Menu

What Data Signals Could Google Use to Build Its Audience Segments? 🕵️

By Max Legrand · GreyFox Ads · Programmatic Advertising

Google audience data signals and Android app tracking protection experiment — GreyFox Ads

What actually goes into Google's audience segments in Display & Video 360 (DV360) and Google Ads? We went looking for clues with DuckDuckGo's Android App Tracking Protection—and the results were surprising. 🍿

In brief: On one Android device, over seven days and 19 apps, DuckDuckGo reported 27,579 blocked tracking attempts. Google-associated trackers appeared in 18 of those apps. The dashboard exposed numerous potential data categories, but it did not prove Google collected every field or used them to build advertising audiences.
In this article

1. What signals might Google use to create audience segments?

We've previously explored the mystery surrounding Google's audience segments in DV360 and Google Ads (original article in French). One question kept nagging at us: what kinds of user and device signals might contribute to audience modelling?

Google's precise segmentation recipes are not fully public. Rather than pretending to have reverse-engineered them, we decided to investigate something more concrete: the types of tracking activity visible on an Android phone.

2. What is DuckDuckGo's App Tracking Protection?

You probably know DuckDuckGo, the privacy-focused search engine. 🦆 Its Android app also offers App Tracking Protection, a feature designed to identify and block certain third-party tracking requests originating from other apps.

Its dashboard can show the companies associated with detected tracking attempts, the apps involved, categories of information that trackers may seek, and the number of requests blocked. It's a useful window into app tracking activity—but it is not an audit of Google's internal advertising systems.

3. Our seven-day test: 27,579 blocked tracking attempts

We enabled App Tracking Protection on one Android device and used 19 apps over seven days. According to the DuckDuckGo dashboard, it blocked 27,579 tracking attempts. 🤯

Google-associated trackers appeared in 18 of the 19 apps; Netflix was the exception in our particular sample.

ObservationOur test resultHow to interpret it
Observation period7 daysA short test, not a longitudinal study
Apps used19One device and one app selection
Blocked tracking attempts27,579Requests blocked, not unique personal data items
Apps with Google-associated trackers18 of 19Detected association, not proof of advertising use

That's a lot of activity. But it is not a representative survey of Android users (and repeated requests can inflate the count without implying additional types of data were transmitted).

4. Which potential data signals did the dashboard identify?

The dashboard associated Google-related tracking attempts with a striking range of potential data categories. Here's the inventory from our test, grouped to make it easier to read.

Identity and advertising identifiers

Location and network

Device and technical information

App information

Quite an ingredient list, isn't it? 👀 Some of these categories are useful for basic app functionality or measurement; others could be relevant to analytics, device recognition or advertising in particular contexts. Their appearance on a dashboard does not establish the purpose of a specific request.

5. Does this prove Google uses all these signals for advertising audiences?

No. And this distinction matters more than the impressive headline number.

DuckDuckGo's dashboard describes tracking activity and categories of information trackers may seek. It does not demonstrate that each field was actually transmitted, received, retained, linked to an individual or incorporated into an audience segment in Google Ads or DV360.

Some requests can support app analytics, security, diagnostics, measurement or other services—not necessarily audience targeting. Applicable consent requirements, privacy controls and contractual restrictions also affect what can be collected and used.

So have we cracked Google's secret audience-segmentation formula? Not quite. 🔍 But we have illustrated how broad the universe of possible mobile signals can be—and why marketers should demand clear explanations of what sits behind an audience product.

6. What should programmatic Media Traders ask about audience data?

When buying an audience segment, estimated reach alone tells only part of the story. Here are five practical questions worth asking:

  1. Provenance: Where do the underlying signals originate—first-party interactions, publishers, apps or other sources?
  2. Permissions: What consent, privacy and contractual restrictions apply to collection and activation?
  3. Methodology: Is the audience directly observed, inferred or modelled? How frequently is it refreshed?
  4. Incrementality: Does the segment reach genuinely additional people, or mainly those you could already reach?
  5. Performance: How will you test outcomes against an appropriate control or baseline?

For related reading, see our analysis of the strange thing about Google data segments, incremental targeting strategies and fingerprinting in programmatic advertising.

7. How GreyFox Ads approaches high-quality Open Web advertising

Here in France, GreyFox Ads is building a more transparent and effective way to activate programmatic campaigns across the Open Web. Our Master Curated Deal brings together:

Want to explore the wider context? Read our articles on programmatic curation and financial transparency in programmatic advertising.

Looking for a quality-focused programmatic advertising partner?

📧 Contact GreyFox Ads to discuss your Display and Video advertising challenges in France.

Originally discussed on LinkedIn.

About the author

Max Legrand is the founder of GreyFox Ads, a French company specializing in programmatic curation and quality-focused Open Web advertising.