How digital advertising can benefit from the growth of AI

11/21 By IAS Insider

Less than a century has elapsed since the idea of artificial intelligence began to stir excitement in the field of computing. In this short time artificial intelligence and machine learning have taken off, spilling over the walls of academia and into everyday life. From personal banking to online shopping, machine learning is a very real and concrete part of our lives, directly impacting the way we interact with our environments, both off and on screens.

Digital advertising has evolved dramatically thanks to artificial intelligence and machine learning. These technologies have made way for automation and efficiencies, replacing archaic manual methods with unparalleled speed in data processing, and decision making.

Integral Ad Science invested in artificial intelligence and machine learning early on and continue to do so, enabling ever larger scale, safety, and efficacy of digital advertising.

Bringing brand safety to the masses

As the digital advertising industry grew, advertisers started asking, What kind of inventory is out there, and how do I avoid placements that don’t align with my brand?  IAS answered with our brand safety solution.

The most natural and immediate way to solve the problem of inventory quality was to take stock and construct lists of inappropriate websites. Doing so would allow advertisers to avoid objectionable content such as pornography, hate speech, violence, etc. However, IAS recognized that this was an unscalable approach to a global issue. With hundreds of millions of new websites launching each day, it was clear that identifying the content, categorizing it, and then blocking campaigns from running on these placements all needed to happen with speed and at scale.

At IAS we employ artificial intelligence in order to automate our brand safety solution. Through machine learning informed by our extensive data science team, our technologies are able to continually improve their understanding of the digital landscape at scale, and are capable of automatically deeming new pages as inappropriate without having to be explicitly programmed to do so. While AI gives scale to our brand safety offerings, we continuously audit and enhance our models to keep up with the ever shifting landscape of risky content.  

Ad Fraud: the digital arms race

In 2015 Ernst & Young reported that invalid traffic cost the U.S digital marketing, advertising, and media industry an estimated $4.6 billion dollars annually. Identifying new bots, analyzing their operation, and designing techniques to catch them is a highly-specialized, time-consuming skill, and the need to do so has never been more apparent.

IAS understood the threat of ad fraud early on, and as fraud grew IAS began to utilize big data and AI to identify fraud by discerning patterns in fraudulent activity. Machine learning models recognize bots and watch 24/7 to adapt to changes in their behavior, for instance bots spend on average 2x the amount of time than humans viewing individual ads – this type of information allows IAS to respond immediately to new threats. AI also prevents us from erroneously classifying some human traffic as fraudulent. For example, instead of blindly classifying activity that comes from a compromised IP as bot traffic, our models are capable of differentiating between IVT and human traffic coming from a single IP.

Programmatic

As advertisers have become conscious of how their campaigns perform relative to important metrics such as viewability, fraud and brand safety rates, they realize that their campaigns tend to score poorly on these metrics unless they take proactive measures.  They have increasingly been looking for solutions to optimize their campaigns to ensure they meet and outperform their goals.  Leveraging the billions of measurements we make on a daily basis, IAS utilizes machine learning that can help advertisers do just that.  

Our machine learning models continuously assess inventory quality based on all the important metrics.  During the programmatic RTB process, our technology enables bids to be approved or rejected based on the judgments of our machine learning models.  With the solutions we provide, advertisers can be confident that their spend is going towards reaching the audiences they want, which in turn will help them meet the goals they set out before spending a single penny.

Publisher Optimization: leveraging machine learning for immediate results

With the advertising ecosystem calling for transparency, publishers are under increasing pressure to deliver quality inventory that is aligned with advertiser expectations. The IAS Publisher Optimization Tool leverages IAS’s proprietary machine learning technology, to allow publishers to prevent over delivery and impression waste on viewability campaigns and customize brand safety and IVT goals based on specific advertiser requirements.

Publisher Optimization predicts an ad impression’s viewability, brand safety, and fraud levels before it’s sold with low latency and high throughput. This platform encompasses continuously trained, tested, and deployed machine learning models and leverages a state-of-the-art technology stack and specialized algorithms.

Looking ahead

As we look ahead, we see the limitless potential for artificial intelligence and machine learning in digital advertising. Current advancements are blurring the lines between human and machine as evident in sentiment analysis – the computational  identification and categorization of opinions expressed in a piece of text, in order to determine whether the writer’s attitude towards a particular topic or product is positive, negative, or neutral. Rapid advances in deep learning are allowing computers to process images and video in a way that’s consistent with human understanding at a un precedented scale. These advancements in computer vision enable a more robust brand safety solutions with  less impression waste, and an improved consumer experience.

The increased investment in and adoption of artificial intelligence and machine learning can bring us closer to not only protecting advertiser dollars but can aid us in intimately understanding consumers, ultimately changing the way we target, reach, and interact with them.