AdTheorent Health, a division of AdTheorent Holding Company, recently announced the launch of Predictive Audiences. It allows programmatic advertisers to target audiences in a more precise, data-driven and less opaque manner than previously possible.
“The data provides a syntactical view of parameters to develop an audience segment not based on IDs, but all these other aggregate data points,” said AdTheorent CEO Jim Lawson. “An audience set might be people diagnosed with high cholesterol within the last 12 months subscribed a certain medication at a specific pharmacy.”
"We first must find the parameters from raw sources data,” he said. "It's the most privacy-compliant way to target advertisements."
AdTheorent Health Predictive Audiences are customized based on advertiser requirements and use primary-sourced healthcare data and statistics to identify the most qualified audiences for targeting consideration, without relying on personalized information or user IDs.
AdTheorent aggregates health data from hospital and healthcare provider claims and electronic health records, as well as pharmacy and contextual data representing more than 300 million patients and Health Care Professionals, which equates to more than 90% of the U.S. population.
This aggregated health data is mined to create statistical representations of desired target audiences such as condition sufferers, caregivers, HCP or prescription users. It is based on non-sensitive data such as age, location, gender HHI and education level.
This data foundation is not ID-based and does not include personal health information (PHI), making AdTheorent Health Predictive Audiences HIPAA- and privacy-compliant.
In the past, companies had to create healthcare audience segments based on data from pharmacies, healthcare providers and other sources. The audience is made up of 10% of condition sufferers that fit the criteria for a specific segment, and 90% of non-condition sufferers, which is intended to adhere to HIPPA regulations.
In the prior method, impressions are served to anyone within the audience. AdTheorent only delivers an impression when an ad opportunity is within the customized target audience and the predictive models indicate a higher probability of completing the advertiser-specified campaign action such as prescription fill, digital action, audience quality, and more.
AdTheorent Health Predictive Audience models evaluate 900 impression-specific signals and millions of potential impressions per second to drive performance, and to continuously optimize throughout the campaign, allowing AdTheorent to drive performance for healthcare advertisers. From this is scores the propensity that a user will carry through with a given action.
The algorithm continually learns. "It's modified and refined as more conversion activity comes in," he said.
AdTheorent began using this method in the healthcare industry, but Lawson said the company is building out predictive audiences for other segments.