
New ways to drive hotel bookings and reach travelers

Ad technology providers widely use machine learning (ML) models to predict and present users with the most relevant ads, and to measure the effectiveness of those ads. With increasing focus on online privacy, there’s an opportunity to identify ML algorithms that have better privacy-utility trade-offs. Differential privacy (DP) has emerged as a popular framework for developing ML algorithms responsibly with provable privacy guarantees. It has been extensively studied in the privacy literature, deployed in industrial applications and employed by the U.S. Census. Intuitively, the DP framework enables ML models to learn population-wide properties, while protecting user-level information.
When training ML models, algorithms take a dataset as their input and produce a trained model as their output. Stochastic gradient descent (SGD) is a commonly used non-private training algorithm that computes the average gradient from a random subset of examples (called a mini-batch), and uses it to indicate the direction towards which the model should move to fit that mini-batch. The most widely used DP training algorithm in deep learning is an extension of SGD called DP stochastic gradient descent (DP-SGD).
DP-SGD includes two additional steps: 1) before averaging, the gradient of each example is norm-clipped if the L2 norm of the gradient exceeds a predefined threshold; and 2) Gaussian noise is added to the average gradient before updating the model. DP-SGD can be adapted to any existing deep learning pipeline with minimal changes by replacing the optimizer, such as SGD or Adam, with their DP variants. However, applying DP-SGD in practice could lead to a significant loss of model utility (i.e., accuracy) with large computational overheads. As a result, various research attempts to apply DP-SGD training on more practical, large-scale deep learning problems. Recent studies have also shown promising DP training results on computer vision and natural language processing problems.
In “Private Ad Modeling with DP-SGD”, we present a systematic study of DP-SGD training on ads modeling problems, which pose unique challenges compared to vision and language tasks. Ads datasets often have a high imbalance between data classes, and consist of categorical features with large numbers of unique values, leading to models that have large embedding layers and highly sparse gradient updates. With this study, we demonstrate that DP-SGD allows ad prediction models to be trained privately with a much smaller utility gap than previously expected, even in the high privacy regime. Moreover, we demonstrate that with proper implementation, the computation and memory overhead of DP-SGD training can be significantly reduced.
We evaluate private training using three ads prediction tasks: (1) predicting the click-through rate (pCTR) for an ad, (2) predicting the conversion rate (pCVR) for an ad after a click, and 3) predicting the expected number of conversions (pConvs) after an ad click. For pCTR, we use the Criteo dataset, which is a widely used public benchmark for pCTR models. We evaluate pCVR and pConvs using internal Google datasets. pCTR and pCVR are binary classification problems trained with the binary cross entropy loss and we report the test AUC loss (i.e., 1 - AUC). pConvs is a regression problem trained with Poisson log loss (PLL) and we report the test PLL.
For each task, we evaluate the privacy-utility trade-off of DP-SGD by the relative increase in the loss of privately trained models under various privacy budgets (i.e., privacy loss). The privacy budget is characterized by a scalar ε, where a lower ε indicates higher privacy. To measure the utility gap between private and non-private training, we compute the relative increase in loss compared to the non-private model (equivalent to ε = ∞). Our main observation is that on all three common ad prediction tasks, the relative loss increase could be made much smaller than previously expected, even for very high privacy (e.g., ε <= 1) regimes.
![]() |
DP-SGD results on three ads prediction tasks. The relative increase in loss is computed against the non-private baseline (i.e., ε = ∞) model of each task. |
Privacy accounting estimates the privacy budget (ε) for a DP-SGD trained model, given the Gaussian noise multiplier and other training hyperparameters. Rényi Differential Privacy (RDP) accounting has been the most widely used approach in DP-SGD since the original paper. We explore the latest advances in accounting methods to provide tighter estimates. Specifically, we use connect-the-dots for accounting based on the privacy loss distribution (PLD). The following figure compares this improved accounting with the classical RDP accounting and demonstrates that PLD accounting improves the AUC on the pCTR dataset for all privacy budgets (ε).
