 Hi, I'm Danny Rojas, and I'm on the Product Partnerships team for Privacy Sandbox. I spent the last two plus years having conversations with folks in the ecosystem on how to prepare for a future without third-party cookies, leveraging technologies like in the Privacy Sandbox. I'm here to tell you what's going on with the three big ads APIs we've been working on for the Privacy Sandbox, Topics, Fledge, and the Attribution Reporting API. We generally expect that they will be used as building blocks to create advertising solutions, both in combination with each other, along with contextual and first-party data. But before I get into that, let me take a step back for a minute. What is an API anyways? There are viewers from a wide range of backgrounds, so let me take a minute to be clear about what I'm going to be discussing. These APIs are not products. These are new capabilities we're adding to the Chrome web browser, new additions to the JavaScript programming language. The web browser's job is to be a platform that people can build products on top of. Folks who will actually be building ad products are the same folks who have been building ad products today. Many of those ad products historically made use of third-party cookies. In a world without cookies, these ad products will need to rely on new tools, new APIs instead. If you want to hear how that's going, you can read case studies by RTB House in Criteo. But what if you're a publisher and an advertiser? What's your call to action if these APIs may be being used by advertising technology companies? Talk to the ad tech companies you do work with and ask them if they're involved in privacy sandbox testing or if they have plans to be later this year. Until now, it's been reasonable to wait. Things have been changing a lot and keeping up with current developments is not an easy job. Now the time has come to figure out how to build with these tools. Speaking of that, let me give a quick view of timelines, where we've come and where we're headed. 2022 has been about early technical integration. Launching the APIs for technical testing and having the ad tech companies effectively kick the tires on the ads API proposals. For the first half of 2023, it'll be very much test and learn. We expect the testing to be focused on individual APIs to give directional utility learnings and will be increasing traffic exposed to these APIs to a sufficient level at a responsible pace. And we're looking forward to feedback in this timeframe to help understand how specific parameters of the API proposals influence the results. In the second half of 2023, we'll launch the APIs to all Chrome users, which will enable testing of the sandbox technologies at scale in combination and across the ecosystem. At this point, ad tech companies will have integrated these APIs into their solutions alongside the rest of privacy preserving capabilities that are available. With this, we'll gain meaningful insights into the results for marketers and publishers in a world without third party cookies. Looking further ahead, 2024 will be about preparation for a third party cookie deprecation. After launching the sandbox APIs later this year, there will be a full year to integrate the solutions and optimize the results. And of course, you can always go to privacysandbox.com for the latest on timelines. Now the first API, Topics. In a world without any cross site information at all, ad targeting choices could be based on contextual information about the surrounding webpage, media user information like their GEO or on same site user information. Topics adds a little bit of cross site user information into the mix. It provides a high level signal of a user's interest based on their recent browsing history. And what's its current status? Right now we're testing with live traffic and you can monitor the blink dev email alias for the current traffic levels. And at the most rudimentary level, data is flowing through the pipes. Cell side platforms are able to call the topics API and pass the resulting topics along to demand side platforms. And we've already heard some feedback about the mechanics of how the API works. For example, on pages with multiple third parties that all want to use the API. The current design involves some browser gymnastics involving iframes and post messages and we've heard it's cumbersome and sometimes slow. We're facing this by introducing a mechanism based on HTTP headers instead. Now, early in the first half of 2023, we're going to hear more on the directional utility learnings. And we actually started hearing this towards the end of 2022 from the earliest testers. For example, folks are evaluating the on device classifier, which Chrome uses to know the topics of the pages that a person visits in the browser. We're happy to hear that the classifier is pretty good. And of course there are suggestions on how to make it better, particularly by using information beyond just the domain of the site being browsed. And we're interested in further discussions from the taxonomy that the topics are drawn from. We also heard that the current way of picking the top topics for a person based on the number of pages about it means that the most popular topics like news or arts and entertainment are too prevalent and less useful as a result. We're evaluating better ways to surface topics so that we're likely to give information of a commercial interest. Oh, and I should say that I'm just mentioning a few highlights of feedback. And if you want to see more, our quarterly report to the UK CMA contains summaries of the feedback we're hearing on all of the privacy sandbox APIs. Now, in the later half of 2023, this is when we expect conversations about the powerful ways that information from the topics API can be most effectively used by the advertising ecosystem. For example, topics should be a valuable input to predictive ML models used in bidding alongside contextual and