In the wake of the ad tech industry’s pivot towards a privacy-centric approach, AlgoriX has been relentlessly striving to equip partners with solutions that balance user privacy with optimal advertising and monetization results. Hence, when Google initiated the comprehensive rollout of the third-party cookie deprecation earlier this year, AlgoriX promptly enrolled to join the Google Privacy Sandbox initiative.
In Q2 2024, we embarked on a sequence of internal exploratory tests to evaluate the potential influence of the Google Privacy Sandbox on our partners and to devise strategies to tackle any emerging challenges. With Google pushing back the Privacy Sandbox deployment to Q1 2025, we plan to further intensify our efforts to future-proof our products and offerings. This blog aims to offer a glimpse into our preliminary findings.
Our Privacy Sandbox Efforts
AlgoriX’s official enrollment into the Google Privacy Sandbox has granted us direct access to relevance and measurement APIs on Chrome and Android, such as Protected Audience, Topics, Private Aggregation, and Shared Storage. In line with this, we have launched the following efforts:
Launched Tests on In-App Envirornments
Most of AlgoriX’s partners are mobile developers and in-app advertisers. This made it crucial for us to design test applications for Android devices and conduct a series of simulations specifically for in-app environments on Google Privacy Sandbox. Our initial findings are outlined below.
- Protected Audience API
With the Protected Audience API, we recently conducted an on-device auction, involving a single seller and buyer. Using a simulated device with AlgoriX’s Privacy Sandbox test app and our SDK, we first played the role of a Buyer and linked the device to a tailored audience segment called “Shoes” with a candidate ad targeted to this audience. We then played the role of a Seller and initiated an ad selection process with the Buyer that is identified by the eTLD+1 domain of the bidding logic URL, who subsequently gave a higher bid amount for the custom audience named “Shoes”. This process culminated in a successful auction, yielding a winning ad.
-
Topics API
Moving on to the Topics API, we deployed AlgoriX’s test apps on a physical device used by our SDK team for testing, where AlgoriX’ SDK triggered the Topics API to record observations. After a shortened epoch setting from the default 7 days to 5 minutes, AlgoriX’s SDK triggered the Topics API from one of the test apps and retrieved the value of “/Games/Computer & Video Games/Simulation Games” from the topic classifier. This can be further used in auctions.
Feedback Sharing with Key Industry Players
The Privacy Sandbox initiative signifies a transformative leap in the digital advertising landscape. Hence, it requires active collaboration among industry players to address potential challenges and identify key opportunities. Currently, our AlgoriX team is in close dialogue with Google, sharing insights and jointly exploring solutions to hurdles encountered during testing. Simultaneously, as a member of the IAB Tech Lab, we have also shared our in-app specific feedback and recommendations to the IAB team. Recently, the IAB Tech Lab, which AlgoriX is a member of, released the final version of their Google Privacy Sandbox Fit Analysis.
Outreach to Advertisers and Publishers
Any privacy-related efforts will ultimately impact not only our business but also those of our partners’. Thus, we are proactively engaging with advertisers and mobile app publishers, conducting joint tests and exploring viable solutions. This is to guarantee seamless operations amid this ongoing transition.
Privacy Sandbox on In-App Environments
Because the threat of cookie deprecation is more imminent, most industry reports on Privacy Sandbox are conducted on web-based environments. However, GAID deprecation will follow sooner or later, so it’s essential to also look deeper into the impact of Privacy Sandbox on in-app environments. After our initial set of tests, primarily on the Topics and Protected Audience APIs, we have the following observations:
Unique Features of Privacy Sandbox on Android
- Custom Audience vs Interest Group: Unlike the web version that uses Chrome browsing history, the Android version creates Custom Audiences based on user behavior within apps. It does not communicate with the Interest Group on the web version.
- Custom Audience Delegation: The Android Privacy Sandbox allows Buyer ad platforms to delegate the task of adding Custom Audiences to specific apps or SDKs with whom they have a collaborative relationship. For example, when a user uses App X, where DSP A does not have an on-device presence, DSP A can delegate App X or MMP SDKs to App X to add a Custom Audience on their behalf. Currently, though there are discussions about implementing a similar mechanism on web-based environments, an official proposal is yet to be seen.
- Protected App Signals: Ad platforms can create and store signals from app events like installations, first-time openings, user actions, and purchases. These signals are written and stored on the device, and when the Protected Audience auction is initiated, they are encrypted and sent to the advertising platform that created these application event signals, which then help make informed ad choices.
- App Install Ads Filtering: The system prevents showing ads for apps already installed on a device, filtering out unnecessary ads during the selection process.
Points for Improvement & Optimization
- Fragmented technical documentation: While analyzing each use case with the IAB Tech Lab Privacy Sandbox task group, the AlgoriX team found that the API documentation is rather fragmented and lacks version control, increasing the difficulty for advertising platforms to fully understand and participate in testing. The documentation of some technical solutions (such as k-anonymity for advertising creatives) is only clearly defined in the web environment’s Protected Audience API, and there is no separate technical documentation for Android applications.
- Limited scalability, increased costs: The serial auction model under the Privacy Sandbox (traditional programmatic auction followed by Protected Audience API auction) has introduced significant traffic and network restrictions that did not exist before, increasing the demands for computing power, internet connectivity, and electricity, thus increasing the cost of supporting the Protected Audience API. Furthermore, the current Trusted Execution Environment (TEE) only supports Google Cloud and Amazon Cloud, lacking sufficient market fairness for advertising platforms that use other cloud services or on-premise data centers.
- Reduced bidding transparency: Compared to the traditional OpenRTB protocol, it is difficult for both SSP and DSP to directly record No Bid and Bid Loss data within the Protected Audience API. Although it is possible to obtain aggregated data with noise through the Private Aggregation API, the traditional bidding optimization algorithms and mechanisms used by the advertising platforms need to adapt to lower bidding transparency.
Engage with Us
AlgoriX is committed to further exploring Google’s Privacy Sandbox, particularly within in-app environments. We’re extending an invitation to industry partners to engage with us, offering your insights and feedback to collaboratively address the challenges at hand. As we delve deeper, we’re eager to share our initial findings and ongoing efforts.
