The distinction isn’t academic: editing the wrong kind of data source silently changes reports you’ve never opened. Most people meet this the hard way, usually when a colleague asks why their dashboard has gone red.
This explains what a data source actually is, the choice you make without noticing every time you connect something, and the two decisions in here that can’t be undone.
Dataset, connector, data source, report
Four things, in a chain. Getting them straight makes everything else in Data Studio easier to reason about.
| What it is | Example | Lives | |
|---|---|---|---|
| Dataset | Your underlying raw data. Data Studio doesn’t own it or store it. | A GA4 property, a spreadsheet, a BigQuery table | In the other platform |
| Connector | The pipe. Knows how to talk to that platform and fetch rows. | The Google Ads connector | Provided by Google or a third party |
| Data source | Data Studio’s model of the dataset: field names, types, aggregations, calculated fields, credentials. | “GA4 — client site (cleaned)” | In Data Studio |
| Report | The charts, arranged on pages. | Your monthly dashboard | In Data Studio |
The key idea: a data source is not a copy of your data. It’s a description of it — what the fields are called, whether a column is a date or text, what should be summed and what should be averaged. Data Studio queries the underlying dataset live, every time, through that description.
Which is why a data source can break without anything happening to your data, and why fixing a field type here fixes it in every chart at once.
Embedded or reusable — the choice you make without noticing
Every data source is one of two kinds, and you pick one implicitly the moment you connect something.
| Embedded | Reusable | |
|---|---|---|
| Lives | Inside one report | In your account, independent of any report |
| Created when | You add data from inside a report | You create it from the Data Studio home page |
| Edits affect | Only that report | Every report using it |
| Travels when the report is copied | Yes | No |
| Good for | One-off reports, templates meant to be copied | Anything where you want one definition of a metric |
Neither is correct in general. Reusable is right when consistency matters — one definition of “qualified lead”, maintained once, used by six reports. Embedded is right when independence matters — a template someone will copy, or a report that shouldn’t be affected by someone else tidying up a shared source.
Why editing a reusable source breaks other people’s reports
Google states the mechanism plainly: editing a reusable data source — hiding fields, changing field types, removing fields — affects every report that uses that data source.
That’s not a warning about being careless. It’s the whole point of a reusable source, and it’s also its failure mode. The specific way it goes wrong is worth knowing by name, because it’s what you’ll see:
- Hide or remove a field, and charts elsewhere show
Invalid DimensionorInvalid Metric. The chart is looking for a field that no longer exists in the source. - Change a field’s type and charts that assumed the old type stop aggregating correctly, often without erroring — which is worse.
So if a report you haven’t touched in months suddenly shows Invalid Dimension, don’t debug the report. Find out who edited the data source.
The related surprise, which comes up constantly on forums: duplicate a report that uses a reusable source, edit the duplicate’s data source, and the original changes too. The duplicate didn’t get its own copy. It’s pointing at the same shared source.
Converting embedded to reusable — and why you can’t go back
You can promote an embedded source: Resource → Manage added data sources → Make reusable.
Two things to know before you click it:
- It’s a one-way door. Google is explicit that a reusable data source can’t be converted back to an embedded one. If you want that, you rebuild.
- You need to own the data credentials to do it. If someone else set the source up with their credentials, you can’t promote it.
The practical read: promote deliberately, when you’ve decided a definition should be shared. Don’t promote to tidy up.
What happens when someone copies your report
This is the agency handover failure, and it catches people at exactly the wrong moment.
- Embedded sources are copied with the report. The copy gets its own independent version.
- Reusable sources are not copied. The copy points at the original shared source — which the person you gave it to may not have access to.
- If an embedded source uses owner’s credentials and the person copying isn’t the owner, the source isn’t copied at all and the charts error out.
Which means: if you’re building a template for other people to copy, use embedded data sources. That’s the whole reason templates work the way they do, and it’s covered in copying and reusing a template. If you’re building a report for one client that you’ll maintain yourself, reusable is usually better.
Editing fields — what you can and can’t change
In the data source editor you can change quite a lot:
- Field names — rename
ga:sessionsto something a client understands - Field types — text to date, number to currency, and so on
- Default aggregation — sum, average, count distinct
- Calculated fields — new fields built from the existing ones
- Descriptions, and hiding fields you don’t want cluttering the panel
What you can’t edit is the field ID. That’s the internal identifier Data Studio uses to track a field, and it’s deliberately fixed — it’s what lets you rename a field without breaking every chart that uses it.
