A Q&A with Dr. Aaron Andalman, Chief Science Officer & Co-Founder at Cognitiv

We know that the cookie has dominated our feeds (and lives), especially with the recent news of Google delaying its removal of third-party cookies on Chrome — again — however, Cognitiv has lived in the cookieless future for years. Our 2024 launch of the industry’s first Deep Learning Advertising Platform (DLAP) comes from the culmination of years of developing technology independent of cookies.

This week, we are excited for our Spring Neural Networking 2024 event, where we are hosting “A Sweet Sendoff Into the Cookieless Future.” At the event, we will hear from industry experts leading the charge in this ever-evolving advertising space. As we gear up for an engaging panel discussion this week, we sat down with Cognitiv’s Co-Founder & Chief Science Officer, Dr. Aaron Andalman, to discuss:

• • The recent delay in the removal of third-party cookies
• • How to adapt to the cookie decline
• • The use of AI technologies, particularly contextual targeting
• • Revolutionizing advertising with multimodal models
• • Evaluating campaign effectiveness post-cookies

Read on for more insight into the cookieless future and what lies ahead for advertisers as we head into the second half of this potentially cookieless year.

Why is it important for marketers to adapt to the decline of cookies now, despite Google’s recent delay in implementing its post-cookie plan following challenges with UK regulators?

In Q1’24, Google initiated its cookie phaseout by testing with 1% of users as part of the initial test group, while the other 99% were expected to experience the phase-out by the end of the year. Although Google assured us they were sticking to the plan, they once again postponed the depreciation of third-party cookies due to regulatory challenges and insufficient readiness.  However, the undeniable reality is that the cookieless future is imminent, and preparation should be a non-negotiable.

Now is the time for advertisers to seize the opportunity to identify and implement the most suitable cookieless solutions for their needs. Given the potential learning curve associated with new insights and outcomes, experimentation will be an essential part of the process.

Additionally, the data shows that advertisers are still fighting for cookied audiences. According to Raptive’s Chief Strategy Officer, Paul Bannister, the company found a 30% decrease in CPMs for non-cookie impressions. Advertisers would be wise to redirect their focus from the dwindling pool of cookie impressions and invest in exploring cookieless solutions, which are poised to dominate the landscape by 2025.

What solutions should advertisers look to implement as they make this transition?

The good news is that advanced AI-based contextual advertising is already performing well for advertisers looking to reach audiences in new ways. Thanks to significant technological advancements, we have seen a seismic shift in technology’s ability to comprehend and leverage context. Large Language Models (LLM) and Multimodal Models are at the forefront of this revolution, like the models powering our ContextGPT™ product.

So, how does advanced contextual AI technology work without cookies?

In Cognitiv’s case, our models are trained to analyze and understand web content with unprecedented nuance, enabling more precise and effective ad placements. ContextGPT™ uses Natural Language Processing (NLP) to “read” the content of a page with incredible richness and gain a human-like understanding of a page. Contextual targeting is more powerful and accurate than ever before, making it a realistic and viable alternative, whereas it was seen as supplemental in the past. A recent study found that it can be 3-4X more effective than contextual targeting that only relies on meta-data. Our company’s investment in this technology has yielded significant improvements in ad relevance and engagement and opened new avenues for scalable advertising strategies.

(Note: In October 2023, we launched two features built on top of our deep learning contextual capability: Inclusivity and Diversity. You can read more about those features here.)

Something you mentioned earlier was multimodal models. Can you expand on what they are?

While LLMs understand a single modality – the written work – Multimodal models can process and understand information from multiple modalities or sources of input. For example, a multimodal model could consider text, images, and audio data simultaneously to perform tasks such as understanding, generation, or classification.

What do multimodal models do for advertisers?

As we look ahead to the rest of 2024, the advent of multimodal models promises to revolutionize contextual targeting further. The exciting part about these models is that they not only understand text but also analyze page layouts, the interplay between text and images, and the overall quality of ad placement. By enabling a more holistic understanding of a webpage and a particular ad placement, multimodal models will allow predictions about viewability, attention, and lift to be much more accurate. This improved accuracy will help offset the loss of the third-party cookie.

Ultimately, this holistic understanding of content will enable even more sophisticated and effective targeting, marking a new era in digital advertising.

How can marketers measure success with new AI approaches?

A significant challenge in a post-cookie world will be measuring the effectiveness of advertising campaigns. Traditional metrics, heavily reliant on user tracking, will no longer be as accessible because they—such as website traffic, user engagement, and conversion rates—depend on cookies to collect and store information about users. However, alternative methods are emerging.

By extrapolating data from the subset of users who still have cookies and employing new attribution models like UID2.0 and LiveRamp’s analogous system, advertisers will continue to be able to gauge the success of their campaigns and refine their strategies.

Any last thoughts you want to leave us with before Neural Networking?

As we are navigating the transition away from third-party cookies, the rise of AI-powered contextual targeting offers a beacon of hope and opportunity for advertisers. Multimodal models are set to transform the advertising landscape, making it more effective, privacy-compliant, and user-friendly. The future of digital advertising is not just about reaching audiences but engaging them in the right context, and the innovations of 2024 are poised to make this more achievable than ever.