February 23, 2024

by Nicole Montgomery and Kevin Synnott

8 min read

In today's digital era, data reigns supreme. Companies are on a relentless quest to gather and decipher data to unearth insights that drive informed business decisions. At Tracer, we recognize the pivotal role of context in data analysis. We created Tracer to empower organizations to seamlessly incorporate relevant business context into their data analysis processes.

With OpenAI's recent launch of Sora, a revolutionary tool that transforms text into videos, we're stepping into a new era of content creation. At Tracer, we believe in the power of leveraging AI tools not only for content creation but also for content analysis. We're keen to investigate how these AI solutions can be combined to draw context and insights from video content. This approach is vital for advertisers looking to sharpen their creative strategies with robust insights.

In this piece, we aim to unpack the potential of AI-driven video analysis, focusing on how the synergy between Google Video Intelligence and ChatGPT can transform advertisers' approach to video content. We'll explore the integration of these sophisticated tools to unlock video data's full potential, enabling advertisers to cultivate more resonant and effective campaigns.

To explore this, we analyzed the television advertisements from Super Bowl 2024 alongside data from the USA Today Ad Meter. Ad Meter invites members of a panel of U.S. adults to rate each Super Bowl ad on a 1-10 scale, with 10 being the highest. These ratings determine the year's most beloved commercial and have been a respected metric for measuring Super Bowl ad impact for over 35 years.

Our analysis focused on uncovering how different features of each Super Bowl commercial impacted its Ad Meter scores and number of online views. Our approach was comprehensive—using Google Intelligence AI, we reviewed Super Bowl commercials, identifying key elements like objects in the ads, the presence of people, logos, speech, and how often scenes changed. We then used ChatGPT 4 to analyze the speech from each ad for tone and sentiment, adding another layer to our creative analysis.

After aggregating this rich contextual information, we utilized Tracer Tags to seamlessly link the creative details of each ad with its corresponding USA Today Ad Meter rating and viewership numbers. This efficient process facilitated the construction of a dynamic dashboard, unveiling the aspects of Super Bowl ads that resonate most with audiences. The insights gleaned shed light on the elements that contribute to an ad's success and pinpoint areas for improvement.


The resulting insights are as follows, revealing key takeaways. Please see the embedded dashboard at the end of this page for more information.


Using Google Video Intelligence AI, we collected data on every object in each frame of each Super Bowl commercial. Leveraging Tracer tags, we quickly analyzed and identified the key objects and labels present in the ads. We then examined the relationship between these objects and labels with both the Ad Meter Score and the number of video views.

The table below provides a summary of the findings from this analysis. Commercials that included a vehicle received a higher USA Today Ad Meter rating, but, on average, these ads garnered fewer views compared to ads without a vehicle. While this insight may not be particularly advantageous for car manufacturers, it could guide their strategy on how often to include vehicles in their advertisements for future campaigns.


People & Number of Shots

Google Video Intelligence enhances our analysis with insights into video structure, such as the total number of shots and the presence of individuals within those shots. It's crucial to note that our methodology counts each appearance of an individual in separate shots as distinct. Therefore, if a single individual appears across 10 different shots within an advertisement, this will be quantified as 10 individual appearances. This approach allows us to capture a more detailed picture of how characters and their visibility play a role in the ad's composition and potential impact.

Furthermore, Google Video Intelligence excels in accurately detecting crowds and appropriately excludes them from the analysis. For example, in the Sleep Number commercial, showcasing a football player making a catch amidst a packed stadium is intelligently recognized as featuring just one key individual. This level of precision bolsters the reliability and depth of our insights.

The graphs provided below illuminate key patterns in how the number of shots and the presence of people influence Super Bowl commercial performance. Each graph plots one aspect—shots or people—with variations represented by bubbles on the chart. These bubbles are arranged according to their Ad Meter rating on the Y-axis and their view count on the X-axis. Bubbles located in the top left corner suggest that a particular feature correlates with higher ratings yet attracts fewer views. Conversely, bubbles positioned in the bottom right corner indicate features that, while associated with a greater number of views, tend to receive lower ratings, showcasing a trade-off between viewership engagement and qualitative assessment.


