Navigating The Future Of Addressability: From Cookies To Chrome Sandbox Innovations 

 In the wake of a heavily cookie-reduced world, advertisers face both challenges and opportunities in maintaining effective addressability. Here we delve into the current state of Privacy Sandbox APIs, exploring their potential for prospecting and retargeting while highlighting the need for the development of advanced, multi-signal approaches. 

Addressability & the Sandbox 

Effective advertising connects the right message to the right audience; a task that has always been challenging. Take a sports brand, for instance: simply targeting ‘sports enthusiasts’ is not enough to drive significant results. True success comes from advanced addressability: using complex combinations of signals and targeting strategies to accurately match brands with audiences. For example, with the same sports brand, we could increase spending in urban areas and target people interested in related topics such as health and competition, alongside existing customers, and those we predict as future customers. This is the standard of relevance that advertisers expect, and which we must assess cookieless alternatives against, including those that use Privacy Sandbox APIs. Basic targeting methods are no longer adequate in today’s digital landscape. 

Google’s recent decision to preference user choice over cookie deprecation does not change the need for urgency in transition and mitigation planning. There is one thing we can be sure about:  prompting user choice will lead to a diminished cookie pool. This is not new news, and things are moving at pace. Below, we expand on how Chrome’s Sandbox fits into the new landscape, and where there are gaps in its functionality.   

But before we get into the Sandbox specifics, there are two pieces of key context to keep front of mind; as outlined in our earlier report on test findings and what needs to be done to improve performance for advertisers: First, Sandbox API functionality is not like-for-like with 3rd party cookies, and nor is it intended to be. Second, Privacy Sandbox APIs are not frontend and ready-made solutions.  They are building blocks for ad tech to build solutions with. 

Our approach: test, test, test 

The availability of Sandbox-enabled auctions and audiences continues to trend upward and has grown from a low base, though still very small in relative terms – c 11% of Chrome impressions containing Topics, c 1.5% Chrome auctions with protected audiences*.  

Ad tech providers have extended testing opportunities throughout the year. We maintain active testing across both addressability-focussed Sandbox APIs: Topics and Protected Audience API (PA API); and across both Prospecting and Retargeting use cases. However, there is still a long way to go, as these testing opportunities have been functional and limited to simplified addressable use cases. 

Let’s dive into the specific findings on prospecting, retargeting, IP protection, and recommendations for the path forward:   

Prospecting 

While prospecting has made significant progress, with many testing options available through SSPs and other ad tech, it’s also the area where we are advocating for the most substantial strategic shift. 

Topics & the multi-signal approach 

In addition to PA API, which is used to form Interest Groups (IGs) of third-party datasets, prospecting use cases can use Topics API. Here, interest audiences are created at the on-device browser level according to user browsing habits, with an audience taxonomy roughly approximate to IAB’s taxonomy.   

Topics API groups are not directly comparable to cookie interest audiences. Additional privacy preserving features alter functionality. Noise is a key variable: 5% of Topics assignments are random, leading to reduced accuracy. Domains are also classified at the hostname level, leading to some further mismatching. Though not all features are net-negative – we welcome the prioritisation feature within Topics toward higher-value signal domains rather than toward high-frequency (and less indicative) domains. 

While we acknowledge that Topics will form the backbone of many durable prospecting strategies, we do not view Topics audiences as an effective tool in isolation. Topics audiences are too coarse to map to all client needs. Further, interest-targeting has long been a blunt tool and weak indicator of clients’ strategic objectives. Topics should be the means by which other data signals are enriched via predictive modelling.  

But here’s the issue: many ad tech partners have moved to develop their cookieless solutions without these signals; choosing instead to prioritise the better known, and currently more scalable, signals. This short termism denigrates the multi-signal approach and leaves the advantage to walled gardens and globally scaled platforms. Advertisers want and need more advanced addressability opportunities now; we urge ad tech to develop them. 

Prospecting insights are coming into focus 

Throughout H1 our tests have been largely functional across both APIs. Partnerships with SSPs have yielded quickest scaled testing via Topics. Scale and workflow implications are promising.  

Industry studies are beginning to yield useful insight into the health of prospecting across the APIs. Criteo1 and Index2 studies indicate potential average publisher revenue drops of 60% and 33% respectively. Google’s study3 reported publisher revenue recovery of c. 20% to 35% when using Privacy Sandbox in comparison to when only using alternative signals. According to Google’s study, while conversion rates also decreased, new optimization techniques limited the decline to 4%. These initial results are concerning, but we caution against final conclusions at this stage. Current market conditions are still in flux, and the present competition for cookieless traffic is distorting prices and performance in ways that may not reflect long-term trends. 

