code (thereby hopefully reducing the incidence of ‘spam’).
Investigating using the User Deletion API to delete users’ GA4 data
Investigating using the User Deletion API to delete users’ GOV.UK GA4 data.
Enabling easier access to site search data
Creating simplified tables or views containing GOV.UK site search data in BigQuery.
Recently released
Resolved issues with some incomplete custom dimension fields in the GOV.UK GA4 flattened dataset
Issues with the query_string, ui_text, response, search_term, autocomplete_input, autocomplete_suggestions, and link_text fields in the GOV.UK GA4 flattened dataset have been resolved.
The processing for these fields was only including string values, but we observed that some data was coming through as having an integer or double data type.
We have now edited the processing so that integer and double type data will be included in these fields in the flattened table.
Page path (cleaned_page_location) field updated in the GOV.UK GA4 flattened dataset
The cleaned_page_location field in the GOV.UK GA4 flattened dataset has been updated following the implementation of the canonical URL.
The cleaned_page_location now defaults to the canonical_url value, with the hostname removed, if it is available.
If the canonical_url is null, the cleaned_page_location is calculated as before - by stripping the hostname, anchors and query string from the page_location.
Redaction added to target National Insurance numbers
Some users were spotted searching for their National Insurance numbers. We have added in additional redaction to ensure this data does not get collected in any search analytics data collection, and are deleting any collected National Insurance numbers that may have previously slipped through.
Changed processing of the content_language field in the GOV.UK GA4 flattened dataset
We have noticed unexpected values appearing in the content_language field - for example, where the language changes between events on the same page, or multiple times in one user’s session.
Previously, we were applying some processing to the content_language field to keep one consistent value across one user’s events on a given page, but we are no longer sure if that is the best approach and so we have removed this calculation.
The values in the field now simply reflect the values collected with each event, to help us debug these unexpected values.
Updated processing of the GOV.UK GA4 flattened dataset to use the custom timestamp
We have added the custom timestamp field to window functions previously only using the event_timestamp.
From the dates we tested this change with, we do not expect this to cause any difference in the values output - this is just being added as a fail-safe.
Resolved issues with some sessionised custom dimension fields in the GOV.UK GA4 flattened dataset
Issues were identified with fields which use window functions to generate session-scoped values in the GOV.UK GA4 flattened dataset.
The processing for the affected fields involved window functions which generated the output values based on the client ID, session ID and (in some cases) page location.
However, in these functions, the ga_session_id was not being correctly selected (the string value was being extracted instead of the integer value), and so data could be incorrect for users who had multiple sesssions in one day.
This impacted the content_language, session_source, session_medium, public_updated_at, updated_at, content_id, browse_topic, publishing_app, first_published_at, taxonomy_all_ids, status_code, withdrawn, document_type, history, taxonomy_main_id, taxonomy_all, taxonomy_main, taxonomy_level_1, full_taxonomy_DEPRECATED, full_taxonomy_ids_DEPRECATED, rendering_app, organisations, schema_name, term, content, and campaign_id fields.
In most cases the difference noticed when fixing the functions was minor (<0.05%), but session_source and session_medium were more significantly impacted (about 8% out).
We have now edited the processing to fix these fields in the flattened table.
Backfilled the GOV.UK GA4 flattened dataset
We have refreshed the entirety of the GOV.UK GA4 flattened dataset so that all processing improvements made over previous months are applied to the whole dataset.
Updates
As we’re still at an early stage, our plans may shift. We’ll update this page when this happens and add more detail when we can.