![]() |
Batch size is a hyperparameter that affects different aspects of DP-SGD training. For instance, increasing the batch size could reduce the amount of noise added during training under the same privacy guarantee, which reduces the training variance. The batch size also affects the privacy guarantee via other parameters, such as the subsampling probability and training steps. There is no simple formula to quantify the impact of batch sizes. However, the relationship between batch size and the noise scale is quantified using privacy accounting, which calculates the required noise scale (measured in terms of the standard deviation) under a given privacy budget (ε) when using a particular batch size. The figure below plots such relations in two different scenarios. The first scenario uses fixed epochs, where we fix the number of passes over the training dataset. In this case, the number of training steps is reduced as the batch size increases, which could result in undertraining the model. The second, more straightforward scenario uses fixed training steps (fixed steps).
In addition to allowing a smaller noise scale, larger batch sizes also allow us to use a larger threshold of norm clipping each per-example gradient as required by DP-SGD. Since the norm clipping step introduces biases in the average gradient estimation, this relaxation mitigates such biases. The table below compares the results on the Criteo dataset for pCTR with a standard batch size (1,024 examples) and a large batch size (16,384 examples), combined with large clipping and increased training epochs. We observe that large batch training significantly improves the model utility. Note that large clipping is only possible with large batch sizes. Large batch training was also found to be essential for DP-SGD training in Language and Computer Vision domains.
The per-example gradient norm calculation used for DP-SGD often causes computational and memory overhead. This calculation removes the efficiency of standard backpropagation on accelerators (like GPUs) that compute the average gradient for a batch without materializing each per-example gradient. However, for certain neural network layer types, an efficient gradient norm computation algorithm allows the per-example gradient norm to be computed without the need to materialize the per-example gradient vector. We also note that this algorithm can efficiently handle neural network models that rely on embedding layers and fully connected layers for solving ads prediction problems. Combining the two observations, we use this algorithm to implement a fast version of the DP-SGD algorithm. We show that Fast-DP-SGD on pCTR can handle a similar number of training examples and the same maximum batch size on a single GPU core as a non-private baseline.
![]() |
The computation efficiency of our fast implementation (Fast-DP-SGD) on pCTR. |
Compared to the non-private baseline, the training throughput is similar, except with very small batch sizes. We also compare it with an implementation utilizing the JAX Just-in-Time (JIT) compilation, which is already much faster than vanilla DP-SGD implementations. Our implementation is not only faster, but it is also more memory efficient. The JIT-based implementation cannot handle batch sizes larger than 64, while our implementation can handle batch sizes up to 500,000. Memory efficiency is important for enabling large-batch training, which was shown above to be important for improving utility.
We have shown that it is possible to train private ads prediction models using DP-SGD that have a small utility gap compared to non-private baselines, with minimum overhead for both computation and memory consumption. We believe there is room for even further reduction of the utility gap through techniques such as pre-training. Please see the paper for full details of the experiments.
This work was carried out in collaboration with Carson Denison, Badih Ghazi, Pritish Kamath, Ravi Kumar, Pasin Manurangsi, Amer Sinha, and Avinash Varadarajan. We thank Silvano Bonacina and Samuel Ieong for many useful discussions.
Posted by Anthony Chavez, VP Product ManagementIn February we announced the Privacy Sandbox on Android, with the goal of bringing new, more private advertising solutions to mobile.
Over the course of 2022, we've published design proposals and released a number of Developer Previews. We appreciate all of the feedback we've received which has helped us refine and improve these proposals.
Beginning early next year we plan to rollout the initial Privacy Sandbox Beta to Android 13 mobile devices, so that developers can take the next steps in testing these new solutions. We'll start with a small percentage of devices and increase over time. Note that Developer Previews will continue to be released and this is where we’ll first deliver the latest features for early feedback before being released on production devices.
Today, we're sharing more details about the Privacy Sandbox Beta so that developers can get prepared.
The Privacy Sandbox Beta will be available for ad tech and app developers who wish to test the ads-related APIs as part of their solutions.
During the initial rollout stages, enrolled developers will also need to join the early testers program. This program will allow developers to test the APIs on a limited number of their own Android 13 devices for internal apps and requested published apps.
For the SDK Runtime, we’ll have a closed beta for developers to test Runtime-enabled SDK distribution to select apps. Because of the coordination required to test the SDK Runtime on production devices, we expect this beta to involve a limited number of partners who can dedicate resources to support this testing. If you’re interested in participating, please register your interest.
To utilize the Beta release, developers will need to compile their solutions with an API level 33 SDK extension update that is coming soon.