first party signals. Topics may also be useful alongside contextual signals for audience building use cases. That is in combination with the Fledge API, which we'll turn to next. Okay, on to Fledge. This is the thing that gets described as the privacy sandbox API for remarketing, but really Fledge is so much more than just the remarketing use case. Remarketing is an ads ecosystem product, and Fledge is a tool, and the tool is made for building custom audiences. And let me be clear up front, using Fledge is a much more difficult task than using topics. Inventing a way for the ecosystem to show ads to custom built audiences without being able to track the individual people in those audiences, that was a huge privacy challenge. Now, one big benefit that topics brings to the table is how easy it is to use. Nobody will say that about the Fledge API. And I'm telling you all this just so that you understand that when we say the current state is functionality testing, it really means something. I mean, the mere fact that it is possible to run an on-device auction with custom audiences from multiple buyers bidding against each other, it's a big deal. Our biggest feedback from this stage is a lot of discussion about latency, about the amount of time and resources that these auctions use. And we've already shipped a lot of control so that the party running these on-device ad auctions, the publisher or the self-side ad text that they choose to do the job for them, can limit that resource usage. And we're involved in a lot of discussions on how to get the most out of the resources that are available. And Chrome also announced support for the bidding and auction services, which provides optional tooling to potentially improve latency. Now, in the first half of this year, as people build out and test their implementations, we expect a lot of discussion of reporting. For example, the ways that developers who are used to performing bidding on their own servers, how they can get insight into what's happening when things are running in browsers instead. And by the second half of this year, as Fledge becomes available to all users, we look forward to much more attention on how the different parties on a single page interact with each other. This includes the dynamics of multiple buyers in a single auction and also the component auctions, the design of Fledge that allows publishers to have multiple SSPs competing against each other. Oh, and I do want to point out that we've been in some discussions about interesting ideas like publishers individually or in coalitions creating and selling their own audiences while retaining control over how those audiences are used. This is indeed something that Fledge could be used for, but this is an idea for an ad's product innovation. The platform changes we're making mean that it will become possible to let someone advertise to an audience without being able to take that audience and run away with it. But building a new business around this will be in the hands of interested ad tech companies, not browsers. Finally, I'll say just a few quick words about the attribution reporting API. This is, of course, built to serve an essential part of the ad's ecosystem. The ability for advertisers to tell that the advertising dollars that they spent were worth it. This API is multifaceted covering both event level and aggregated reporting. The early functional testing has been essential for understanding debugging use cases. Across the privacy sandbox work, but for attribution reporting in particular, for both in detailed error reporting and understanding of failures has been a key piece of feedback we've received and acted on. The early utility testing has highlighted some valuable use cases that were not covered in our first releases. For example, we heard from advertisers with multiple separate websites where ad clicks lead to landing pages that are on a different domain from the one where an eventual conversion happens. As we get more testers next year, we expect to hear about additional use cases like these that don't work like the textbook example. As we move to the integrated utility testing phase later this year, we also have plans to support attribution across web and app contexts, as part of integrating the Chrome and Android sides with the privacy sandbox. And while we have the first stage of integration that allows the use of attribution reporting from within Fledge, I'm sure we'll get refined as people begin testing the APIs together. So that's where the privacy sandbox ads APIs are today and where they're heading through the rest of 2023. As I hope you can see there's a lot going on here. None of this is going to happen quickly and we understand that. We've been through big transformations before. When the web transition from HTTP to HTTPS so that users could more safely browse websites, this took a while to get the industry on board. Many ad serving stacks didn't support HTTPS and there were concerns about perf issues and losing SEO. Chrome team took this feedback seriously and worked together with the industry and with the rest of Google to address concerns so that we can have a more secure web. Similarly, though at a much bigger scale, Privacy Sandbox is developing new technologies to ensure a more private web and app ecosystem. Change and adaptation are going to be necessary, but we can do this together. So what I want you to do is this. First, talk to your advertising partners. Ask them about their plans for durable privacy forward solutions. We hope that the tools that we're providing in Privacy Sandbox will be valuable contributions to those plans. And second, you willing to experiment? When those advertising partners come to you and say that the future is going to be a little bit different than the past, be open to it. We're going to accomplish the same goal, but maybe not in all the same ways and we're all going to need to be flexible to get there. Thank you.