Last Updated on: 5th July 2022, 10:13 am
RTB House, a global company providing state-of-the-art retargeting technology for top advertisers worldwide, presents a new upgrade to its recommendation mechanism using a combination of deep learning and computer vision. The new method enables ultra-precise predictions of possible user’s buying needs, leading to product recommendations up to 41% more efficient than previously.
During a real-time bidding auction in personalized retargeting, computing occurs in very short time frames. The recommendation mechanism has only milliseconds to decide what to present on a creative. The decision of “what to display” is made on the basis of what the particular user was looking for, taking into account click data, information about the product, categories of interest, shopping behavior and search tactics. Because of time limitations, the mechanism has to act quickly to pick one creative from billions of combinations, presenting content that will be most interesting for a potential buyer.
RTB House has recently implemented new algorithms to its recommendation mechanism, enabling more accurate advice in the decision-making process. The final display is based on a full spectrum of information, which takes into account not only the referencing patterns made by other users with a similar buying profile, but also what was previously presented on creatives.
The new approach employs deep learning, the most promising subfield of AI research, which imitates the way the human brain works at solving problems. It makes decisions about what a user is most likely to click, browse, or buy. Without deep learning, it wouldn’t be possible to exploit dynamic personalization from multiple dimensions, not only based on standard recommendation systems, but on where products on banners are also chosen according to the user’s impressions history.
It uses technology referred also to computer vision, which allows for automated extraction, analysis and understanding of information from a single image or a sequence of images. It looks for similarities between products checked by potential buyers.
The end result is that recommendations become even further optimized, as shown by traffic and clicks made by RTB House clients. Users clicked on ads up to 41% more than usual with the new RTB House’s deep learning recommendation mechanism over current approach. Growth is noted especially in sectors such as: fashion and multi-category e-shops, where the possibilities to use cross-categories recommendations are almost endless.
Bartlomiej Romanski, Chief Technology Officer RTB House, notes that over the past few years, the industry has worked on tools that in some ways exceed the human intuition or eye’s capabilities. “Our goal is to make retargeting ads delighting customers on the one hand and performing extremely effectively on the other. The innovative recommendation mechanism we’ve implemented brings personalization to a new level. Thanks to deep learning our mechanism evolved to adeptly select products that should be shown on banners and have the biggest potential to be bought. In combination with computer vision we have the ability to analyze thousands of images per second, define patterns with a great precision and adjust recommendations to every small change in the customer’s behaviors. At the end of the day, higher performance brings our clients bigger return on ad spend and help to multiply ROI,” Romanski summarizes.
RTB House is one of few companies in the world that managed to develop and implement its own technology for purchasing advertisements in the RTB model (real-time bidding) – a solution in which buyers participate in real-time advertising space auctions. The company operates worldwide and runs more than 1,000 unique campaigns for global brands in more than 40 markets across Europe, Latin America, Asia and Pacific, Middle East and Africa. The RTB team is made up of over 200 people: performance marketing experts, analysts, sales and customer care specialists, programmers and others.