We’ve heard from many advertisers and publishers about the role they can play in testing these new technologies. For companies that rely on third party solutions for ad serving or ad measurement, we recommend working with your providers to understand their testing roadmaps and how you can participate in early testing of Privacy Sandbox.
We want to thank everyone who has engaged on the Android Privacy Sandbox, and look forward to continued feedback as we enter this next phase of testing."
Posted by Anthony Chavez, VP Product ManagementIn February we announced the Privacy Sandbox on Android, with the goal of bringing new, more private advertising solutions to mobile.
Over the course of 2022, we've published design proposals and released a number of Developer Previews. We appreciate all of the feedback we've received which has helped us refine and improve these proposals.
Beginning early next year we plan to rollout the initial Privacy Sandbox Beta to Android 13 mobile devices, so that developers can take the next steps in testing these new solutions. We'll start with a small percentage of devices and increase over time. Note that Developer Previews will continue to be released and this is where we’ll first deliver the latest features for early feedback before being released on production devices.
Today, we're sharing more details about the Privacy Sandbox Beta so that developers can get prepared.
The Privacy Sandbox Beta will be available for ad tech and app developers who wish to test the ads-related APIs as part of their solutions.
During the initial rollout stages, enrolled developers will also need to join the early testers program. This program will allow developers to test the APIs on a limited number of their own Android 13 devices for internal apps and requested published apps.
For the SDK Runtime, we’ll have a closed beta for developers to test Runtime-enabled SDK distribution to select apps. Because of the coordination required to test the SDK Runtime on production devices, we expect this beta to involve a limited number of partners who can dedicate resources to support this testing. If you’re interested in participating, please register your interest.
To utilize the Beta release, developers will need to compile their solutions with an API level 33 SDK extension update that is coming soon.
We’ve heard from many advertisers and publishers about the role they can play in testing these new technologies. For companies that rely on third party solutions for ad serving or ad measurement, we recommend working with your providers to understand their testing roadmaps and how you can participate in early testing of Privacy Sandbox.
We want to thank everyone who has engaged on the Android Privacy Sandbox, and look forward to continued feedback as we enter this next phase of testing."
Posted by Anthony Chavez, VP Product ManagementIn February we announced the Privacy Sandbox on Android, with the goal of bringing new, more private advertising solutions to mobile.
Over the course of 2022, we've published design proposals and released a number of Developer Previews. We appreciate all of the feedback we've received which has helped us refine and improve these proposals.
Beginning early next year we plan to rollout the initial Privacy Sandbox Beta to Android 13 mobile devices, so that developers can take the next steps in testing these new solutions. We'll start with a small percentage of devices and increase over time. Note that Developer Previews will continue to be released and this is where we’ll first deliver the latest features for early feedback before being released on production devices.
Today, we're sharing more details about the Privacy Sandbox Beta so that developers can get prepared.
The Privacy Sandbox Beta will be available for ad tech and app developers who wish to test the ads-related APIs as part of their solutions.
During the initial rollout stages, enrolled developers will also need to join the early testers program. This program will allow developers to test the APIs on a limited number of their own Android 13 devices for internal apps and requested published apps.
For the SDK Runtime, we’ll have a closed beta for developers to test Runtime-enabled SDK distribution to select apps. Because of the coordination required to test the SDK Runtime on production devices, we expect this beta to involve a limited number of partners who can dedicate resources to support this testing. If you’re interested in participating, please register your interest.
To utilize the Beta release, developers will need to compile their solutions with an API level 33 SDK extension update that is coming soon.
We’ve heard from many advertisers and publishers about the role they can play in testing these new technologies. For companies that rely on third party solutions for ad serving or ad measurement, we recommend working with your providers to understand their testing roadmaps and how you can participate in early testing of Privacy Sandbox.
We want to thank everyone who has engaged on the Android Privacy Sandbox, and look forward to continued feedback as we enter this next phase of testing."
The year is 2012. You and a record-setting 8 million people are watching live as skydiver Felix Baumgartner breaks the speed of sound with his 24-mile fall to Earth. In another tab, PSY’s “Gangnam Style” plays for the umpteenth time. Later in the year, it will become the first YouTube video to hit 1 billion views. You’re doing all of this, of course, on your desktop, a device that still accounts for more than 90% of web traffic globally.
That same year, we introduced Think with Google to give readers insights into behavioral and cultural trends based on Google data.
