MIAMI — Deep learning is everywhere in advertising these days.
But for an ad platform like Cognitiv, deep learning has been a core part of its offering since its founding a decade ago. For Jana Jakovljevic, SVP of Partnerships at Cognitiv, real-time adaptation to consumer behavior isn’t a new trend but a fundamental principle of effective advertising.
“Cognitiv chose deep learning for its ability to ingest vast amounts of data, find complex patterns in that data and make accurate predictions,” Jakovljevic told Beet.TV contributor David Kaplan at the POSSIBLE conference. “This means advertisers don’t need to understand all of that data they have. Deep learning will understand it for them.”
Beyond traditional audience targeting
Being able to aggregate data in a clearly digestible, understandable way is one thing. Putting that understanding in context takes a bit more finesse and dialogue between the platform provider and the advertisers. To get there, Jakovljevic noted that updating what and how media consumption is measured is key. And that process depends on moving past the limitations of traditional audience segmentation strategies.
“Deep learning can evaluate user and context in unison,” Jakovljevic said. “Advertisers don’t need to have a separate strategy for audiences or decide whether to target certain demographics, which is quite static. They can just provide all of their first-party data and the deep learning algorithm will understand what’s important in that moment.”
This capability allows marketers to focus on more meaningful metrics than simple clicks or impressions.
Driving true incrementality
Cognitiv’s platform distinguishes itself by stressing business outcomes over proxy metrics that don’t necessarily translate to real business impact.
“When we built Cognitiv, we built everything with custom in mind,” Jakovljevic said.. We wanted to deliver a custom algorithm to every marketer using deep learning, which optimized to their desired KPI.“This means we love to optimize towards business metrics over proxy KPIs like click-through rate.”
The platform’s approach to incrementality involves a scientific testing methodology that determines when ads should actually be shown.
“We’ve always taken a very scientific approach to this, having very strict testing control groups so we can understand what’s the conversion rate if we don’t show ads,” she added. “A deep learning algorithm for every ad opportunity is predicting what’s the probability of conversion if we show an ad, and what’s the probability of conversion if we don’t show an ad.”
This allows for smarter allocation of advertising budgets. “If we’re predicting this user is going to convert regardless of whether they see an ad, then we’re not going to show them an ad,” Jakovljevic said. “We want to make sure we’re spending the marketer’s money as efficiently as possible and driving true incremental lift.”
Real-time adaptation delivers results
Unlike many ad platforms that rely on historical data or static audience segments, Cognitiv’s deep learning approach continuously adapts to changing consumer behavior.
“Most of the industry today is still relying on stagnant audiences or they’re using historical data to make decisions in the now,” Jakovljevic said. “However, what worked yesterday may not work today. As people’s behavior shifts, deep learning is able to deliver advanced performance because it can evaluate all of that data in real time.”
The system’s constant feedback loop strengthens its predictive capabilities. “It’s learning all the time. It understands, ‘Hey, I chose to serve this person an ad. Did they do that desired action that I wanted them to do?’ And then it’s constantly reinforcing itself,” she added.
Context remains undervalued
Back to the topic of context, Jakovljevic believes many marketers are still underestimating the importance of how that concept works now in driving advertising effectiveness.
“A lot of the industry is still underestimating context, and I’m not talking about context in terms of keywords or these static lists of contextual segments,” she said. I’m talking about considering context alongside the user.”
This take on context enables more precise message targeting. “What works in the morning might be different to what works in the evening and what works for you might be very different to what works for me,” Jakovljevic said. “You really need technology that can evaluate all of that real-time information to not only determine, yes, I should serve this person an ad, but actually what message do I want to serve this person.”
You’re watching “The New Era of Intelligent Advertising,” a Beet.TV Leadership Series at POSSIBLE 2025, presented by Cognitiv. For more videos from this series, please visit this page.
