Header Bidding vs Waterfall

Header bidding has become a practical tool for publishers in the innovative world of programmatic advertising due to its ability to maximize yield and create true competition for ad inventory. But before header bidding was developed, waterfall bidding dominated the programmatic advertising landscape. In the second installment of our header bidding series, we’ll explore the history of programmatic advertising to understand the groundwork that led to the development of these techniques. We’ll then compare header bidding vs waterfall to identify the strengths and weaknesses of both.

History of Programmatic Advertising

The basis for programmatic advertising took place in the late 1990s with direct, manual deals between publishers and advertisers. Publishers would normally sell their premium inventory directly to advertisers at agreed rates, while leftover inventory would go to ad networks at lower rates. This process relied fully on sales teams and manual insertion orders, taking up a lot of time and effort.

The introduction of ad exchanges marked the first step toward programmatic advertising. Unlike direct deals’ one-to-one approach, ad exchanges created a digital marketplace where all participants could buy and sell with each other. But these early exchanges still lacked real-time capabilities and sophisticated targeting.

Programmatic advertising as we know it today, truly began in the mid-2000s with the introduction of real-time bidding (RTB) technology. Real-time bidding created an opportunity for publishers to earn more per impression, especially for high-value audiences or peak times, as advertisers were willing to compete hard for specific ad placements. These publishers were now able to get a bigger revenue from individual impressions rather than relying on lower-paying bulk purchases from direct deals.

As RTB became more established, publishers needed a way to manage multiple demand sources efficiently. This is one of the reasons behind the development of waterfall bidding around 2009. Using the waterfall technique, publishers were able to prioritize their demand sources, typically based on historical performance. This meant that the highest-paying ad network would be called first, followed by the next, and so on.

Finally, one last big development took place with header bidding. Header bidding, introduced around 2014, allowed publishers to gather bids from multiple demand partners simultaneously before making a call to their ad server, eliminating the sequential constraints of waterfall bidding and potentially increasing overall ad revenue.

Now that we have a general overview of the history behind these processes, let’s take a closer look at waterfall bidding.

Waterfall Bidding

Around the year 2009, publishers found themselves with two different needs – having to manage multiple demand sources efficiently and also having to deal with remnant inventory that wasn’t sold via direct deals. And so, waterfall bidding (also known as daisy-chaining or waterfall tag) was born.

This process follows a hierarchical structure where ad impressions are sequentially offered to demand sources based on predefined priority levels, creating a waterfall effect. These demand sources are usually ranked by either their size or by the average historic yield they have produced for the publisher. This means that an ad network where premium inventory has been sold in the past (for a higher price) will then get first chance on further impressions from the same publisher. Typically, publishers lower their price floor with each hop down the waterfall, in hopes of finding the highest-paying ad in the shortest time possible.

In the waterfall bidding set up, the process is as follows: 

1. A user visits a webpage with ad space in it.
2. The waterfall bidding begins when the publisher’s ad server calls the first ad network in the sequence.
3. If the first ad network can’t fill the ad space, it passes on to the next network.
4. This process continues until the ad space is filled or all networks are exhausted.
5. The winning ad is displayed to the user.

The entire process typically takes between 800 to 1500 milliseconds.

There could be several reasons for why an ad network is not able to fill up the ad space:

• The publisher’s price floor might have been higher than the price offered by the demand source.
• The demand source was timed out because it took too long to reply to the ad request.
• There may not have been any relevant ads for the user and/or website based on the targeting criteria. For example, the website could be focused on traveling while the demand source only had ads related to sports. The ad networks receive information about the website and user from the publisher’s ad server, so they are able to decide whether the impression is worth filling.
• The advertiser purchasing through ad networks might have implemented impression capping, a common strategy to restrict the number of times their ads are displayed to the same individual within a specific timeframe.

If none of the demand sources are able to fill the impression, it remains unsold leading to a passback scenario, meaning that the ad request is sent back to the publisher’s default ad server. The publisher might then serve a fallback ad, such as a house ad (an ad promoting the publisher’s own content or services) or a PSA (public service announcement).

Benefits of Waterfall Bidding

Waterfall bidding helped publishers in a few different ways.

• Remnant Inventory: The main benefit of using waterfall bidding is the fact that it allows publishers to sell off leftover ad inventory that would have been wasted otherwise.
• Easy Implementation: Waterfalling is pretty easy to implement and requires very little technical knowledge. In order for it to work, the only thing needed to set up is a tag on an ad network.
• Management: Waterfall bidding is pretty easy to manage, making it a great option for smaller publishers or those with limited resources in technology. 
• Direct Deals: If a publisher has strong relationships with certain advertisers or ad networks, using  a waterfall setup to prioritize them can be a smart move.

Challenges of Waterfall Bidding

But waterfall bidding also brought with it some obstacles.

• Ranking: The predefined order of demand sources is often determined by the size of the sources or by the average yield in the past, rather than the amount they are willing to bid in a specific moment. This means that the ad revenue is not maximized.
• Loss of Opportunity: If the highest bidder was lower in the ranking, they would never get the opportunity to bid.
• Latency: The practice of passing an impression down the chain until it is purchased results in delayed ad loading times.
• Inventory Worth: Waterfall bidding results in publishers managing their ad inventory not based on the real value, but on an assumption of what an impression was worth. 

