How Server-Side Header Bidding Can Elevate Your Programmatic Strategy
Like so much in the programmatic ecosystem, header bidding continues to evolve. Today publishers are beginning to explore server-side header bidding, which promises to reduce page latency while optimizing programmatic revenue. This post will discuss this new form of header bidding, how it differs from older programmatic methods, and how this approach might evolve in the near future.
In an earlier article, I discussed some of the technologies required for a publisher to build a profitable programmatic ad business.
As mentioned then, two of the primary mechanisms for optimizing programmatic yield are:
- Upgrading as much inventory as possible out of Open Exchanges and into Private Marketplaces (PMPs)
- Implementing a “super auction” across all programmatic inventory utilizing what is today called “header bidding” mediation
Focusing on the latter, let’s first explore why header bidding is superior to the traditional “waterfall” approach of optimizing across all of your programmatic demand partners (Supply-Side Platforms or SSPs).
What Is Header Bidding
As mentioned in my previous post, with a waterfall, publishers manually dictate the competition for each of their programmatic ad impressions. In the example below, let’s assume that, based on in-depth and expert analysis by your yield optimization team, you are confident that your demand partners are properly prioritized in the order below (#1 = best performer or top priority). For most ad impressions, this will provide optimal yield as Partner #1 will typically produce a higher bid than all of the other three SSPs. However, for the ad impressions where this is not the case, you will be leaving money on the table.
In this simple example, where you have set a $2 CPM floor for this given set of inventory, Partner #1 will pass because its highest bid is only a $1.50 CPM. Partner #2, though, will win the impression in a Waterfall as, at $2.20, it does have a bid that is greater than $2. Meanwhile, Partner #3 and Partner #4 never even get a chance to submit their highest bid as someone above them in the Waterfall, Partner #2, already filled the impression. In a “super auction,” though, all Partners get to submit their highest bids, and Partner #3 would win with a $3 CPM. As a revenue-obsessed publisher, would you rather get a $2.20 or a $3 CPM? If, like me, you prefer the latter, then implement header bidding.
Client-Side vs. Server-Side Header Bidding
As also mentioned earlier, “client-side” was initially the more popular technical implementation of header bidding. Here, every SSP, including the mediator and all of those being mediated, are required to place a tag on your pages potentially causing significant page and ad load latency (it’s worse in a mobile app environment as the SSPs would all want to install a Software Development Kit or SDK). In addition to increasing consumer frustration with digital ads, these performance problems extend directly into programmatic revenue concerns, as, for many pageviews, a few of your mediated SSPs will timeout and not get to submit bids.
These performance and yield concerns have accelerated the move to “server-side” header bidding mediation which only requires a tag from the primary partner. Once called, this primary mediator will retrieve bids from all other demand sources via technology running on the mediator’s servers. The bid with the highest CPM across all of these SSPs is then returned as the winner.
Based on our experiences with our publisher clients and conversations with others in the industry, this practice effectively addresses the latency issue. However, when you move to server-side header bidding, your non-primary partners are likely to call foul. These other demand partners are suspicious of the mediator biasing the auction in the mediator’s favor. Additionally, without a tag on the page, non-mediators are hit with “ID leakage” limiting their ability to match audiences, which is a key factor in determining bid rates.
To address this concern major technology players in the space are looking at a shared identification across all ecosystem players. We expect to hear more on shared identification in the next year.
As always seems to be the case with ad technologies, there are complexities and then trade-offs in every effort you take to simplify and optimize your ad tech stack. Header bidding mediation, though, is a proven winner: despite some trade-offs, server-side deader bidding is currently the best implementation.
The effect of page latency and user experience is difficult to factor into simple equations, but my earlier math does not even take into account revenue sacrificed due to high bounce rates and the loss of repeat users (also, eliminating complaints from your peers on the editorial side will likely help you sleep better at night). Thus, until the next best thing comes along (and, in ad tech, it always does), testing server-side header bidding mediation is worthy of consideration. With attention being paid to the key concerns with server-side header bidding and the results we have witnessed firsthand with our publisher clients, we only expect its appeal to grow.