Web analytics
Advertising platforms show you their own view of your data and nothing else. Web analytics gives you the event stream somewhere you control: your own warehouse, so you can join it to revenue and keep history beyond a platform’s retention window, or the analytics tools your teams already read.
Sources and destinations
Section titled “Sources and destinations”All three destinations take the same sources: the website tag, through dhPixel or NeoTag, and the e-commerce platforms.
Supported destinations
| Category | Supported |
|---|---|
| Warehouse | |
| Analytics |
The same sources that feed your conversion destinations feed these. You do not instrument the site twice.
Choosing a destination
Section titled “Choosing a destination”The three answer different questions, and nothing stops you connecting more than one.
- BigQuery is where the raw event stream lands in your own Google Cloud project so you can query it yourself. Nothing is modeled for you, and the analysis happens in your own tooling.
- Google Analytics 4 is server-side delivery into a property your team already reports on, so events survive ad blockers and dropped page loads that the browser tag loses.
- Mixpanel is product analytics: journeys, feature adoption and retention, with events attached to a person by the primary identifier you nominate.
Identifiers behave differently on each, and that is the thing to settle first. BigQuery receives what the source sends without hashing, because it is your own warehouse. GA4 matches on a client ID rather than on personal identifiers, and Google’s terms prohibit sending personal data into event parameters. Mixpanel receives identifiers hashed, so an analyst sees a consistent hash rather than a readable email address.
Connecting BigQuery as a destination
Section titled “Connecting BigQuery as a destination”This connector uses a Google Cloud service account key. As a destination the service account needs both of these roles, because it writes rather than reads:
- BigQuery Data Editor
- BigQuery Job User
Grant the roles on the dataset rather than the whole project, and keep the JSON key secure. It is a credential rather than a configuration value, and Google will not show it again after it is created.
What lands
Section titled “What lands”Events from the connected source are written to the BigQuery dataset you nominate. Signals does not hash on the way, so what lands is what the source sent. Events collected by the website tag arrive hashed, because the tag hashes identifiers in the browser before they leave the page. Plan your joins around hashed values rather than expecting raw email addresses or phone numbers. To join against your own tables, hash the other side the same way: Apply SHA-256 hashing to PII has the BigQuery expression.
GA4 and Mixpanel receive the events themselves rather than a table. What you see there is whatever the connected source sends, named by the event names you agreed at source, so settle those before you go live: renaming an event afterwards leaves the old name in every saved report.
What this is not
Section titled “What this is not”This use case lands data in a warehouse, or delivers it to an analytics tool you already run. It does not provide dashboards, reports or an analytics interface inside Studio. Analysis happens in your own tooling.