LAS VEGAS — Artificial intelligence models trained on low-quality data risk generating meaningless insights that then degrade further as models train on their own flawed outputs, creating a destructive cycle that obscures actual campaign performance.
“One thing that I fear, especially with all the excitement around AI, is that the quality of the training data that goes into a lot of the AI models being used more and more often in the marketing ecosystem, that that training data is not up to snuff,” Justin Rosen, SVP Data & Analytics at Ampersand, told Beet.TV Editorial Director Lisa Granatstein at CES. “What that will risk is the AI models that you have built and that may become commoditized are going to generate insight that doesn’t have a lot of signal to it that isn’t very meaningful.”
This creates potential for doom loops where degraded data trains subsequent models, compounding inaccuracy over time.
Combatting fragmented planning
Household-level insight provides the granular, scaled data needed to plan confidently across fragmented platforms, though achieving complete data sets remains challenging due to overlap issues between different sources.
“In order to plan across a fragmented ecosystem, you need to have as granular a data set as possible and as complete of a data set as possible to have the breadth of insight and the depth of knowledge,” Rosen said. “If you can get those assets to be joined together in the right way, your planning capabilities will be that much more powerful.”
The fragmentation extends beyond screens and viewing to the underlying data infrastructure itself.
Stability and scale
Ampersand’s identity foundation built through MVPD partnerships combines privacy-compliant anonymous IDs associated with stable postal addresses and IP addresses from direct sources rather than third parties.
“Our identity spine is privacy compliant, anonymous ID that can be associated with postal address, which is stable and at scale, as well as IP address, which can be an unreliable signal, but not for us because we’re the source of that IP address,” Rosen said.
This combination enables confident linkage between household and device levels while maintaining accuracy and privacy compliance for performance measurement.
Healthcare complexity
Healthcare advertising confronts heightened privacy regulations that vary by state without near-term prospects for national standardization, increasing fragmentation challenges for sensitive health data.
“There are frequent differences in regulation at the state level,” Rosen said. “I don’t see a national regulation coming anytime soon for that category. That even increases that fragmentation even more because the way one state deals with these policies might be different than another.”
Scaled data assets become critical for reaching niche condition-specific audiences while navigating varying compliance requirements across geographies.
Speeding towards new data sources
The ecosystem must accelerate insight generation while fragmentation continues expanding through new data sources and media tactics that require incorporation into unified views.
“We need to get a little bit faster and we are getting faster, but we’re getting faster while the world behind us gets more fragmented,” Rosen said. “Every time that we solve a challenge to reaggregate what has been fragmented, there’s a new data source or a new media tactic that we want to incorporate into that complete view.”
Insight as differentiator
Ampersand focuses on ensuring viewing insight, identity, and ad exposure accuracy at the highest possible scale as AI becomes normalized across marketing functions.
“The predictive insight is one of the major upsides of something like an AI first marketing world,” Rosen said. “We are focused on making sure that the viewing insight, the identity, the ad exposure that goes into that program is accurate and at the highest scale possible. And we think that that will increasingly be a differentiator as AI becomes the norm across everything that we do.”
You’re watching “Omnichannel Starts with TV: The Evolving Role of TV in Healthcare Marketing”, a Beet.TV Leadership Series, presented by Swoop. For more videos from this series, please visit this page.