So much has changed in the decade since. For one, there’s a much higher chance you’re reading this article on mobile, which now makes up almost 60% of global web traffic. In that time, Think with Google has charted the evolution of consumer behavior as new digital technologies have emerged.
To mark the 10-year anniversary of Think with Google, here’s a selection of insights that tell that story.
The opening ceremony for the 2012 Olympics drew a global TV audience of 900 million. By the time the Games were over, NBC declared it the most-watched television event in U.S. history.
But an important change was underway: As broadband access increased, people were ditching TV programs and heading online. In a 2012 Think with Google study, 44% of 13-to 24-year-olds said they spent more time watching online video than TV.
Source: Ipsos/Google, 2012 Teens and Twenty-Somethings research study, 2012.
In 2010, when sports fans tuned into the World Cup, they were so glued to their TV screens that online searches, which tended to take place on desktop, plummeted.
Ahead of the 2014 World Cup, we drew on data from the Union of European Football Associations Champions League to make a prediction: “Second screening,” where fans tune into an event on a big screen while following along on mobile, was going mainstream. That prediction held up. In 2018, research revealed that 70% of adults looked at a second screen while watching TV.
By 2015, long gone were the days where shoppers had to write a list before heading to the store. Instead, people were turning to their smartphones the second they wanted to do something, discover something or buy something.
We called these “micro-moments,” and they were becoming increasingly widespread. That year, 82% of smartphone users consulted their phone while in a store.
In 2017, it took, on average, 22 seconds for a mobile webpage to load.
That might not sound like much, but according to Google research that year, it was enough to sink an online business. The research, which analyzed the landing pages of 900,000 mobile ads, found that as page load time went from one second to 10 seconds, the probability of a mobile visitor bouncing increased 123%.
We’ve all seen people with their heads buried in their phones, unaware of what’s happening around them. Maybe that’s why the internet has long had an unfair reputation as an isolating experience.
But YouTube data from 2018 revealed an emerging trend that would explode just a few years later: the rise of video as an interactive, social experience. For example, we saw a huge interest in videos that encouraged viewers to do something — study, clean, read — at the same time as a YouTube creator and their followers.
No amount of Search data could have predicted what would happen in 2020, as the coronavirus pandemic shuttered offices, stores and schools.
Many people used the stay-at-home mandates to develop new skills. YouTube data revealed that globally, videos with variations of “beginner” in the title earned more than 7 billion views.
If anyone thought the changes brought on by the pandemic were temporary, Search data from 2021 suggested otherwise.
While people were understandably eager to go back to IRL experiences, virtual social events never fully went away. Around the world in 2021, we saw a 90% year-on-year increase in searches containing the term “watch party.”
Now, here we are in 2022. We’re continuing to crunch the numbers and share trends. We’re shedding light on evolving consumer behavior. And we’re sharing the lessons Google is learning, in real time, on topics ranging from inclusion to privacy. In an industry that has changed so much over the past decade, and will continue to do so, insights like these are crucial — and we’ll be here to keep sharing them.
‘Creativity’, ‘diversity’ and ‘the economy’ are becoming increasingly associated with one another. The findings from two landmark studies, launching today at YouTube Festival, demonstrate that they are strongly symbiotic.
We have always known that creators are at the core of YouTube’s success and the latestYouTube Impact report by Oxford Economics on the state of the creator economy, highlights the wider, significant economic impact of YouTube’s creative ecosystem in the UK. A second piece of research, Mirrors and Windows, conducted with MTM, explores identity and media choice in the UK and the role and responsibility of brands in improving authentic representation.
This year I celebrated my 15th year at Google and there are some key themes that have progressed significantly during my time here. The main is the growth in sheer reach – new audiences continue to come to us with 96% of online adults in the UK visiting YouTube at least once a month. As a platform that attracts nearly the whole of the UK to our creative content, it should be no surprise that Oxford Economics found that YouTube’s creative ecosystem contributed over £1.4billion to the UK’s GDP, with 80% of creative entrepreneurs agreeing that YouTube provides an opportunity to create content and earn money that they wouldn’t get from traditional media.
