A deep dive into the role computer vision plays in the valuation and measurement of sports sponsorship

Sports Sponsorship is one of the most lucrative industries in the world with an estimated $501.4 billion being invested in it in 2022. But how accurately is the return on these enormous investments being measured?

The ROI Measurement Challenge

A big challenge for sports sponsors is the fact that ROI measurement tools have not evolved as quickly as their counterparts in other sectors. The chasm in reporting becomes obvious when you look at things like web, steaming and advertising analytics where a brand can precisely identify a number of key metrics, often in real-time. This includes exact impressions, CPM CPC, share voice, touch points, demographics, psychographics and final acquisition. For a number of reasons, this integrated approach doesn’t exist in the sponsorship analysis and valuation sector. Stakeholders must therefore pull data from multiple sources and ultimately make references from what they see. This is not ideal or accurate. 

Studies have show that: 

• There was an average efficiency level of less 0.3, meaning that on average sponsors were spending 70% more than needed in order to achieve the same results. 
• Only 12.5% of the investigated sponsorships were fully efficient 
• 39% of fans will watch non-live content 
• 47% of people who watch live sports simultaneously interact with other live content 

In short, those managing and evaluating sponsorships are not maximising their investments by analysing all the relevant data accurately. 

Lack of Core Accuracy

This issue cannot simply be attributed to a lack of platforms that can bring all the key metrics into one portal. Even if this portal existed today there is a more fundamental issue: the lack of accuracy in measuring impressions. 

Digital ads are binary, they are shown or they are not shown, therefore measuring impressions is pretty cut and dry. Sports footage, on the other hand, is fluid. Player movements and camera coverage is unpredictable, speed of movement changes, and viewpoints differ. This all means that impressions are not easily determinable without delayed analysis.

What should brands be measuring?

Computer Vision For Sponsorship Monitoring

Computer Vision is a branch of artificial intelligence (AI), and specifically machine learning, dedicated to the extraction of meaningful information from unstructured visual sources (images, video and live streams). The extracted data is then categorised into structured classes, allowing insights to be identified and to take actions/make recommendations based on those insights. In simple terms, if AI is the equivalent of a computerised brain, then Visual-AI adds eyes to that brain, allowing computers to see, observe and understand the visual world around us this likeness to human vision is simultaneously less advanced and more advanced than human capabilities. Although humans have a lifetime of learning that allows us to identify a massive range of objects. We also have the ability to learn new objects from a single exposure and once learned we can inherently identify those objects even if they are significantly different to the original ‘training data’ and even if they are at an odd angle, distorted or significantly obscured. This is not the case for computer vision.

These systems must be trained using extensive training data, although more modern systems can significantly reduce or even eliminate this need, allowing it to be more human-like in its learning. However, traditionally training sets would contain many different examples of an object in different positions, sizes, colours, etc. Either way, the outcome is a model that can be used to detect that object in subsequently submitted media to be analyzed. Visual-AI systems can also suffer from relatively high levels of false positives/negatives and these must be checked and corrected by humans. Modern systems can also mitigate this need for human intervention, drastically reducing the number of missed/false detections.

However, Visual-AI surpasses human capabilities once the model is trained and tuned because it is a computer system, which can work at orders of magnitude faster than any human is capable of. So where a human, working at peak capacity, may be able to review ~1,500 images in a best case working day (8 hours), a computer vision system would process in excess of 30,000 images in the same time period. However, unlike humans, computer vision doesn’t suffer from fatigue or boredom, it doesn’t need to take breaks and it can work 24/7.

This means that an average system could process ~100,000 images in a 24 hour period. The other key difference is the ability to add computing resources in order to meet whatever processing demands are required. This effectively means that there is no upper limit to how much a computer vision system can process, making it infinitely more scalable and efficient than a human workforce The process of analyzing media is facilitated via API calls that deliver the content to be processed plus instructions for the technologies to be used. Once processed, the extracted data is returned in JSON or similar file format, which allows the receiving software/platform to integrate it into their own dashboards.

A summary of the challenges of computer vision

Computer Vision is an amazing technology that has revolutionised many industries, but it presents many challenges, which are amplified by the choices that leading providers of these technologies make. These choices are made in order to focus their offering on the needs of different sectors, and so not all computer vision systems are equal for every application.

When it comes to sports sponsorship monitoring there are a number of specific challenges that a computer vision system must meet in order to be fit for purpose – as follows:

Conclusion

This document has discussed the current state of the sports sponsorship monitoring and valuation sector, highlighting the gaps in the capabilities as these platforms lag behind other digital marketing, brand monitoring and advertising focused platforms.

It has reviewed the broad range of metrics across multiple paid and earned channels required for a rounded platform and shown the important core role of impressions data in sponsorship monitoring. It outlined the many metrics that must be measured to accurately report on impressions, including time-on-screen, size, prominence, clarity, visibility and share-of-voice.

Finally, it has shown the critical role of computer vision in surfacing insights from visual media as the foundation of determining exposure value. But these systems must meet key needs, such as flexible brand libraries with instant logo learning, placement recognition, real-time and

full frame rate processing, as well as cloud and on-device deployment options. As well as meeting these requirements it has highlighted challenges that platform providers must ensure their chosen computer vision can meet to be successful. The right computer vision system will sit at the heart of any successful sports sponsorship valuation platform, while the wrong one will cause loss of subscribers through user friction and mistrust of the data.

About VISUA

VISUA develops best-in-class, enterprise Visual-AI that powers the world’s leading sponsorship monitoring, brand monitoring, brand protection and authentication platforms. VISUA delivers solutions ranging from logo/mark detection and counterfeit product detection to holographic authentication and ad detection. Its Visual-AI technology is proven to deliver the highest precision with instant learning, at unlimited scale, and is adaptable for any use case. VISUA believes in People-First AI, we see a world where Visual-AI will lift humanity out of the mundane, empowering a society that focuses more on creativity and collaboration and less on binary tasks, and empowering services and solutions that humans alone simply can’t deliver.