Researchers at the University of Notre Dame report that conventional measures overstate the effectiveness of traditional television advertising. The result is important because advertisers continue to allocate far more to linear TV — roughly $139 billion this year — than to streaming and connected TV, which are expected to receive $33 billion.
The team, led by Shijie Lu with Tsung-Yiou Hsieh and Rex Yuxing Du, linked second-by-second viewing data from LG smart TVs to purchases on a food delivery app. They analyzed activity for millions of opt-in households over a four-month period and focused on broadcast networks such as NBC and ABC; the study did not track streaming apps like Hulu or Amazon. By exploiting natural timing differences — for example, whether a household watched the part of a live game when a given ad aired — the researchers created a natural experiment that separates true ad effects from confounding factors.
The authors find that methods based on ratings and aggregate market data overestimate ad effectiveness by 55% in this context. The study also offers practical insights:
- Promotions for first-time buyers increase retention.
- Viewer responsiveness peaks within two days of a purchase, with highest engagement among customers who ordered two to four times before.
- Younger, tech-savvy sports fans respond better than older news viewers.
The researchers conclude that isolating ad exposure timing allows more precise measurement of return on investment and can help target TV ads based on actual purchase behavior rather than only demographic data. Source: University of Notre Dame
Difficult words
- overstate — present something as larger or more effective than true
- allocate — set aside money or resources for a purpose
- opt-in — agree to join or allow participation or data use
- confound — make cause and effect hard to separate or see clearlyconfounding
- retention — continued use or repeat purchase by customers
- responsiveness — speed and strength of people's reactions to ads
Tip: hover, focus or tap highlighted words in the article to see quick definitions while you read or listen.
Discussion questions
- How could advertisers use second-by-second TV viewing data to improve ad targeting and spending decisions?
- What privacy or ethical concerns arise from linking smart TV viewing data to purchases, even for opt-in households?
- Given the study's finding that conventional measures overstate TV ad effectiveness, would you change ad budgets between linear TV and streaming? Why or why not?
Related articles
Rethinking 'the Human' in AI
Artist and writer Xonorika Kira argues that centring the human in AI can exclude other forms of knowledge and intelligence. She proposes practical changes like small, consent-based datasets and slower models to support communities and cultural sovereignty.
AI expands sexual and reproductive health information in Peru and Argentina
Health groups in Peru and Argentina use artificial intelligence to give sexual and reproductive health information to young and marginalised people. Projects include a Quechua chatbot and a WhatsApp platform, but experts warn of bias and access problems.
Africa uses AI to strengthen health systems and self-reliance
At the CPHIA conference in Durban, Africa CDC said AI and digital tools can help protect 1.4 billion people, improve surveillance and support primary health care. Data governance, infrastructure and domestic financing are key concerns.