Rameez Tase – Co-Founder & President, Antenna
Gilles Duterque – VP, Data Science, Antenna
In this presentation, Rameez Tase and Gilles Duterque introduced Antenna’s approach to understanding how streaming viewership drives business outcomes like subscriber acquisition, retention and churn. Because traditional TV ratings do not capture the fragmented viewing behavior in streaming, they cannot tie ratings or viewing behavior to subscription success. Rameez and Gilles described the process and methodology behind Antenna’s solution, “Subscriber Views,” which addresses the gap by linking viewership data with subscription data, enabling analysis of what content drives sign-ups, keeps users engaged or leads to churn. The presentation emphasized that the goal is not just audience measurement, but actionable insights into user behavior across cohorts and platforms.
Download Presentation
Member Only Access
Dan Reines – SVP, Research & Insights, SmithGeiger Group
Brian Sabella – Director, Original Content & Research Innovation, ESPN
Dan Reines of SmithGeiger and Brian Sabella of ESPN explored how conversational AI research methods can deepen understanding of sports audience behavior in an increasingly fragmented media landscape. Methodologically, the study employs a “conversational survey” approach, where AI moderators simulate real-time dialogue with respondents, blending qualitative depth with quantitative scale. Across two waves totaling 2,000+ respondents, this method enables longer, richer, open-ended responses (≈45 words vs. ~10 in traditional surveys) and iterative probing similar to in-depth interviews, while maintaining statistical rigor. This approach allows researchers to uncover not just what audiences do, but why behaviors are shifting across platforms and formats.
Download Presentation
Member Only Access
Pedro Almeida – CEO, Mediaprobe
Ed Hunt – Research Associate, Medill Spiegel Research Center, Northwestern University
Larry DeGaris – Executive Director, Medill Spiegel Research Center, Northwestern University
Pedro Almeida (Mediaprobe), Ed Hunt and Larry DeGaris (both from Northwestern University) examined how media context influences advertising effectiveness. They introduced the concept of reverse context effects, which describes the impact of ads on programming. Their study used Mediaprobe data, including galvanic skin response and declarative measures, across more than 10,000 ads and multiple content types such as sports, scripted and unscripted programming. Results revealed that higher emotional engagement in programming increases subsequent ad engagement, while ads also elevated emotional responses in the surrounding program. Sports programming was identified as producing stronger emotional carryover effects. Additional findings showed that ad recall declined over time, although strong creative or sustained engagement mitigated the decline.
Download Presentation
Member Only Access
Joe Garland – Director of Data Analysis, Comscore
Michael Vinson, Ph.D. – Chief Research Officer, Comscore
Bill Engel – Board of Directors, Tenetic
Comscore’s Michael Vinson and Joe Garland, along with Tenetic’s Bill Engel, examined how to measure audience reach and overlap across linear TV and digital platforms by comparing traditional panel-based approaches with a simpler alternative—the independence assumption, which estimates separate audiences’ overlap by multiplying their individual reach.
Using Comscore’s large-scale dataset that includes set-top box data covering about one-third of U.S. TV households and digital census data linked via household identifiers, the researchers created a “ground truth” dataset of roughly 2 million households. They then simulated panels of varying sizes and compared their overlap estimates against both the true observed overlap and estimates generated using the independence assumption. The methodology also included weighting panels to reflect U.S. Census demographics and calculating required sample sizes under different confidence intervals, revealing that accurate panel-based estimation often requires impractically large samples. As a result, the presenters concluded that panels are better suited for calibration and validation rather than primary measurement, and that future solutions should rely on hybrid approaches, including adjusted assumptions and machine learning models, to better account for real-world audience behavior.
Key takeaways:
- Panels are not scalable as a primary measurement tool. They are expensive, hard to recruit and require unrealistic sample sizes to be accurate.
- The independence assumption is a strong baseline. Despite being simple, it often delivers results close to reality in low-correlation scenarios.
- Correlation is the critical challenge. When audiences are behaviorally linked (e.g., fans, binge viewers), both panels and independence assumptions can break down. Scenarios where the independence assumption do not hold will require continued methodological development, supported by richer data and more flexible modeling frameworks.
- The future is hybrid and model-driven. Combining large datasets, adjusted assumptions and machine learning will produce more accurate and scalable overlap measurement.
Watch the Presentation
Download Presentation
Member Only Access