These obstacles deterred publishers from maximizing their ad revenue, had a negative impact on the UX, and created more work for the programmatic team. It is why header bidding came along as a natural progression.

Header Bidding

As we reviewed last week, header bidding came along after waterfall bidding and provided a much better way of working for publishers. 

The client-side header bidding process works in the following way:

1. A user lands on a site that has available ad space in it. 
2. The code on the website’s header sends out a request to multiple advertising demand partners with details about the user, the webpage they have landed on, and the ad space on the page.
3. Demand partners who are interested in the ad space respond with their bids. 
4. These bids are gathered and forwarded to the site’s ad server. 
5. The ad server evaluates the data from the bids to pick the highest one, deciding which ad to display. 
6. Finally, the chosen ad is rendered on the site and the impression is logged.

Though it may sound lengthy, this entire process is very fast – around 400 to 800 milliseconds on a desktop computer and 800 to 1200 milliseconds on a mobile device!

Benefits of Header Bidding

Header bidding brings many benefits to publishers.

• Increased revenue: Increased competition maximizes the potential ad monetization.
• Transparency: Publishers are able to control which demand partners can bid for their inventory, they can also prioritize some advertisers over others, and they can even prevent certain advertisers from taking part in the auction.
• Improved UX and SEO Results: Server-side header bidding helps with latency issues.
• More Insights: Header bidding gives publishers a lot of data, which helps them make decisions regarding their website and their inventory.

Challenges of Header Bidding

Although header bidding revolutionized the world of programmatic advertising, it does have its challenges.

• Technology: Implementing header bidding is more complicated when it comes to technology, as every SSP/ad exchange/ad network needs to be added into the publisher’s site’s code.
• Duplicate Bidding: An advertiser connected to multiple ad networks might compete against itself without even knowing so, potentially harming performance and affecting the amount of data exchanged within the ecosystem.
• User Identification Challenges: Server-side header bidding often struggles with matching users accurately, as much of the data is stripped away during server processing.
• Browser Compatibility: Client-side header bidding operates directly within the browser, which requires it to support a variety of browsers, including older versions. This can lead to technical difficulties. Moreover, some browsers restrict or block external pixel connections, reducing the effectiveness of header bidding auctions. 

Key Differences Between Header Bidding and Waterfall Model

Header bidding and waterfall bidding differ significantly in how they work, impacting everything from ad revenue optimization to user experience. Here’s a breakdown of the key distinctions.

1. Auction Process: In the waterfall model, ad impressions are offered to demand sources sequentially, following a predefined priority order. In contrast, header bidding allows all demand partners to bid simultaneously, ensuring the highest bid wins regardless of rank or past performance.
2. Revenue Optimization: The waterfall model often results in missed revenue opportunities because impressions may not reach the highest bidder if they are lower in the chain. Header bidding maximizes revenue potential by fostering real-time competition among all demand sources, ensuring impressions are sold to the highest bidder.
3. Set-Up: Waterfall bidding is easy to implement, making it an accessible option for smaller publishers with limited resources. On the other hand, header bidding is more complex to set up, requiring technical expertise and integration with multiple demand partners.
4. Ad Inventory: The waterfall model is often used for remnant inventory that hasn’t been sold through direct deals. Header bidding, however, allows publishers to monetize all inventory, not just leftovers, ensuring maximum yield.
5. Latency: Due to its sequential nature, waterfall bidding introduces high latency as impressions move down the chain, negatively affecting page load times.. Header bidding, especially server-side, reduces latency significantly by processing bids simultaneously, leading to faster page rendering and a smoother user experience.. This is crucial for maintaining strong Google Core Web Vitals scores, particularly Largest Contentful Paint (LCP) and Interaction to Next Paint (INP), which are key metrics for search ranking and user satisfaction.
6. Fill Rate: In the waterfall model, the fill rate decreases when no demand partner purchases the impression. Header bidding’s increased competition results in higher fill rates, as more demand partners vie for the same inventory.
7. Performance Control: Waterfall bidding offers little to no ability to evaluate ad performance, leaving publishers with minimal insights. Header bidding provides detailed reporting data, enabling publishers to analyze and optimize their ad performance effectively. Furthermore, the waterfall model often suffers from high ad discrepancies due to inefficiencies in tracking and reporting. Header bidding reduces these discrepancies significantly by providing more accurate, real-time data.
8. Demand Partners: Waterfall bidding is limited to a small number of demand partners, restricting competition and ad revenue potential. Header bidding accommodates multiple demand partners simultaneously, further enhancing competition and driving better results.

Conclusion

In the debate of header bidding vs waterfall, the former clearly provides advantages in most aspects, but waterfall bidding does provide unique strengths. By understanding the distinct value proposition of each technique, publishers can decide which best suits their needs.

As the industry evolves from waterfall bidding to header bidding and beyond, platforms like Opti Digital continue to drive innovation, offering sophisticated tools that help publishers maximize ad revenue while providing a seamless integration and detailed real-time analysis. Please get in touch if you would like to know more about Opti Digital’s revenue optimization technologies.