Learnings from Privacy Sandbox Testing

We’re grateful to see the engagement and feedback from companies across the industry who have deployed and tested products based on Privacy Sandbox and other privacy-preserving approaches. At the end of June, these efforts reached a significant milestone with over a dozen companies submitting results to the UK’s Competition and Markets Authority to help inform their assessment of the Privacy Sandbox.

In this post, we’ll share some key learnings and insights from the testing process.

First and foremost, testing has shown that solutions integrating the Privacy Sandbox can serve and measure digital ads on the web without third-party cookies. Testing has also indicated that ecosystem participation is a critical factor in understanding the performance of these solutions. Because testing at low volumes of third-party cookieless traffic cannot produce the marketplace dynamics and incentives that will exist next year, today’s results reflect what’s possible at 1% in the first half of 2024 — not what to expect in 2025.

A critical learning is the need for breadth and depth of adoption to support a thriving ecosystem. This includes the quantity and diversity of participants integrating with Privacy Sandbox, as well as the extent of feature support and optimization applied to the APIs and other building blocks. In particular, we see four key areas of opportunity that can help unlock the full economic potential of Privacy Sandbox-enabled solutions.

1. Increased DSP and SSP integrations to improve publisher results
2. More categories of solutions providers to unlock advertiser and agency demand
3. Expanded feature support by ad tech to enable more volume
4. Scaled model training to optimize performance outcomes

Increased DSP and SSP integrations to improve publisher results

Demand-side (DSP) and supply-side (SSP) platforms provide the core infrastructure of programmatic digital advertising on the web. Many of these platforms have integrated with Privacy Sandbox, including the Protected Audience API, to support running ad auctions in a privacy-preserving manner.

Yet there’s still a significant opportunity for increased adoption with DSP and SSPs who haven’t yet adopted the Sandbox APIs. And even amongst current adopters, Privacy Sandbox integrations between DSPs and SSPs are a fraction of what they are in today’s programmatic environment. For example, at the time of H1 2024 testing, an SSP may have only integrated with a few DSPs, when normally they support many dozens. This limited number of integrations means less buyer demand and auction pressure, resulting in lower publisher monetization than what would be expected as integrations increase.

More categories of solutions providers to unlock advertiser and agency demand

DSPs and SSPs aren’t the only tools used by participants across the ecosystem. We’ve heard feedback from advertisers and agencies that they’ll be ready to run campaigns through Protected Audience once it’s supported by their key vendors. This includes measurement, verification, data management, and audience companies whose solutions play a critical role in enabling digital ads to be transacted on the web. We’re working with many of these companies and as more of them add support for Privacy Sandbox, more advertiser demand will be unlocked — which in turn benefits publisher monetization.

Expanded feature support by ad tech to enable more volume

During the testing process, we’ve seen the importance of robust feature support for ad tech solutions built on Privacy Sandbox APIs.

For example, in early testing of Protected Audience we heard feedback that support for viewability — a key metric for advertisers, agencies, and publishers — was missing, which prevented buyers from accurately assessing the value of their campaigns. As vendors have added support for viewability, testers have seen this metric trending to expected ranges, supporting their ability to buy through Protected Audience.

And other key capabilities in today’s ad tech products — such as video ads and deals — are still being incorporated into solutions built on the Sandbox APIs. As these solutions are launched, it will enable advertisers to reach more of their desired audiences, and publishers to monetize more of their inventory, when third-party cookies aren’t available.

There’s also opportunity for ad tech vendors to further optimize their solutions to operate without third-party cookies. For instance, today it’s common for buy-side solutions to only bid when a third-party cookie is present. As new technologies like Privacy Sandbox enable effective monetization of cookie-less traffic, these solutions will need to be adapted to fully unlock this opportunity.

Scaled model training to optimize performance outcomes

Today’s advertiser and publisher solutions rely heavily on machine learning models to optimize performance, whether that’s for bidding or publisher yield management. These models are powered by historical data that can be used to predict future events. As ad tech companies adopt new technologies like Privacy Sandbox, these predictive models need to be retrained on data from operating without third-party cookies. As ad tech participation and cookieless traffic expands in the future, we can expect the performance of these models to improve as data sets expand, leading to better advertiser and publisher outcomes.

Path forward

Working together, we can unlock the opportunity from wider and deeper adoption of these new technologies across the ecosystem. And by doing so we can deliver improved privacy for users and the best possible results for advertisers and publishers.