I hear from creators every day about how our open platform has lowered the barrier to entry for creatives from every corner of the UK, and of every background and demographic. Munya Chawawa, who hosted YouTube Festival, is testament to how YouTube has the power to launch life-changing careers and ensure audiences can find content that reflects the diversity of the UK today. The research reinforced these anecdotes, with 79% of Mirrors & Windows respondents saying that they believe YouTube represents a wide range of voices, people and perspectives, a sentiment that was shared by those who identify as LGBTQ+ and those identifying as Asian, Black, Mixed or multiple, or other ethnic groups. Across the UK, from Scotland to Yorkshire to the South West, users also agree that YouTube has the content that people in the UK love.
Munya Chawawa at the MOBO Awards, 2021
While we’ve gained a better understanding of the tremendous benefits that YouTube’s creative ecosystem provides to our economy, society, brands and culture, the agencies and advertisers MTM spoke to indicated that they sometimes still have apprehensions about their ability to speak about the issue of diverse representation without misstepping and facing scrutiny. Saying that, as YouTube continues to introduce new ways to support diverse creatives, it is becoming a powerful vehicle for brands on the road to more inclusive campaigns.
When fashion and beauty destination ASOS, were planning to tell young people that their brand is for people of all sizes, races and celebrates self expression, YouTube was there to connect them with a unique opportunity to amplify this message. ASOS’ partnership with MOBO – a global brand dedicated to celebrating and championing Black culture and music – was a good example of how authentic and inclusive marketing equals reach and brand results, with ASOS’ pre-roll ads reaching almost three million people watching the Awards and associated music content on YouTube alone. The partnership, which also included ASOS sponsoring the Awards’ Best Newcomer category, not only demonstrated YouTube’s ability as a driver of cultural moments, but also its ability to bring brands closer to diverse audiences, deliver strong brand metrics, and support the creative development and distribution of representative content.
While we’ve known this for a while, the insights released today demonstrate that there’s an urgent need for cultural fluency – where brands show not just a familiarity with different cultures but an ability to communicate effectively in different contexts and be able to change and adapt as much as our platform and its audiences are.
I am so excited to continue to work with brands to help them thrive on this open and constantly evolving platform, where engaging with creative talent, ideas and audiences means direct investment in the UK economy, helping to maintain the UK’s position as a global leader in the creative industry.
People come to Google to find information from a wide range of sources and formats, from big brands to individual creators, across text, images and video. As we’ve introduced features and design elements to help you explore information in new ways, we’ve also continued to bring greater context to the search results page, helping you make sense of the information you see. Today, we’re making a few new updates to the search page that build on this work, providing even more information about the sites that you see so you can feel confident about the websites you visit.
We’re adding site names to search results on mobile, so you can easily identify the website that’s associated with each result at a glance. We’re also updating the size and shape of the favicon (a website’s logo or icon) that appears in Search, to make it easier to see on the page. We’ll extend these changes to Search ads to increase clarity and advertiser transparency at a glance.
Part of helping you make sense of the information you see is ensuring that ads are clearly labeled, which is why our label will now be featured on its own line in the top-left corner of Search ads. We also want the label to be prominent and clear across different types of paid content. That’s why when ads show on mobile search results, they will now be labeled with the word “Sponsored” in bold black text. This new label and its prominent position continues to meet our high standards for being distinguishable from search results and builds on our existing efforts to make information about paid content clear.
This search page update is starting to gradually roll out on mobile and we’ll soon begin testing a similar experience on desktop, helping people more easily find what they’re looking for, no matter where they’re searching.
The following is adapted from a speech given by Matt Brittin, President, Google EMEA, at DMEXCOin Cologne.
Across the world, we’re seeing increased uncertainty. We’re living through a pandemic, seeing rising prices, a global energy crisis, increasing climate disasters and a horrific war in Ukraine. Access to quality information has never been more important — to help people search for answers, find ways to save money, make more sustainable choices and stay safe and informed.
But the web as we know it is at risk. People are more concerned than ever about their privacy online. Regulators across the world are demanding a more private internet — with some critics calling for a ban on personalised ads completely.
The future of the web depends on earning people’s trust — building responsible, private advertising to secure a sustainable internet that is safer for people, stronger for businesses and successful for publishers.
For generations, ads have funded our favourite content: from newspapers, magazines and entertainment to the web. Today 66% of the world is online. The ad-supported internet model has become a remarkable resource for humanity: putting an explosion of tools, information and content at our fingertips.
But nearly 40 years after its creation, the internet needs a grown-up attitude to responsibility.
