“Bumper results for brands” is how Financial Times put it. Six quarters and two Centers of Excellence for AI later – one for advertisers, another for user experience – it seems like Meta is digging itself out of the ATT rabbit hole.

Last quarter, Meta advertisers saw 20% more conversions on ads compared to the year before, with outcomes like clicks and sales. At the heart of this transformation is Advantage+, Meta’s AI-led ad tool that yields a ROAS of seven – about as high as before Apple’s privacy changes, to a 20 – 30% lift in revenues compared to non-Advantage+ adverts, agencies say.  

In the new privacy landscape with less data to track consumers, Meta uses AI in both creative and targeting. To tweak ads, to make small enhancements like adjusting brightness, aspect ratio, to test placements, to improve performance of creative Advantage+ deploys generative AI. Being able to relate to individual searches on an almost one-to-one basis, reflecting on the users’ search terminologies and preferences, is going to be the biggest benefit of generative AI in advertising. Now, we have an at-scale mechanism to sync with the voice of the customer by incorporating it in the ad copy and the associated imagery used – at least on an aggregate basis. Composable ad creatives, in situ advertising, in vitro advertising – call it what you want to, generative AI is helping endear end users to brands.  

From thereon, conventional AI takes over, serving these variants of ads to cohorts of hundreds of thousands of viewers and optimizing iteratively for higher efficiencies. Since the ads are going to be more efficient, in terms of being able to reach out to targeted audiences with a high probability of converting, CTRs, CVRs and CPCs will likely be headed north. Any wonder then, that Meta’s Advantage+ receives more funding than Metaverse, per an insider?

The jury is still out on the AI-based Variance Reduction System that Meta launched to minimize bias in the distribution of its ads as part of a settlement with the DoJ, however.  

Different blokes, diverse strokes

In using AI to fill in the shoes of deprecating cookies, players adopt myriad playbooks. Google’s Topics API mines users’ browser history to find interests and serve relevant ads. Meta’s Conversions API bypasses the browser altogether and transfers data from ad servers directly to Meta’s databases.  

IBM’s Watson Advertising focuses on an AI-based contextual targeting approach to serve timely and relevant ads to viewers.  

Many other companies are building identity graphs using probabilistic data, to connect disparate datasets and get a holistic view of the consumer. Pattern Mining algorithms seamlessly match online and offline shoppers’ identities and improve the accuracy of identity graphs. Natural Language Processing (NLP) helps in gauging sentiments to serve suitable ads. This way, marketers fine-tune their recommendation systems and avoid wasted ad impressions.

Affinity based clustering of consumers is one such pattern-recognizing model that is beginning to catch the fancy of DTC marketers. These models describe a consumer as a collection of strong or weakly held affinities. For instance, a middle-aged man in California wearing a Patagonia vest (as seen from his selfie on Insta), might also rave about Eight Sleep on Twitter. So, the model would recommend targeting all Patagonia fans with Eight Sleep ads.

Pattern-mining based on probabilistic data and generative AI for copies and images are going to become native features in most social ad platforms. Meta’s about-turn validates this further.

‍