Data Manager in GA4: What the New Admin Hub Changes
Ask three people on a client account where a data connection was set up and you get three answers. One points at Admin, then Product links. One points at Data streams. The third opens the Google Ads account and is certain the link lives there. All three can be right, because the controls have sat on separate screens for years.
On 6 November 2026 Google put them on one page. Data Manager arrived in the GA4 admin console in beta: a single hub for product links, data streams and first-party data, with diagnostic alerts and connection health in the same view. It is a real change to how a property gets administered, and it ships with one warning that matters more than the interface.
What Google actually shipped
The changelog entry describes a unified hub in the admin console that lets you connect and control product links, data streams and first-party data. It adds two things the older screens never offered in one place: addressing diagnostic alerts, and monitoring connection health.
The help page is more specific. It calls Data Manager a centralized data management tool for managing data streams and connected products from a single home page, with a visual link between data sources and their usage. The word beta sits in the page title, so treat the layout as provisional and the concept as settled.
Where it sits, and who can open it
Admin, then the Data Manager tab in the left navigation, below My preferences. The documentation names no enablement step and no rollout toggle.
Two limits are worth knowing before you promise a client a complete view. The help page says you need the appropriate permissions for the property and does not name a role, so test with the account you actually use instead of assuming Editor is enough.
The harder limit: subproperties and roll-up properties do not support the Data Manager landing page. If your reporting model is a rollup over regional properties, this is not the audit screen for it, and the older screens still carry that work.
Four jobs the page claims
- Centralized management. One place to set up, manage and troubleshoot a property's data setup.
- Simplified ingestion. The documentation promises expanded connector availability. The help centre already carries separate setup pages for Meta, TikTok, Pinterest, Snap and Reddit, so the connector set reaches well past the Google-only links the old screens implied.
- Data visualization. The connection map, which gets its own section below.
- Proactive data strength. The data strength guide is meant to surface measurement gaps and recommend actions based on specific signals. The documentation stays general here, so read it as a prompt to check something, not a score to report to a client.
Reading the connection map
The map puts the property in the middle, sources flowing in from the left and destinations flowing out to the right. Three views stack: a mini view as the default, a collapsed list of tag or SDK, connected products, property and streams, and data export. Then an expanded view. Then a full page view.
The expanded view carries the detail, in four columns. Data sources covers Tag Manager, a CRM platform, a Google Ads product link or an offline file. Data types covers the import types in use. Property and streams shows how the data attaches to your property, including web and app stream detail. Data export lists the connected products receiving events.
Product-linking integrations skip the data types column. That absence is a useful tell: a node with no data type is a link, not an import.
Four interactions do the work. Hovering a node traces the full lineage from source to destination. Clicking opens a summary card. The card carries a filter icon that narrows the map to the selected node. Grouped nodes expand and collapse, and the filter accepts a column name as well as a node name, so every export can be isolated in one move.
The expanded view also resolves identifiers the old screens hid. Google gives Tag IDs and Google Sheet names as examples of what appears at that zoom. That is what makes the map usable for tracing one specific import rather than admiring the overall shape.
When a connection is broken
A source with a problem shows a broken state or a red status indicator. Clicking it opens a summary card with the issue and a Fix action that resolves it in place. The diagnosis and the repair end up on the same screen as the evidence, which cuts a support round trip out of the process.
Two lists below the map
Data streams and Connected products sit underneath. Streams get an Add stream button, connected products get Connect Product. The connect flow asks for a data source, then how you will use that source, then begins setup.
Deletion is not in the list. It lives in the Manage Connection panel, opened from the connected product. Worth knowing before you walk a client through removing a stale link over a screen share.
The warning that matters
Google's documentation is blunt about direction. Data ingested through Data Manager in GA4 flows downstream to linked Google Ads and Google Marketing Platform accounts, and not the other way around.
Three rules follow, and this is the part of the release that can cost money:
- New connections start in GA4. Configure them in the property, and the flow runs from there to the linked accounts.
- Existing connections stay where they are. Keep managing them in Google Ads or GMP, and do not recreate them in GA4.
- Existing imports appear in Data Manager automatically, so an empty-looking map is not permission to rebuild a pipeline.
The explicit warning is about duplication: do not activate an identical data pipeline in both interfaces, because conversions can be logged twice.
Double counting is not a cosmetic error. It inflates conversion counts, distorts bidding in any connected ad account, and produces a client report that cannot be reconciled against the platform's own figures. If a conversion import was already running through Google Ads and someone sets the same import up in GA4 because the new screen made it look unconfigured, the second pipeline is the fault.
Two data types are GA4-only and must be configured in Data Manager: item data and third-party campaign data. There is no alternative path for either, which makes them the exception to the leave-existing-connections-alone rule.
How to check for duplication in ten minutes
- Open the expanded map and trace every node that ends in a Google Ads or GMP destination.
- For each one, write down which interface owns the setup. If a conversion import appears on both sides, one of them has to go.
- Pick one conversion action, take a 30 day range, and compare the count in GA4 with the count in Google Ads.
- If GA4 reads close to double, find the second pipeline before changing any bidding settings.
What this does not fix
Data Manager is an administration surface, and that limits what it can tell you.
- It reports connection health, not tagging quality. A stream that loads the config tag on the wrong trigger still records a session with no source, and the map shows a healthy connection the whole time.
- It does not touch consent. Denied analytics_storage is not a broken connection, and no Fix button brings those events back.
- It leaves subproperties and rollups out of the landing page.
- It is beta. Names, positions and view labels can move without a changelog entry you would notice.
Keep the tag-side tools for tag-side questions. Tag coverage and tag diagnostics check the pages. Data Manager checks the pipes between products. A property can pass one and fail the other.
The first-party data question
The third item in the hub carries legal weight. Uploading customer records, order history or hashed email addresses into GA4 is a personal data processing activity under UK GDPR, and a new admin screen does not change the obligations attached to it.
Four things to confirm before an import runs. There is a lawful basis and a stated purpose a customer would recognise. The fields are minimised, and identifiers arrive hashed where the connector supports it. The retention you configured matches what the privacy notice says. The processor terms with Google cover what you send, including any category of data you should not be sending at all.
The rules on storage and access cover how data was collected on the site. They do not authorise an import. Those are two separate questions, and a client who can point at a cookie banner has answered only one of them.
Do this while the beta is still young
Nothing forces a migration today. Existing connections keep running, and the help page names no deadline.
What the release does offer is an audit view that did not exist last week. The duplication risk was always there, buried across three interfaces that nobody compared side by side. The map puts it on one screen, which is reason enough to spend half an hour on it now, while a client is still willing to hear that a conversion may be counted twice.
Running several properties and unsure who owns which connection? Book a free tracking audit and we will map the pipelines, the owners and the duplicate imports.