As people manage more of their lives online, their concerns over how personal data is gathered, used and shared have increased. People want great online experiences — delivered with the privacy they deserve, by brands they can trust.
For advertisers, that presents a clear responsibility - but also an opportunity. And the good news is this: privacy safe ads are effective ads.
This year, we asked 20,000 Europeans about the consequences of good and bad privacy experiences. Our findings show that users view bad privacy experiences as almost as damaging as a theft of their data. It’s enough to make many of them switch to another brand entirely. And, because the impact of a negative privacy experience outweighs that of a positive one, it’s very difficult to recover from.
Instead, brands need to get it right the first time. People prefer to buy from brands that give them more control over their privacy — almost three quarters said they would prefer to buy from brands that are honest about what data they collect and why.
In times of uncertainty, companies may be tempted to put privacy on the backburner - but that would be a mistake. In tough times you need to invest for the future. Privacy is that investment.
Making these changes won’t just lead to successful advertising — but a sustainable web.
Digital advertising needs to be safer for people. They need to feel protected online and able to trust what they view. It needs to be successful for publishers — funding quality journalism while giving us access to authoritative and diverse perspectives. And, it needs to be stronger for business — allowing businesses of all sizes the opportunity to grow and build a global customer base.
Across Europe, we’re investing in that vision. We’re one of the world’s biggest financial supporters of journalism, committing billions of dollars every year; we’re delivering authoritative information and creating privacy-first technology.
At our Google Safety Engineering Center in Munich, hundreds of engineers are creating tools and technology that combine two German traditions: exemplary engineering and rigorous privacy standards.
The privacy-first technology they’re creating is minimising the amount of data used, simplifying data downloads and deletion, and helping root out hijacked passwords — building on our shared values and breaking new ground in the global industry.
Today, as part of our commitment to that transition, we’re announcing two new tools.
The first is the Google Ads Privacy Hub, launching today with the rollout starting here in Germany. It will show you the latest on product innovations and how best-in-class marketers are doing it — helping you take the first steps on this journey, whatever your company size.
The second tool we’re launching focuses on users. Last year, 300 million people visited Ad Settings — choosing to make the ads they see more specific to them. So we’ll soon launch the new My Ad Center globally — expanding our existing Ad Settings to give people a single place where they can control the ads they see across Google Search, Discover and YouTube — seeing more of what they like, and less of what they don’t. Because the best ads are helpful, relevant and safe — benefiting the user, and responsible businesses too.
Matt Brittin speaking at DMEXCO conference
Moving to a world without third-party cookies means rethinking the tech on which much of the web advertising system is built and building new, privacy-first solutions.
We’re doing that through the Privacy Sandbox: sharing and testing new technologies with the industry, while staying on course to deprecate third party cookies by the end of 2024, in line with our commitments to the UK Competition Authority, which we are applying globally.
There are those that say that efforts like the Privacy Sandbox aren’t enough. Some say that we should ban personalised advertising altogether — that “contextual” advertising can fill the gap. But that won’t pay for the web everyone wants.
It has been estimated that if personalised advertising were to suddenly go away, as much as $32 to $39 billion would shift away from those who rely on open web technology — including publishers, at a time where authoritative information has never been more important.
There are others that say that all services should simply be paid for. But that turns the web into a luxury good — shutting billions out. It’s why Netflix, a pioneer of the modern subscription model, and others like Disney Plus and HBO, are introducing ads for users who want — or need — to pay less. Now, the advertising industry is a big tent. There is plenty of room for newcomers.
But recent events underscore how flawed these arguments really are - and unpopular to boot. Research by IAB Europe shows that 75% of Europeans would choose today’s experience of the internet over one without targeted ads, where they would need to pay for access to websites, content and apps.
So there should be no question as to whether the ads-supported internet model is important: only what kind of advertising industry we want to see. We want an advertising industry that makes room for businesses large and small; that supports value for publishers, media and journalism, and that protects people’s privacy from tracking.
But it’s not just enough to want that future. We have to actually choose it. For online advertising, and the future of the internet, this is a now or never moment: without people’s trust, the future of the ad-supported web is at stake.
The next two years are critical. The industry must embrace the journey and invest in privacy. It has to build stronger relationships with customers, create better ad campaigns and navigate the uncertainty. If we do nothing, the web as we know it will be under threat.
Together, we can build an ad-supported web fit for the future — giving us better content, richer perspectives and further protection online.