On April 18th, Google published the results from Google Ads’ interest-based advertising testing, where, among other tools, Topics API was analyzed. We are happy that the Google Ads team joined RTB House and Criteo in openly publishing findings from experiments with Privacy Sandbox APIs.
In the report, the Google Ads team compares the effectiveness of applying Topics API, contextual signals, and publisher first-party IDs instead of cookies for Interest-Based Advertising (IBA). Third-party cookies were allowed in this experiment for use cases unrelated to targeting, such as frequency capping and measurement.
Table of Contents:
- What we, participants in cookieless preparations, read from the report
- What can advertisers read from the report?
- RTB House recommendations
What we, participants in cookieless preparations, read from the report
First of all, as noted above, we warmly received the publication of this report, since it helps to facilitate fruitful discussion about the Privacy Sandbox APIs. We acknowledge that there are numerous limitations in the current ecosystem that make it difficult to simulate the future world without cookies, as third-party cookies are currently available to everyone in Chrome. In theory, Google could simulate traffic without third-party cookies, but any other ad tech provider could not currently repeat such a test.
As mentioned in previous publications, testing on a small sample of users with third-party cookies turned off is inevitable and should be started as soon as possible. However, it must be available to any entity willing to experiment to make testing constructive. Initially, such a sample would also have limitations; for example, there would be a limited number of buyers capable of bidding on such inventory. On the other hand, such an environment would be closer to what can be expected in the cookieless future and could allow for the best use of all new privacy-preserving technologies.
We would also like more transparency on the data used for this experiment. We acknowledge Google’s message that we can provide similar results to what is available today only by combining Topics API with other technologies. Still, it would be great to understand how specific tools were used. This is especially related to the simulated first-party IDs derived from third-party cookies. Google assumed that all publishers would be willing to share their first-party IDs with them. It is hard to say how it would work in practice, especially since such IDs could be shared in multiple ways. Perhaps it would be worth exploring whether Google could reach similar results when using the Secure Signals feature – available to third-party buying platforms to ensure a fair and level playing field.
Lastly, we welcome Google’s recommendation that Topics should be more granular. This has been our impression ever since the original Topics API announcement. We believe that it would make Topics a more effective tool for advertising use cases.
What can advertisers read from the report?
Advertisers may be interested in learning more about how this experiment reflects their specific use case. Advertisers operate in multiple verticals, differing by the type of business (e.g., marketplaces, click-through websites, classifieds) and product portfolio (e.g., fashion, home & garden, finance). The differences between these businesses imply different approaches to e.g., what data is used for advertising. Simply speaking, advertisers would be interested in learning if their campaigns could achieve these specific results.
Also, advertisers would definitely want to learn about the measurement and frequency capping possibilities when using Topics. They would like to assess the effectiveness of the specific tools used in their campaigns, which requires proper measurement. At the same time, frequency management has become essential in most advertising campaigns today.
RTB House recommendations
As noted above, the current environment makes it challenging to run quantitative experiments properly, and any results will be questionable. Therefore, moving to an environment that mirrors the future state as closely as possible is inevitable and should happen soon. We believe that starting a discussion on when to separate a no-cookie sample and how to do it is needed.
In the report, Google frequently mentions the use of publisher-partitioned IDs, which can help profile users. In this context, it is worth looking into the Protected Audience API (PAAPI, FKA FLEDGE), as it could further improve results. PAAPI allows for building audiences on the publisher’s website and using it on the publisher’s inventory and across the web, with full publisher control and no leakage of user-level data.
Google has also indicated that it intends to run more tests. We would like to see more analyses done in a way that any third-party company could repeat. It would also require more transparency on what data has been used and in what way. This is inevitable if we want to reach a competitive, privacy-protecting advertising ecosystem without any company building an unfair advantage.
If you have any questions, comments or issues, or you’re interested in meeting with us, please get in touch.