- App Monetization
- Header Bidding
- iOS Ad Mediation
- App Monetization
- Header Bidding
Previously, programmatic ad buying was mainly mediated using the ad server waterfall architecture, but since the arrival of the header bidding approach, a lot has changed. How do stakeholders such as publishers and ad networks in the industry leverage both technologies for an incremental revenue drive?
Although a number of advertisers are yet to fully understand the mechanism of programmatic ad buying, it has brought about significant impacts in the digital/mobile advertising industry. For instance, publishers now enjoy better and more rewarding interactions with demand sources.
In recent times, more publishers are transitioning to header bidding. A 2019 survey published by Statista showed that 76 percent of publishers sampled now use header bidding as part of their ad tech stack. Since the benefits are very much obvious, mobile and in-app publishers have also joined in on the train.
Many years ago, before the advent of digital media, publishers interacted directly with advertisers who were looking for spaces to market their brands, products, or services. The era of digitization seriously disrupted this pattern particularly because some ad spaces were never sold, giving rise to remnant inventory.
Ad networks stepped in, and among other things, brokered the aggregation of remnant inventory from publishers (website owners and app publishers), and sold off the inventory to advertisers (demand sources) via ad exchanges. Therefore, after direct bids, publishers used ad networks to sell their remaining ad inventory. But again, this pattern didn`t handle all the ad-buying needs efficiently. There were more ad networks and it was difficult knowing which one would yield more revenue.
So, new ad tech platforms rose to the occasion to offer more dynamic solutions in the programmatic ad buying ecosystem. They are known as supply-side platforms. Although several other technologies have emerged and made the ad tech ecosystem more sophisticated, they all have generally simplified many processes (e.g. real-time bidding and header bidding) and improved yield for each party involved in ad operations. Real-time bidding (RTB) basically involves auctioning off available ad inventory in real-time – to demand-side platforms.
In real-time bidding (RTB) exchanges, publishers use supply-side platforms (SSPs) to make their inventory available for purchase by advertisers. Advertisers, also known as demand sources, then use a demand-side platform (DSP) to bid on the inventory (also called impressions) exposed by publishers. An ad exchange serves as a platform or marketplace for this exchange to happen.
Ad exchanges provide a means for publishers to connect to advertisers who need ad spaces. Both SSPs and DSPs are powered by similar technologies but they are different in the purposes they serve. In addition, these platforms can be used to manage ad inventory.
Ad networks, on the other hand, function as aggregators of ad inventory, serving as intermediaries (companies) who host ad servers and broker the selling and buying of ad inventory between publishers and advertisers. A common simple analogy is to look at ad networks as stockbrokers and ad exchanges as the floor of the stock market where the trading takes place.
The Waterfall model is an ad selling model in which the ad request cascades down from one ad source to the next in sequential order until it gets to the ad network that is willing to fill the request. The downside to the model is the inability of publishers to earn premium or even winning bid rates on their ad requests most times.
In the waterfall architecture, each impression is offered for sale sequentially to each ad buyers – one at a time. The waterfall is set up on the ad server and is simply an ordered list of buyers which is traversed.
If the first buyer in the sequence of buyers rejects the offer, it automatically goes to the next buyer down the waterfall sequence and so on, until some ad source fills the impression. Each ad source might be an SSP or an ad network, and internally connect the publisher’s inventory to as many exchanges, DSPs, and networks as possible, thereby increasing the chances of filling the available inventory at the specified price floor.
Advertisers on the other hand, use DSPs to analyze and purchase impressions released by publishers as cheaply and as efficiently as possible. In practice, a major demerit of the waterfall architecture is that publishers’ inventories are devalued as they get passed down from one buyer to another.
Because of this loophole, most publishers believe that selling their ad impressions using waterfall limits profitability. In addition, some inventory may remain unsold completely or sold after a number of unsuccessful attempts which leads to a high latency – again leaving money on the table. Advertisers who buy these remnant impressions low in the waterfall are not willing to pay high prices for them.