The analysis depicted in the graphs above highlight a notable trend: advertisements with 30-39 shots are most effective at maximizing views. Conversely, for securing top USA Today Ad Meter scores, commercials with 50 or more shots lead the pack, suggesting that ads with a faster pace and more visual variety are favored in audience evaluations.

The data also reveals a specific trend for the number of people included in the ads: commercials featuring 10-19 individuals achieve the highest view counts. Interestingly, commercials that include a larger number of people generally receive higher Ad Meter scores, indicating a preference for ads with more diverse and crowded scenes among viewers. It is worth noting that there are 13 commercials that showcase 50 or more people. Yet, the Kia commercial is the only ad which uniquely combines over 50 shots and 50+ individuals.


Tracer tagging enables the integration of nuanced context into advertisements—details that AI might not yet detect but are crucial for business insights and analyses. Recognizing the significant impact celebrities often have in Super Bowl commercials, we delved into evaluating how their presence influences both viewership numbers and the USA Today Ad Meter scores.


It's not unexpected that commercials showcasing multiple celebrities achieved top performance in both viewership and Ad Meter ratings. Yet, it's particularly fascinating to note that ads devoid of any celebrities secured the second-highest viewership, surpassing those featuring only one celebrity. This insight suggests that when budget constraints restrict the inclusion of numerous celebrities, omitting celebrity endorsements altogether could be more beneficial than including just one or two. Conversely, if a celebrity presence is essential, opting for a single well-chosen celebrity might be the best strategy unless the budget allows for a more extravagant ensemble.

Number of Words Spoken, Tone, & General Feeling

In the final part of our investigation, we turned our attention to the scripts behind the ads, a process greatly simplified by Google Video Intelligence AI. We then utilized ChatGPT 4 to categorize the commercials according to their tone and general feeling. This step enabled us to delve into the emotional and thematic nuances of the ads, further enriching our understanding of their effectiveness and appeal.


As the graph above suggests, commercials containing 60-79 words strike the most effective balance, resonating well with audiences in terms of both views and Ad Meter scores. However, a notable finding is that ads with few words spoken tended to underperform in terms of both views and Ad Meter score. The cluster of data points to the left represents instances where fewer than 60 words were spoken, indicating that ads with minimal dialogue or narration may not capture audience attention or fare well in evaluations as effectively as their more verbose counterparts.


Utilizing ChatGPT 4 to analyze the tone and general feel of the commercials revealed several noteworthy insights. Ads with empowering and frustrated tones emerged as top performers in terms of views, highlighting their ability to strongly resonate emotionally with viewers. Additionally, a playful tone, while commonly employed, consistently proved to be both safe and effective. In contrast, ads adopting a serious tone tended to underperform in both view counts and ratings, suggesting that such a tone might not be as effective at connecting with viewers. Remarkably, ads characterized by an imaginative feel distinguished themselves by securing the highest view counts, suggesting that ads with a creative and imaginative approach have the potential to capture audience attention and engagement.

Concluding Thoughts

Our deep dive into Super Bowl 2024 ads, powered by the synergistic use of Google Video Intelligence AI and enriched by the insights from USA Today's Ad Meter ratings, underscores Tracer's unparalleled capability to transform complex datasets into digestible, actionable insights. Tracer's platform was instrumental in seamlessly integrating vast amounts of creative data with audience perception metrics, facilitating the creation of a dynamic dashboard that not only unravels the elements behind successful ads but also pioneers a new benchmark in advertising strategy optimization. By leveraging Tracer's advanced tagging and data mapping technologies, advertisers are equipped to make data-driven decisions with precision, ensuring their creative content captures attention, while also resonating deeply with audiences. At Tracer, we are not just enhancing the landscape of ad analysis—we are setting new standards for leveraging AI in marketing intelligence.