Executive Summary
​
A consumer study to measure the impact on viewer attention and brand perception of CTV advertising revealed that ads targeted using AI-enabled contextual data outperform targeting using standard demo and publisher-declared metadata.
Viewers paid more attention to the ad, learned more about the product and were more interested in the products shown in AI-enabled contextually targeted ads.
Respondents frequently commented that ads were engaging even if the product wasn’t relevant and that they would remember the product and bring it up in conversation.
​
The study also revealed that viewers are sensitive to irrelevant and brand unsuitable ad placements resulting in reduced attention and increased negative perception of the brands and products appearing in the ads.
​
Objectives
​
At its current pace, CTV ad spending in the US is expected to grow to $41 billion by 2027*. As linear TV ad budgets follow viewers into CTV, buyers are mindful of the significant differences in targeting capabilities and their impact on viewers attention and brand perception.
​
Linear TV is delivered to millions of viewers who typically see the same ads simultaneously. Targeting is based on the network and viewership demographics, while brand suitability is determined by program restrictions at the show level.
​
In CTV, millions of viewers stream their favorite shows, but viewers see different ads because targeting is based on identity and demographic data sets. Due to technology and legal reasons, buyers cannot apply the same brand suitability program restrictions in CTV. Unlike a mobile device which is personalized to the viewer, a smart TV is shared in a household of several viewers. As a result, buyers are unclear who saw the ad what type of content the ad appeared within.
Contextual targeting enabled by AI is being used to increase ad relevance and therefore drive attention and brand perception. This study seeks to measure its effectiveness in driving these metrics compared to demo targeting and publisher-declared metadata.