Our request to ad tech 

Advanced addressability options need to be developed by ad tech. Here are the top asks from advertisers: 

  • DSPs, SSPs and curation platforms utilising Topics as part of multi-signal predictive, AI modelled and interest-based audiences 
  • Publishers using Topics signals to enrich their PMP or Direct deals 
  • DSPs and Retail Media Networks using PA API to scale their 2PD audiences across the open web 
  • DSPs and data partners collaborating to rebuild audience marketplaces fit for a privacy-first world, using PA API 
  • DSPs, CDPs or data partners extending data collaboration opportunities for advertisers and publishers beyond clean rooms and authenticated inventory 

Retargeting 

Retargeting opportunities have lagged, but opportunities are opening across DSPs into Q3. Our early testing has launched, with more in the pipeline. Here’s what we have learnt so far – 

PA API nuances emerge 

Interest Groups diverge significantly from site-pixel cookie audiences. For example, they may only be updated once every 24 hours, and recency of site visit is noised in 1% of calls; both elements affecting the accuracy of recency targeting. It is therefore an important acknowledgement that our ability to remarket site-visitors will be somewhat reduced. Our testing schedules aim to map the effects and extent of the reduced functionality, and to mitigate against it.  

Another key yardstick is the effect on upstream operational procedures and setup parameters. We are not expecting an exactly comparable workflow, but require a retargeting solution that is efficient and does not require cumbersome processes. Encouragingly, site-tagging has not been an issue, with existing site-tags largely sufficient to enable PA API capabilities. 

We have seen a move to integrate PA API into algorithmic tech stacks in preparation for Retargeting use cases. Enhanced Automation, a feature from Google Marketing Platform’s DV360, is a case in point. Use of this feature requires a mindset shift since there is an additional data-sharing element to the feature – albeit anonymised and aggregated. Such shifts need to be properly planned with advertisers, and made explicit and transparent across ad tech.  

Retargeting takes shape 

Initial testing has been iterative and rudimentary, with audiences built and maintained on the backend by DSP engineering, and with limited optimisation and reporting functionality. We welcome this approach. As we move into a scaling phase, we will be looking for a simplified frontend workflow, and a far higher degree of practitioner control.  

While the recent momentum in testing is positive, the retargeting use case remains our biggest concern. First, Google’s recent study3 indicates a reduction in scale of up to c. 50%. This is significant and not easily replaced. Second, latency issues continue to emerge throughout industry testing, caused by new auction mechanics. Latency has negative downstream effects, including declining quality metrics (e.g. viewability). We understand testing is relatively nascent and that these early signs are not necessarily indicative of future state. Further and quicker testing is required. 

Our ask of the industry: quicker solutions to market 

There is a dependency for DSPs to develop this functionality since PA API uses Interest Groups using DSP site tags. Our ask is that DSPs maintain and increase momentum to bring more testing online quicker. 

We expect ad tech to continue to engage with the existing IAB Tech Lab’s Privacy Sandbox workstream, and with Chrome directly. Similarly, we expect Chrome to continue working with ad tech on emerging issues. 

IP Protection 

IP Protection, the obfuscation of IP addresses within Chrome Incognito, is also on the horizon, with Google recommitting to it in its recent blog. Its effects must be assessed within the spectrum of the Privacy Sandbox. 

A loss in the geographical signal 

To maintain geographical targeting under IP Protection Google proposes the assignment of “GeoIP”, a privacy-preserving method for assigning coarse user locations. 

The current proposal suggests geo-fencing based on a threshold of one million observed users within a specific region over two weeks. Population movement means observed numbers won’t mirror census data. Google’s US example4 indicates thresholds closer to 500,000 people – equivalent to a third of Manhattan, or the entire state of Wyoming. In any case, this is a significant loss in granularity by a factor of over 100. This nullifies the utility of precision geotargeting; albeit within Chrome Incognito. 

During the public consultation period we expect more details on how IP Protection will operate in other geographical regions, and on the wider ramifications for targeting and signal loss. We expect Google to continue actively engaging the industry on these developments. 

Keeping ahead of developments 

IP Protection impacts 3rd party IP tracking cross-site, with first-party usage excluded. We expect solutions on the first-party data side (supply and advertiser) to enable greater degree of control over geotargeting while maintaining privacy. We ask that ad tech and advertisers lean into these developments. 

Path Forward

The Privacy Sandbox APIs are not a complete solution for addressability in the cookie-reduced world. Instead, they form part of a broader strategy that should include other cookieless signals and approaches. Diversification is key, with addressable strategies incorporating clean rooms and alternative IDs where authenticated signals are available.  

At GroupM, we’re committed to testing and comparing these new solutions against existing strategies to ensure we’re delivering the best possible outcomes for our clients. As the digital advertising ecosystem continues to evolve, our focus remains on developing advanced addressability solutions that balance effectiveness with user privacy. 

Call to action 

Advertisers should urge their ad tech partners to develop the solutions they need, while maintaining testing momentum.  

The ad tech industry must deliver more solutions quicker, prioritising advanced addressability options which go beyond the basic use cases, while continuing to engage with the IAB Tech Lab and Chrome. 

  1. https://www.criteo.com/blog/privacy-sandbox-testing-results-show-shortfalls-to-meet-cma-requirements/ 
  2. https://www.indexexchange.com/2024/07/02/insights-privacy-sandbox-testing/ 
  3. ads-privacy/mode_b_testing_whitepaper.pdf at master · google/ads-privacy · GitHub 
  4. https://github.com/GoogleChrome/ip-protection/blob/master/Explainer-IP-Geolocation.md  
  5. *Source: Global SSP partner, Q2 2024