Popular ad servers include Google Ad Manager, PubMatic, Rubicon, and AppNexus. These ad servers typically allow publishers to set up their inventory, targeting, and delivery. They also facilitate the setup of a waterfall with several ad sources that are willing to pay for the publishers’ inventory. In essence, the ad server is between the publisher’s app and supply-side platforms.
A major break, for publishers especially, in the programmatic ecosystem, is the arrival of the header bidding technology. It does more than compliment waterfall; it stands alone as a subject. However, leveraging the header bidding model successfully requires putting in the necessary work.
Header bidding, and its varieties known as parallel bidding or advanced bidding, offer publishers a model to request bids simultaneously from as many demand sources or ad networks as possible. They also get to see the highest bidders and decide who buys the ad impression in the end. Sometimes, the highest bidder – e.g a particular ad network offers a price higher than the price floor stipulated by the publisher – yielding added revenue. The benefits of this simultaneous auction method outweigh that of the waterfall architecture. For publishers, they can make more money from running ads on their apps and websites while the advertisers get equal opportunities to bid for available inventory and secure the best ad spaces or inventory depending on their bid price.
But this method has some drawbacks, which is mainly the increased cost of infrastructure to implement header bidding successfully – publishers have to do more setup work. Publishers will need to – in addition to their SSP – install suitable software development kits (SDK) or header bidding adapters in their apps to be able to run the ads from advertisers.
First, they’ll need to set up a multi-line-item order on the ad server, and then integrate header bidding adapters for the client-side header bidding. In the case of in-app header bidding, an additional SDK will need to be integrated besides the Primary Ad Server SDK. For server-side header bidding, publishers will have to either host their own or choose a hosted server like that of Prebid Server. A Prebid Server improves page loading speed by running the header bidding auction on a separate server whereas in traditional header bidding it is run on the client.
Ad latency affects user experience and is not good for business. In this whole process, one thing that cannot be compromised is UX. This is all aimed at maximizing yield while improving the users’ experience.
Seeing that waterfall fails in maximizing yields from ad requests which leaves ad publishers feeling shortchanged, the ad selling process can, therefore, be complemented or substituted by introducing header bidding into the mix.
Hence, instead of only obtaining bids via the flawed waterfall method, publishers can also interact with ad networks through header bidding to get the best bids for every inventory served.
So, publishers integrate header bidding into the waterfalls – mainly because they have an existing waterfall and also want to increase their inventory fill rates by augmenting the waterfall with header bidding. In addition to this, the inventory is more likely to be sold for a better price.
While the waterfall structure is used to sell ads sequentially to advertisers, the header bidding model is simultaneously deployed to make the available impressions available to as many buyers as possible, increasing the chances of optimizing the yield from ad inventory sales.
To achieve this, a multi-line-item order is set up on an ad server. Thus a header bidding auction is plugged in as an ad source at multiple positions within the waterfall. Each position is assigned a price point – each line item corresponds to a particular price point and acts as a separate ad source in the waterfall. This is the price in which a winning bid in the header bidding auction can be equal to. It is sent to the ad server in the form of a keyword that targets a particular line item to participate in the waterfall.
Thus given the additional sources when waterfall is traversed as usual, there is a higher chance to fill. In the end, publishers are able to maximize revenues while also filling their impressions optimally.
This strategy has been the mainstay in the monetization of in-app header bidding. Although the industry is getting more dynamic and sophisticated, it is a widely adopted strategy for yield optimization.
Today, digital advertising is a highly versatile and competitive market. The technological development in the advertising ecosystem can be complex and confusing. Although header bidding supersedes the waterfall method, the market isn’t ready for pure parallel bidding right now. All leading companies prefer a hybrid method that combines waterfall with header bidding.
Since programmatic ad buying is majorly automated, key players need to understand the nuances of the system. This way, yields can be optimized and more sales generated.