This one fact explains two things that otherwise look inconsistent. It’s why renaming is safe but removing isn’t. And it’s why filter controls behave the way they do across data sources, which matters enough that the filter controls post deals with it separately.
Ownership and credentials
Every data source carries credentials, and they decide who can see the data through it.
- Owner’s credentials — the source queries using your access. Viewers see the data whether or not they can reach the underlying platform. This is the default and normally what you want.
- Viewer’s credentials — the source queries using whoever is looking. People without access see nothing.
Two consequences that catch people out:
- Viewer’s credentials stop hyperlinks and images in dimensions from rendering. If your landing-page table has clickable links in one report and plain text in another, this is why.
- Owner’s credentials can fail on BigQuery external tables backed by Drive or Sheets. A service account is the usual answer there.
Also worth knowing if you’re in a managed environment: a Workspace admin can disable users’ ability to set data credentials at all. If the option is greyed out, that’s your answer and it isn’t a bug.
Credentials are the most common reason a shared report looks empty to the person you sent it to.
How to check you’ve set this up right
Before you hand a report to anyone, four checks:
- Open Resource → Manage added data sources. Every source is labelled embedded or reusable. Confirm each one is the kind you meant.
- Check the credentials on each. Owner’s for anything a client will open.
- Check field types — especially dates, which are the most common thing to come through as text.
- Open the report as a viewer, not as yourself. It’s the only way to see what they’ll see.
One thing worth being honest about: there’s no good way to find which charts use a given data source. It’s one of the most-asked and least-answered questions about the tool. The nearest workaround is to open Manage added data sources, remove the source, and see what breaks — then undo. Not elegant, and worth knowing that you’re not missing a feature.
Which to use, depending on how you work
In-house, a handful of reports you own: use reusable sources. You’re the only editor, consistency is worth more than isolation, and fixing a field type once is the entire benefit.
At an agency, one report per client: reusable per client, embedded for anything you hand over. The rule that prevents the most pain: if someone else will own it after you, embed it.
Building templates for other people: embedded, always. A template with a reusable source is a template nobody else can use.
Frequently asked questions
What’s the difference between a data source and a dataset?
The dataset is your underlying raw data — the GA4 property, the spreadsheet, the BigQuery table — and it lives in that platform. The data source is Data Studio’s model of it: field names, types, aggregations, calculated fields and credentials. Data Studio doesn’t store your data; it queries the dataset live through the data source.
Should I use an embedded or a reusable data source?
Reusable when consistency matters and you’re the one maintaining it — one definition, many reports. Embedded when independence matters, especially for templates other people will copy.
I edited a data source and now another report is broken. Why?
You edited a reusable source, and edits to a reusable source affect every report using it. Hiding or removing a field produces Invalid Dimension or Invalid Metric errors in charts elsewhere.
I duplicated a report, edited the copy’s data source, and the original changed. Why?
The report used a reusable data source, and duplicating a report doesn’t duplicate that — both reports point at the same shared source.
Can I turn a reusable data source back into an embedded one?
No. Embedded to reusable is possible via Resource → Manage added data sources → Make reusable, and it requires that you own the data credentials. The reverse isn’t supported — you’d have to rebuild.
If I copy a report, do the data sources come with it?
Embedded ones do. Reusable ones don’t — the copy points at the original. And if an embedded source uses owner’s credentials and the person copying isn’t the owner, it isn’t copied and the charts error.
How do I find which charts use a particular data source?
There’s no built-in way, which is a genuine gap. The workaround is to remove the source under Manage added data sources, see what breaks, and undo.
Can I edit a field’s ID?
No. Field IDs are fixed internal identifiers, which is exactly why you can rename a field without breaking the charts that use it.
What’s the difference between owner’s and viewer’s credentials?
Owner’s means the data source queries with your access, so viewers see data whether or not they can reach the underlying platform. Viewer’s means it queries with theirs, so people without access see nothing. Owner’s is the default and right for most client reports.
Is Looker Studio still called Data Studio?
Data Studio is the current name — Google renamed it Looker Studio and then back again. Same product either way. More on that in what is Data Studio.