5 Struggles of Real-Time Bidding and How to Solve Them
The reasons why your marketing budget might not be spent optimally and how to change it
Online advertising is a thriving field considering that more and more time is spent online. The importance of online advertising multiplies as in 2020, it is estimated that worldwide daily internet consumption will overtake the consumption of TV [1]. Regardless of the marketing strategy and campaign type, the goal of advertising is ultimately to draw people’s attention to a service or a product and increase the number of visits and engagements — tracking points on the marketer’s website. As the entire internet grows larger every second, its diversity increases. There’s a vast variety of topics, website designs, number of visitors per day, bounce rate, length of visitor sessions, etc. Due to this fact, it becomes essential to choose which sites you want to place your ads on and don’t spend your budget uniformly on all websites. In this article, I will describe the criteria based on which you can distinguish websites and how to choose the ones you want to advertise on.
This is the second article of our series dedicated to digital advertising. In this article, we focus on Real-Time Bidding (RTB) as it is the only fully automated way of digital advertising and, therefore, it can be optimized and augmented using machine learning techniques. We’ll describe the issues and struggles that come with it, together with ways of improving them with the use of AI. If you’d like to find out more about the ways how online ads can be bought, what real-time bidding is, how it works and why it’s perfectly suitable for machine learning applications, read the previous article of this series.
What Are the Struggles and Where Do They Come From?
RTB is a programmatic and fully automated way of buying ad banners via an auctioning system. By default, RTB applies uniform bids across the entire internet, which means that the price of the bid is the same for every web page and every user who is auctioned. In AdForm and similar advertising platforms, it is possible to bid only for users from given countries or AB test your campaigns based on the user’s postcode, but that’s basically it regarding the ways of targeting. What it means, in the end, is that you pay the same amount of money for ads of different visibility, audiences, and many times, the user doesn’t even scroll down to see the ad you’ve paid for. But is there a possibility to improve it instead of just admitting it as a fact? Of course, there is, let’s take a deeper look at them!
1. Ad Visibility
Most likely, the worst money spent is on ads that nobody can see. Some websites often offer ad spots at the bottom of articles, which most people just skim and close the page before reaching its end, whereas other websites offer spots at the beginning of the page or in sidebars that go down as you scroll. Other websites are only click-through pages (e.g. category disambiguation) where people spend no more than one second on average. On these, even if you have bought a super nice and huge banner, it most likely has little to zero impact. These types of websites are the ones you don’t want to buy ads on, or at least you’d like to buy the ads for less.
2. Ad Targeting
Do you know what your target audience is, yet your ad is shown to everyone, including people that are not interested in your products and are out of the target group? RTB in its raw form targets all people without differentiating them, therefore, you pay the same amount for an ad for users both in and out of your target group. In many cases, you might want to target only a group of people that share the same interest, for example, people that like to work out or people interested in buying a new car. In these circumstances, it might be more efficient to use the same budget on ads only for people who are likely to be interested in your product.
3. Ad Quality
No two websites are the same. They vary in the average banner
visibility ratio, number of banners per page, ratio of robot visits, and
many more aspects. Even when you have a big, visible banner at the top of a page, if the website is filled with other adverts, your ad will get lost in the sea of all the banners, and the user might even get annoyed by the overwhelming marketing content. And if your ad is bought for a robot user, the impact of your ad most definitely equals zero.
The overall “quality” of an ad can be assessed by multiple factors:
- for how long the ad was visible
- what percentage of the ad was seen
- how many ads were on the page
- whether the ad was interacted with in any way (hovered over, clicked)
- whether the banner was clicked through or not
- whether the ad was shown to a robot or not
All these characteristics can be used to evaluate how much was a bought ad worth. Altogether, they represent the impact your ad had and remember that without any enhancements, you pay the same price for ads of all different qualities, and therefore, impacts.
4. Ad Engagement (Click-Through Rate)
Recently, native campaigns have become more and more popular. As these advertisements are displayed usually on social networks among the organic content, it’s difficult to distinguish between ads and the non-payed content. In our opinion, for native ads, the visibility time might not be the most impactful metric to follow since these banners are usually seen as a whole as people scroll down through the news feed page. That’s why for some campaigns, and native campaigns especially, we suggest optimizing on the click-through rate instead of visibility time or ratio.
5. Inappropriate Content
Your ads can be sometimes displayed on websites with inappropriate or sensitive content, such as news about war, terrorism, child abuse, fake news, etc. This can have a negative impact on people’s awareness of your business. Hence, it is important to “keep an eye” on the content that people link to your brand and strive to place adverts on websites with positive and neutral content.
How to Approach These Issues?
In every case mentioned above, you pay the same price regardless of how good the ad is according to the given aspect, which can lead to your budget not being spent effectively.
Our AI-driven product AdPicker can help you avoid all of these problems. AdPicker features multiple solutions, each of which is focused on the optimization of one aspect of online advertising. Depending on your needs, we help you choose the best solution for your specific case.
The strength of all AdPicker solutions comes from machine-learning models. We take historical data about impressions and tracking points, process them and extract features from them (such as click-through rate, number of banners on the website, visibility time, etc.). These features then serve as an input for the training of a machine-learning model. Based on the specific solution, the model can be XGBoost, Linear Regression, a complex mathematical function, or the best-performing model found using cross-validation. The final model is chosen, registered and used to produce an outcome that is directly used in an advertising platform (e.g. AdForm) and impacts future spending.
Based on the type of model outcome, AdPicker solutions can be divided into two groups:
1. Domain-Based Solutions
The first type of solutions is domain-based. In these solutions, the model output is in the form of a domain whitelist — a large pool of domains together with their corresponding bid multipliers. These multipliers are then used in the RTB auctions and express how much you are willing to pay for an impression on a particular domain based on its characteristics. That means your spending will decrease on pages with worse attributes and increase on higher quality or relevant ones.
2. Cookie-Based Solutions
The second group of our solutions is cookie-based, which means that the model outcome is a list of cookies with corresponding bid multipliers. These multipliers are used in the RTB auction in the same way as described above. In this case, they represent how likely will the cookie be interested in your ads and products.
How Do Our Solutions Help?
Based on the principle of how our solutions work, we can look at improving the budget spending from two perspectives:
- it helps by increasing the number of ads bought on better websites or for people likely interested in your products, and
- it adjusts the price offered for an ad according to its predicted impact and “quality”.
As a result, you can buy more ads with the same budget and target these ads on pages with relevant topics and quality banner spots at the same time.
Final Words
Right now you should be familiar with the main drawbacks of using default RTB without enhancements and how can these issues be solved. If you think that your budget could be spent more efficiently too, feel free to reach out to us!
Related Sources and Further Reading
[1] https://nmc-mic.ca/2018/12/06/is-tvs-reign-nearing-its-end/
Thank you for reading up to this point. If you have any further questions or suggestions, feel free to leave a response. Also, if you have a topic in mind that you would like us to cover in future posts, let us know in the comments below.