How AI Can Save Ad Tech

Artificial Intelligence (AI) centered solutions are increasing in popularity in today’s ad ecosystem. And for good reason. With increasing demands for greater efficiency, AI has the power to transform Ad Tech and expand the traditional solutions available to solve complex problems and achieve full, autonomous advertising to improve advertisers’ ROI. In this article we will explore how AI can greatly improve numerous parts of the advertising world such as data ingestion, revenue optimization, creative optimization and the move away from cookies.

Autonomous Advertising With AI

In the not-so-far future, AI will be able to automate every step of the digital advertising process. Whether it is predicting, generating, or optimizing ads, AI will be a key component in scaling and improving digital advertising.

Wait But How?

Hundreds of programs across varying industries are already applying deep learning language models and computer vision to train AI to take in video, audio, images and text to analyze context.

While this might seem like sci-fi, Natural Language Processing (NLP) and computer vision have come a long way in improving AI, similar to how humans see, learn, and interact with their environment. Except AI can process infinitely more amounts of data than humans.

Now that the general democratization of machine-learning tools has made it easier for existing businesses to utilize AI, it is less a matter of if AI will become fully autonomous, but when.

Using Gobs Of Data And Processing Power For Greater Efficiency

Every company and brand on the buy side and sell side of Ad Tech ingest incomprehensible amounts of data every day. It is impossible for humans alone to analyze that data for useful purposes but through machine learning and AI, many companies are already figuring out how to apply that data to drastically improve parts of their business. At Sharethrough, for example, we have developed our own SmartSuite™ technology that leverages deep learning algorithms to help auto-enhance work processes from start to finish. This includes Smart Throttle and Smart Floors, among other features.

Smart Throttle

Smart Throttling examines each of our billions of daily impressions based on dimensions like location and user match to pair it with DSPs that are most likely to bid on it. By only sending the highest-value impressions to each DSP, we can avoid arbitrary queries-per-second (QPS) caps from DSPs and earn more revenue for publishers.

This has been so successful that we are in the process of further improving the sophistication of our AI to form a model that would enable us to analyze even more dimensions to throttle on like  device, time of day and seasonality.

Smart Floors

We also ingest billions of data points per day on bid prices and win rates, which also can alter depending on many factors such as the site, the end user, time of day, device type, etc. Our  Smart Floors feature uses machine learning to determine fair prices for the inventory to maximize yield. SmartSuite automatically experiments with different floors for publishers to find the optimal floor for each placement, every day.

This yields a significant average revenue improvement versus manually set floors and is only continuing to improve over time.

More Possibilities With Component Parts

More and more modern omnichannel ad exchanges are now operating with component parts. Component parts are essentially the metadata that is used to construct an ad, such as the headline, thumbnail image, banner, video, call to action, brand name, etc. By leveraging this technology, AI can be used to gather data via component parts to modify the metadata and find the best combinations to deliver optimal performance and custom ad experiences tailored to each user.

One Campaign, One Thousand Creatives

This increases the possibilities to build fluid designs (fonts, colors, layouts), calls to action, and bespoke thumbnail images and focus areas on images combined with the tried and true contextual and behavioral targeting will result in a new world of advertising where no one ad looks the same for any site, user, time of day, browsing mode.

As it was mentioned in one of our previous blog posts on why advertising has only been scratching the surface of AI, one of the strengths of AI is that it can become familiar with the creation of “derivatives” of an original creative concept; then move on to make automatic alterations and spin offs of that concept on its own. This would thereby alleviate the manual work creative teams would otherwise have to do and could in fact produce even better results.

AI & The Era of Uncertain Data

Data is the lifeblood of AI. In light of increasing privacy and data collection concerns continuing to grow, the change in tracking & targeting will simply result in a different set of data that AI can now optimize towards.

It is worth noting that AI does not have a direct correlation to the new era of tracking and targeting. The lack of this data will certainly mean existing models built around correlating a specific user’s interests and their probability to click or convert will no longer be changing over time.

AI will be critical in trying to bridge the gap now created by this old user-level data that will be going away; and it can achieve this by tying together disparate non-targeted datasets and finding correlations in near real-time that could result in efficiencies and improvements akin to targeted data. One promising way AI will bridge this gap is through more sophisticated contextual targeting.

Improving Contextual Targeting Through AI

Contextual targeting involves displaying ads based on a website's content. For example, placing an ad for dining chairs on an interior design website or an ad for running shoes on a running forum.

With AI, the process becomes even more powerful, enabling the analysis of languages, images, videos, and overall sentiment of the web page. AI is capable of turning all of this information, down to the colors and thematics of a series of images, into insightful data.

AI will also help expand contextual advertising beyond reading a webpage to also reading video content online and on CTV, audio signals from podcasts and music, even digital out of home DOOH). Instead of context only being limited to a video or audio's channel, AI can be used to interpret and identify the context of the actual content of the video or audio file. With the increase of personalization and new data from the proliferation of IoTs, AI can combine these data points to provide new contextual signals not just in the content but potentially in a user's surroundings and situations.

Thanks to AI, targeting will not involve tracking on a personal level, but rather automatically configuring the ads to complement the context of the page.

From what we know about AI & machine learning, the more data we have, the better it is trained to predict the outcome. AI could also expand the possibilities to automate A/B testing. In effect, advertising costs may decrease in the future to run more campaigns to find what is most optimal.

There’s Still Time

AI is already improving the sophistication of advertising technology and, while it is not currently able to achieve full autonomous advertising, AI is still making headway on optimizing contextual targeting to limit reliance on old user-level data, improving user engagement by enhancing the creative process, and maximizing the revenue generated for advertisers, publishers, agencies and brands.  

The progress and vision is very promising for the future of AI. Allowing for a performant advertising ecosystem that can address many of the challenges that face the advertising industry with great efficiency. The question now is what are the limits of how AI can solve advertising challenges and problems in the future? And what would the ads of a fully autonomous AI look like?