From User IDs to Cohorts, the Future of Ad Tech Is Now Clear

Ad-tech executives and product designers should be breathing a sigh of relief now that Google has announced it will not support a cross domain identifier when Chrome deprecates the third-party cookie by 2022. Until now, the industry has been focused on finding backward compatible work-arounds, such as redesigning device graphs to work in the first party and focusing on industrywide identity solutions, such as Unified ID. Now that cohorts are the future of audience management in digital advertising, we can finally focus on building systems to support them. Shortly after Google announced in 2018 to stop sending UUIDs in ad server log files, I sponsored a project alongside a team at MIT. We investigated the level of loss in predictive power of an algorithm given the artificial constraint of grouping users into minimum size cohorts of 50. Randomized cohort creation retained about 65% performance while guided cohort creation improved to around 80% retention. As such, generalized cohort management for optimization is a viable alternative to ID-based optimization. The loss is even more acceptable if you consider that the current algorithms are challenged by the red herring of measurement and model bias. It makes sense that Google’s announcement that FloC retains 95% of its predictive power was quickly followed by Google’s announcement to not support personal data-enabled cross publisher tracking across Google products. As long as the market is willing to accept a somewhat throttled (yet acceptable) level of output, the consumer privacy benefits of ID-less cohort management outweigh the costs. Forward-thinking ad-tech companies have been making plans to redesign their systems to operate in the ID-less anonymous web as a complement to and backup of the Unified ID effort. Google, Apple and Mozilla’s announcements open the door to a field of disruptive entrants that enable cohort-based methodologies. We have worked hard on quantifying audiences, and spent some time doing the same with media, but we have neglected quantifying creative. Most people in the industry know that creative and context are the most likely alternative solutions to rise as the audience dimension becomes more challenged. As an industry, we have worked hard on quantifying audiences, and spent some time doing the same with media, but we have neglected quantifying creative. It is the last mile in digital media effectiveness that proves that the creative asset is the most powerful dimension of ad effectiveness. Below are some lesser-known categories that enable marketing effectiveness to thrive against the backdrop of cohort-based methods. Creative solving for banner blindness Digital media suffers from banner blindness. One of the ways to address this is through alternative ad formats that are closer to native. This category is being covered by startups with display ads that feel like social media, which make banners highly engaging and conversational. The creative elements drive ad attention and engagement that is far superior to traditional digital banner ads. Creative as code The audience and media dimensions have been parameterized by digital tech companies. However, creative design is still highly grounded in the legacy of print, TV and OOH and has not been quantified. Some use survey methods to score creative, and others are making advances in computer vision that allow research professionals to quantify images and videos as they would text. Experimental design There are two fundamentally different approaches to measurement and optimization: observation and experimentation. As backward-looking data sets across multiple sources become harder to obtain, experimental methods are favored. Many ad creative software companies use experimental design methodologies that can be repurposed to support platforms. Advertiser cohort management The ex-Adobe/Omniture team at marketing data governance platform Claravine is a good example of a new entrant that helps brands create data governance and standards across their digital ad buys, which become a necessary foundation to enable cohort management. Publisher cohort management Look to first-party DMPs, such as publisher data platform Permutive and the redesigned data management company Lotame, to grow in relevance. Both will increasingly focus on publisher first-party data enrichment as the buy-side of the programmatic market develops support for ID-less and cohort-based programmatic trading. First party to first party onboarding Even though the cross entity personal data bridges enabled by email login will not be supported by the Google ad stack, the announcement does not prohibit direct advertiser to publisher onboarding. Companies such as infrastructure organization InfoSum and software company Habu give first parties the ability to make a security-compliant direct connection between advertiser and publisher data without the necessity of a device graph in between. These direct connections provide a powerful sample for brands and publishers to create their own cohort management methods. Measurement panels Sampling and probabilistic projection take a big leap in importance as log level deterministic tracking is reduced in scale. The pending Nielsen One solution and the attention-based panels from advertiser TVision are category examples that should gain from the move to cohort-based methodologies. With the increased likelihood that cohorts are here to stay, the industry now has the confidence to make technical investments that have mostly been on hold for the last three years. Ad tech can now build for a cohort-ready future that upholds both user privacy and can increase marketing performance. Walled Garden Walls Will Get Higher Under Google’s New Privacy Policies