Login is restricted to DCN Publisher Members. If you are a DCN Member and don't have an account, register here.

Digital Content Next


InContext / An inside look at the business of digital content

How machine learning improves header bidding

October 15, 2020 | By Ankit Oberoi, CEO & Co-Founder – AdPushup @oberoiankit

If you have worked in the digital space in the past few years, you’ve likely encountered the term Machine Learning (ML) in some context. For those of us who aren’t tech experts, this buzzword may be a little daunting. However, these technologies are actually widely used across industries including digital advertising.

So, what exactly is Machine Learning? To explain that, let’s take a step back and talk about Artificial Intelligence (AI). AI describes a computer program’s ability to conduct a task that would normally require human intelligence, such as drawing conclusions based on data or recognizing patterns in a dataset. Machine learning is a step above this. A subset of AI, ML is the ability of programs to “learn” and self-improve as they get more data. This allows them to make more accurate decisions and predictions over time. This greatly reduces the human input required in performing crucial tasks.

How is machine learning positively impacting header bidding

Many publishers across the world rely on header bidding as the most efficient way to boost programmatic revenue. However, the technique in itself has undergone significant advancements. Initially, the majority of publishers used client-side header bidding to maximize yield through multiple demand partners. But it had its own challenges such as latency. 

Then, server-side header bidding arrived. It helped publishers combat latency issues but had its own shortcomings. Publishers now use a combination of both, which is generally known as hybrid header bidding. 

Among all of these advancements, one thing has remained constant: the use of machine learning in improving header bidding for all publishers. Therefore, the future of header bidding relies on how well technology companies can incorporate ML into improving header bidding tech. This will ultimately yield a robust solution that does not require any workarounds for its peripheral issues. 

Impact of machine learning on header bidding dynamics

Automatic selection of demand partners

One of the key challenges that publishers dealt with before heading bidding was a lack of control over demand partners. That is key to optimizing revenue. Without knowledge of demand partners, the entire system would lack critical transparency. 

Through the use of ML in header bidding, publishers get access to multiple demand partners. In fact, the selection process ois automatic, while simultaneously providing maximum yield. This massively benefits publishers as it saves them time for other activities while ensuring a consistent increase in revenue.

Timeout management 

For every publisher, managing timeouts is a tricky process. Timeouts can often result in lost revenue and publishers need to spend some time troubleshooting. However, ML has completely altered the game. Publishers can now use managed heading bidding solutions that utilize machine learning to fine-tune timeout settings in a way that maximizes yield but reduces latency.

Server-side header bidding

Server-side header bidding was introduced to reduce the latency issues associated with client-side header bidding. Instead of a user’s browser, the auction takes place on a server. Many publishers have gravitated towards a combination of client-side header bidding and this to get the maximum out of their investment. 

The credit however, goes to ML for helping the ad tech industry in advancing this technology. Machine learning helps in tracking deal performances through automated systems. All this is done in real-time with multiple demand partners in place.

Price floor optimization

Another added advantage with using ML in header bidding is optimizations of price floors. As mentioned above, ML is what makes algorithms smart. With a consistent tracking of data, patterns, trends, demand partners, ad requests, and related factors, ML can help in automated price floor optimization. Over time, this can considerably improve header bidding performance. 

Why machine learning is here to stay and scale

ML capacity to learn and improve is uncapped. This means that with time, as the programs amass and go through more data, their predictions will only become sharper, quicker, and more accurate. While ML may not replace humans anytime soon, these tools can definitely help marketing teams free up time to focus on high-level strategy, instead of repetitive data analytics.

Advertisers across the world are already witnessing increasing returns on their ad spends, thanks to ML-based optimization of everything from creatives, targeting, and bid values. However, advertisers are not the only ones who stand to benefit from the adoption of ML tools. Publishers typically receive a percentage of the revenue (through CPM, CPC or other models) and more effective advertising campaigns mean a chunk of a larger pie for publishers. 

However, publishers can by no means afford to be merely passive bystanders in the ML revolution. Thanks to the data at their disposal, advertisers, today, have a clearer picture of how, where and when their ads work best. In contextual advertising, for instance, publishers need to continually create content. However, they must also evolve based on their audience’s changing consumption patterns. This requires knowing your target audience deeply, and often working with advertisers to create a synergistic experience for the audience. Publishers, too, can leverage ML solutions for layout optimization to create greater value for advertisers, and by extension for their own platforms.

Overall, AI and ML tools offer an opportunity for all parties involved in advertising technology. Early adopters of these solutions will likely see the best outcomes and returns, and given that these tools are based on the premise of self-improvement over time, it is likely that those who adopt a wait-and-watch approach will face an uphill task of catching up.

About the author

Ankit is the CEO and Co-Founder of AdPushup, a leading technology company which helps web publishers and e-commerce companies accelerate their revenue growth. At AdPushup, he has led the product development, market strategy, and business development efforts, which has allowed the company to scale to 4 billion+ monthly impressions for over 300 publishers worldwide.

Liked this article?

Subscribe to the InContext newsletter to get insights like this delivered to your inbox every week.