Skip to Content Skip to Footer
How Second-Screen Users Respond to TV Ads at the Spot Level

How Second-Screen Users Respond to TV Ads at the Spot Level

Rex Yuxing Du, Linli Xu and Kenneth C. Wilbur

second screen users

When NBC coined the term “Must-See TV” Thursday in 1993, a generation of Americans stayed home to watch Seinfeld, Friends, and ER. Today, media producers are hard-pressed to capture viewers’ full attention because 178 million Americans regularly use a second-screen device while watching TV. On the other hand, ready access to a second screen empowers TV viewers to take immediate actions after seeing an ad, such as search for product reviews and prices, express opinions on social media, or place an order on the advertiser’s website.

This phenomenon has inspired a new class of attribution vendors that promise TV advertisers to link post-ad spikes in online activities to the individual TV ads that caused them. Advertisers can then use these measures to assess the relative effectiveness of ad spots to improve ad copy and media placement decisions. The ultimate goal is to improve the cost-effectiveness of TV as an advertising medium. However, conversations with practitioners indicate that many mechanically refine their TV ad creatives and media schedules without a deeper understanding of the TV-to-online spillovers. It is one matter to show that TV ads can cause statistically significant immediate post-ad spikes in online activities; it is another to measure spot-level responses with such precision that one can quantify the relative performances of different ad creative and media placements. A new study in the Journal of Marketing explores the inner workings of this phenomenon to empower advertisers with the method they need to make these important choices.


Our research team presents a rigorous and yet practical framework that links TV ad insertions to minute-by-minute online search, which is decomposed into the sum of a baseline, an immediate response caused by TV ads, and an error term. The baseline includes an hourly fixed effect and a within-hour trend that can vary by hour of the week. The immediate response to an ad spot has an empirically determined duration and a flexible decay pattern. The immediate impact of each ad on online activity is modeled as the product of ad audience size (or cost, when audience data are not available) and response rate, which in turn depends on the characteristics of the ad creative, media placement, and audience. The error term is serially correlated, with the pattern determined empirically.  
We focused on three top pickup truck brands, for which we compiled a rich dataset by stitching together information from multiple sources, covering a span of nearly a half million minutes of online search and TV exposure. We focused on brand search and price search. We observed 27,562 ad spots on national TV and 750,672 spots on local TV. By merging the spot level ad data with the minute level search data, we built a comprehensive testing ground to demonstrate the worth and insights available from estimating the linkage between TV ad spots and immediate online response. 
Our research offers several key takeaways. First, for both brand search and price search, there is a detectable spike immediately after a regular ad insertion, be it on national or local TV. Second, nearly all of the immediate response occurs within five minutes of an ad insertion—brand search response peaks during the minute after the ad is aired and then dissipates quickly; price search response is spread out more evenly over the five post-ad minutes. Third, besides generating immediate own-brand searches, national TV ad insertions also lead to significant competitor-brand searches but little competitor-price searches. Fourth, national spots appear to be more cost-effective at generating immediate brand search response, whereas local spots appear to be more cost-effective at generating immediate price search response. Finally, given ad audience size, immediate search responses vary with ad creative characteristics, audience category interest, slot of the break, program genre, and time factors. 
Managerially, our findings about positive lifts of certain media placements (e.g., first slot, prime time, live sporting event) and audience category interest suggest that when TV advertisers intend to focus on maximizing one particular type of online response, large gains in effectiveness could accrue from quantifying and balancing the multiplier effects of various media and audience factors against their cost differentials. That said, our findings about divergent effects of broadcast/cable, weekend/weekday, national/local, and ad creative characteristics on brand vs. price search caution advertisers against relying on any single immediate online response metric to assess media placements and ad copies. There is not likely a media plan or ad creative that is optimal for all types of online response. 
Practically, unlike the proprietary methods used by advertising attribution vendors, our proposed framework for modeling behavioral response at the minute level is transparent and readily replicable. Advertisers, agencies, and networks can use our method as a benchmark when evaluating the proprietary solutions offered by attribution vendors. TV advertisers could further extend our modeling framework to include website traffic, online transactions, social media activities or other important behavioral indicators that vary at the minute level.

Read the full article.

From: Rex Du, Linli Xu, and Kenneth Wilbur, “Immediate Responses of Online Brand Search and Price Search to TV Ads,” Journal of Marketing, 83 (July).

Go to the Journal of Marketing

Rex Yuxing Du is Bauer Professor of Marketing, Bauer College of Business, University of Houston.

Linli Xu is Assistant Professor of Marketing, Carlson School of Management, University of Minnesota.

Kenneth C. Wilbur is Associate Professor of Marketing, Rady School of Management, University of California, San Diego.