Machine learning is used to solve many problems related to optimizing processes and increasing advertising efficiency. Each advertising platform allows you to quickly analyze large amounts of data, create personalized offers, optimize targeting, predict user behavior, prevent fraud and generally make advertising campaigns more effective. However, it is of great importance for advertisers to achieve maximum conversions (purchases) within the price they are willing to pay, depending on business economics. While this might be the case in the ideal world, in reality something can go wrong.Advertisers have become accustomed to viewing all advertising algorithms as a black box and that they must pay the price for algorithm errors and inadequacies. Typically, various ad platforms say it takes two to four weeks to train their algorithms. This process becomes a problem for advertisers because they have to spend their time and money training the advertising system before getting conversions at the right cost.
Additionally, if campaign settings or the targeted return on ad spend (ROAS) or cost per action (CPA) have changed, additional time may likely be required to collect the required set of data with the changed inputs. Therefore, the estimate is recalculated. Cost-per-click (CPC) based payment model; It can force advertisers to pay for ad platform errors and sponsor learning periods.
Stating that the algorithms in Yandex’s e-commerce advertising product are designed to not wait for the first clicks and conversions, but to predict conversions instantly and accurately based on available information, Yandex Ads International Product and Growth President Elmira Agzamova made a statement on the subject:
“Due to the use of machine learning (ML) algorithms and specific data from ecommerce, our product does this instantly by learning at the viewing stage to predict what conversions will be. In this way, the advertiser can get instant conversion and pay the price he wants, instead of spending weeks waiting for the system to collect data and learn. The key to the speed and quality of predictions lies in the fact that the ML model learns from large amounts of heterogeneous data, and this is not the only data on conversions in the ad system. “Our product also takes into account data about users’ interests and business behavior on the Internet, the website on which ads are displayed, the pages of the online store and the specific product as well as the advertising banner and its characteristics.”
Another problem with the PPC-based model is that it creates the intention to split the marketing budget for retargeting and user acquisition. Advertisers know that retargeting is something they can control, that they can get conversions at a lower price. However, user acquisition (UA) can sometimes bring bots to the website or traffic without any conversion. So advertisers are forced to pay for this traffic even though there is no real impact on their business.The differences between retargeting and user acquisition cannot be easily distinguished. For example, “If the user was on the website 180 days ago, is that retargeting or UA?”, “What if it was 365 days ago?”, “Are you sure your user acquisition campaigns don’t include a retargeting audience?”, “Is this retargeting or UA?” “When was the last time you checked?” Questions like these come to mind.
In fact, much more advanced than manual segmentation, Machine learning (ML) models take into account the diversity of specific e-commerce data as broadly and effectively as possible. ML is also used by advertisers to deliver their products to additional buyers.
It is also stated that the ML model takes into account users’ commercial profiles, including both short-term and long-term interests. While short-term data is defined as data about recently viewed products, for example; Long-term data, on the other hand, is considered information about recurring purchases made by users at a certain frequency.
The advertising system combines data on users’ characteristics and purchasing behavior, as well as data on the products they purchased or viewed in online stores on websites. By analyzing this information, a model based on machine learning algorithms can predict which ads and which products are most likely to convert for similar users. For example, when it comes to e-commerce, especially in the clothing industry, the image on the banner becomes much more important here. In general, the color, size, brand and other features of the products that play a big role in advertising.
ML algorithms also allow filtering out irrelevant or incorrect clicks. The model is not for one click; It makes a difference by optimizing for longer clicks, such as when the user takes a conscious action after visiting a site and clicks on another page. Yandex’s product sees this as a useful click. Thus, it eliminates accidental clicks and ad fraud. This allows companies to bring only target customers who are truly interested in purchasing to their websites.
This approach allows the advertiser to achieve maximum conversions with a fixed budget and target KPI, both through retargeting and by attracting new customers for existing products. But this only works if the advertising platform can guarantee business results for the advertiser.
Explaining how the system works, International Product and Growth President Elmira Agzamova said, “We believe that advertising platforms should be a part of the advertiser’s business. We want businesses to only care about what price they can pay for conversion. We can provide them with maximum conversion in the price range they want. Essentially, we’re always thinking about how we can be an off-site extension for our advertisers. For example, we created algorithms such as ‘Users similar to you bought these products’ or ‘who usually buy this product also buy these’, which are common on large e-commerce sites. “This business-focused approach allows us to guarantee prices to our customers and only get paid as revenue share when conversion occurs,” he said.
About Yandex Ads
Yandex Ads, part of the international information technologies company Yandex, is the leading advertising platform in Eastern Europe and the CIS. Yandex Ads offers a complete set of tools for effective marketing, promotion of products and services, monetization of websites and apps from ads, as well as detailed app analytics and retargeting solutions. Our products enable data-driven marketing decisions and use artificial intelligence technologies to create advertising.
For details: ads.yandex.com/welcome_tr
27601
