Deep Learning Brings Evolution To Advertising

Recently, conversations on artificial intelligence (AI) have exploded into the tech scene, with frontrunners like Mark Zuckerberg participating in the debate. Not only Facebook’s creator believes that working with AI is a chance for us to live fuller, more comfortable lives. Also such global giants like Google or Microsoft have already deployed new AI technologies inside their advertising systems, projected to boost ad revenue.The case study of RTB House, one of the first retargeters that extensively uses deep learning (currently the most promising subfield of AI-oriented research) in every stage of its technology, shows that a new approach has stepped into the advertising industry for good, changing the way advertising work, as well as the results that can be delivered.

20 years have passed since IBM’s supercomputer Deep Blue defeated world chess champion Gary Kasparov, in a historical first victory for AI. But what was once a futuristic concept in 1997, is now part of our daily reality. Today’s computers can not only imitate the way human brains work, but also solve problems in a quicker, more efficient way than humans do.

There are many ways to use AI not only in science and tech, but in every conceivable industry. In advertising, AI-based solutions are becoming a ‘must-have’ especially with enormous amounts of data and cutthroat competition. Super-advanced algorithms have become the answer to a marketer’s biggest challenge – getting in-depth, machine-interpretable information about customer’s buying potential and investing in them wisely.

RTB House has recently implemented deep learning algorithms, which makes its solution tackling three of the main challenges every advertiser faces at every stage of carried campaigns.

Challenge #1: Accurate Estimation of User’s Value
Potential customers have a different value every time they visit an e-shop. It depends on a variety of personal factors: the products they desire, their momentum of buying, or even customer lifetime cycle. The question that hangs over every marketer’s mind becomes, “How does one estimate and measure buying potential?” This has always been one of the biggest marketers’ challenges, as the answer directly correlates with how much they should invest in a customer. The best prediction is one that can correctly estimate the ‘possibility of making a purchase’ and ‘potential value of user’s basket’. These two parameters ultimately define the return on ad spend.

The goal of RTB House was to optimize the process of buying ads in a way that attracted only users who are most valuable to clients and bid higher for them automatically.By implementing deep learning algorithms, the company’s conversion rate and value mechanism started to collect and interpret not only click data (the typical approach), but it began to take into consideration how users browse particular offers, comparing categories of interest, basket behaviors or search tactics to have a broader picture of every individual’s buying potential. With this broader spectrum of information, RTB House was able to increase overall performance from retargeting activities powered by deep learning up to 29 percent.

Challenge #2: Ultra-precise Estimation of Customers’ Desires (Recommendations)
A recurring challenge in digital advertising is what to show on banners to maximize the potential of purchase. The closer you are to presenting what users desire, the larger the chance of finalizing a purchase.
During a real-time bidding auction in personalized retargeting, a recommendation mechanism has only milliseconds to pick one creative from billions of combinations and decide what to present on a banner. By using deep learning algorithms and computer vision, RTB House displays ads with decision-making that not only takes into account referencing patterns made by other users with a similar buying profile, but also what was presented on creatives and a specific user’s reaction to previous offers. By analyzing click data, information about the product, categories of interest, shopping behavior and search tactics, product recommendation efficiency increased by up to 41 percent, compared to campaigns that did not utilize a deep learning mechanism.

Challenge #3: In-Depth Assessment of Click Through Rates
A common success metric of an ad campaign is the click-through rate (CTR), so the ratio of clicks on a banner against the total number of impressions. The objective of every marketer is simple – draw in potential buyers by offering the most click-worthy creatives.

Deep learning algorithms implemented by RTB House choose which banners will be the most effective in given placement and display it to the users. This better prediction of CTR, ultimately yields better ROIs for customers.
By using deep learning algorithms, RTB House was able to boost total CTRs by 16 percent more clicks within the same budget limitations for its clients.

Whether you buy into the ‘Doomsday scenarios’ of AI taking over the world or AI as the salvation of human comfort and ease, it’s clear that artificial intelligence will be the defining technology of the near future. RTB House is an example of innovators exclusively using deep learning and a proof that what was a bit sci-fi just a few years ago seems to be now a natural process to make online activities more efficient than ever before.It means the evolution in advertising, from the perspective of both – advertisers and users is especially visible in such highly-demanding industries like travel where a long list of metrics are taken into consideration and user shopping behaviors are difficult to predict. And that is just a beginning. By having more information-fueled and highly-efficient technology, possibilities are simply endless.

Add Comment